Index _ | A | B | C | D | E | F | G | H | I | K | L | M | N | O | P | Q | R | S | T | U | V | W | X | Z _ __aenter__() (nemo_rl.experience.rollout_manager._Deadline method) __aexit__() (nemo_rl.experience.rollout_manager._Deadline method) __all__ (in module nemo_rl.algorithms.async_utils) (in module nemo_rl.algorithms.loss) (in module nemo_rl.algorithms.single_controller_utils) (in module nemo_rl.algorithms.x_token) (in module nemo_rl.data.datasets) (in module nemo_rl.data.datasets.eval_datasets) (in module nemo_rl.data.datasets.preference_datasets) (in module nemo_rl.data.datasets.response_datasets) (in module nemo_rl.data.packing) (in module nemo_rl.data_plane) (in module nemo_rl.models.generation.dynamo) (in module nemo_rl.models.generation.megatron) (in module nemo_rl.models.generation.vllm) (in module nemo_rl.models.megatron.draft) (in module nemo_rl.models.value) (in module nemo_rl.telemetry) (in module nemo_rl.telemetry.instrumentation) (in module nemo_rl.weight_sync) (in module nemo_rl.weight_sync.xferdtensor_python) __call__() (nemo_rl.algorithms.loss.interfaces.LossFunction method) (nemo_rl.algorithms.loss.loss_functions.ClippedPGLossFn method) (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossFn method) (nemo_rl.algorithms.loss.loss_functions.DistillationLossFn method) (nemo_rl.algorithms.loss.loss_functions.DPOLossFn method) (nemo_rl.algorithms.loss.loss_functions.DraftCrossEntropyLossFn method) (nemo_rl.algorithms.loss.loss_functions.MseValueLossFn method) (nemo_rl.algorithms.loss.loss_functions.NLLLossFn method) (nemo_rl.algorithms.loss.loss_functions.PreferenceLossFn method) (nemo_rl.algorithms.loss.wrapper.DraftLossWrapper method) (nemo_rl.algorithms.loss.wrapper.SequencePackingFusionLossWrapper method) (nemo_rl.algorithms.loss.wrapper.SequencePackingLossWrapper method) (nemo_rl.data.cross_tokenizer_collate.CrossTokenizerCollator method) (nemo_rl.data.interfaces.TaskDataPreProcessFnCallable method) (nemo_rl.data.interfaces.TaskDataProcessFnCallable method) (nemo_rl.distributed.worker_groups.RayWorkerBuilder method) (nemo_rl.models.automodel.train.FullLogitsPostProcessor method) (nemo_rl.models.automodel.train.LogprobsPostProcessor method) (nemo_rl.models.automodel.train.LossPostProcessor method) (nemo_rl.models.automodel.train.ScorePostProcessor method) (nemo_rl.models.automodel.train.TopkLogitsPostProcessor method) (nemo_rl.models.megatron.train.LogprobsPostProcessor method) (nemo_rl.models.megatron.train.LossPostProcessor method) (nemo_rl.models.megatron.train.TopkLogitsPostProcessor method) (nemo_rl.models.value.workers.dtensor_value_worker_v2.RightShiftLossWrapper method) __contact_emails__ (in module nemo_rl.package_info) __contact_names__ (in module nemo_rl.package_info) __deepcopy__() (nemo_rl.data.multimodal_utils.PackedTensor method) __del__() (nemo_rl.distributed.virtual_cluster.RayVirtualCluster method) (nemo_rl.environments.reward_model_environment.RewardModelEnvironment method) (nemo_rl.models.generation.megatron.megatron_generation.MegatronGeneration method) (nemo_rl.models.generation.trtllm.trtllm_generation.TrtllmGeneration method) (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) (nemo_rl.models.policy.lm_policy.Policy method) (nemo_rl.models.policy.teacher_worker_group.TeacherWorkerGroup method) (nemo_rl.models.value.lm_value.Value method) (nemo_rl.utils.logger.Logger method) (nemo_rl.utils.logger.MLflowLogger method) __description__ (in module nemo_rl.package_info) __download_url__ (in module nemo_rl.package_info) __enter__() (nemo_rl.distributed.refit_watchdog.RefitAbortWatchdog method) (nemo_rl.utils.checkpoint.CheckpointManager method) __eq__() (nemo_rl.distributed.named_sharding.NamedSharding method) __exit__() (nemo_rl.distributed.refit_watchdog.RefitAbortWatchdog method) (nemo_rl.utils.checkpoint.CheckpointManager method) __extra__ (nemo_rl.algorithms.loss.loss_functions.ClippedPGLossDataDict attribute) (nemo_rl.data.interfaces.DatumSpec attribute) (nemo_rl.models.generation.interfaces.GenerationDatumSpec attribute) (nemo_rl.models.generation.interfaces.GenerationOutputSpec attribute) __getattr__() (nemo_rl.models.value.workers.dtensor_value_worker_v2.RightShiftLossWrapper method) __getitem__() (nemo_rl.data.datasets.processed_dataset.AllTaskProcessedDataset method) (nemo_rl.data.datasets.response_datasets.oai_format_dataset.PreservingDataset method) (nemo_rl.modelopt.models.policy.workers.utils._DictDataset method) __getstate__() (nemo_rl.models.generation.dynamo.dynamo_generation.DynamoGeneration method) __homepage__ (in module nemo_rl.package_info) __iter__() (nemo_rl.data.dataloader.CyclingDataLoader method) (nemo_rl.data.dataloader.MultipleDataloaderWrapper method) (nemo_rl.data.datasets.response_datasets.oai_format_dataset.PreservingDataset method) __keywords__ (in module nemo_rl.package_info) __len__() (nemo_rl.algorithms.async_utils.replay_buffer.TQReplayBuffer method) (nemo_rl.data.datasets.processed_dataset.AllTaskProcessedDataset method) (nemo_rl.data.datasets.response_datasets.oai_format_dataset.PreservingDataset method) (nemo_rl.data.multimodal_utils.PackedTensor method) (nemo_rl.experience.rollout_recovery.RolloutRecoveryLedger method) (nemo_rl.modelopt.models.policy.workers.utils._DictDataset method) __license__ (in module nemo_rl.package_info) __new__() (nemo_rl.models.generation.vllm.vllm_backend.NixlVllmWorker method) __next__() (nemo_rl.data.dataloader.MultipleDataloaderWrapper method) __package_name__ (in module nemo_rl.package_info) __post_init__() (nemo_rl.data.interfaces.TaskDataSpec method) (nemo_rl.data_plane.interfaces.KVBatchMeta method) (nemo_rl.experience.rollout_manager.RolloutRetryPolicy method) (nemo_rl.models.generation.fleet_health.FleetHealthPolicy method) __repository_url__ (in module nemo_rl.package_info) __repr__() (nemo_rl.distributed.named_sharding.NamedSharding method) (nemo_rl.models.generation.trtllm.trtllm_worker_async.TrtllmAsyncGenerationWorkerImpl method) (nemo_rl.models.generation.vllm.vllm_worker.BaseVllmGenerationWorker method) (nemo_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) (nemo_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl method) (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) (nemo_rl.models.value.workers.dtensor_value_worker_v2.DTensorValueWorkerV2Impl method) (nemo_rl.models.value.workers.megatron_value_worker.MegatronValueWorkerImpl method) __setstate__() (nemo_rl.data.multimodal_utils.PackedTensor method) (nemo_rl.models.generation.dynamo.dynamo_generation.DynamoGeneration method) __shortversion__ (in module nemo_rl.package_info) __slots__ (nemo_rl.algorithms.async_utils.replay_buffer.DataPlaneMutationCut attribute) __torch_dispatch__() (nemo_rl.models.generation.vllm.vllm_sparse_delta._SparseWeightLoadMode method) __version__ (in module nemo_rl.package_info) _abort_stale_inflight() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _abort_subcommunicator() (in module nemo_rl.weight_sync.xferdtensor_python) _Abortable (class in nemo_rl.distributed.refit_watchdog) _ABSENT_STATES (in module nemo_rl.models.generation.fleet_health) _accepted_forward_kwargs() (in module nemo_rl.models.automodel.data) _ACK (in module nemo_rl.utils.weight_transfer_zmq) _active_replica_signature() (in module nemo_rl.weight_sync.xferdtensor_python) _add_multimodal_generation_payload() (in module nemo_rl.experience.rollouts) _add_noise_to_weights() (nemo_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) (nemo_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl method) _add_r3_fallback_metrics() (in module nemo_rl.experience.rollouts) _add_verification_samples() (nemo_rl.utils.weight_transfer_sparse_codec.DeltaCompressionTracker method) _adjust_bin_count() (nemo_rl.data.packing.algorithms.SequencePacker method) _admit_reserved_prompt_groups() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _advantage_input_fields() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _advantage_stage() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _aggregate_megatron_flops_metrics() (in module nemo_rl.models.policy.lm_policy) _aggregate_multi_turn_rollout_metrics() (in module nemo_rl.experience.rollouts) _aggregate_rollout_metrics() (nemo_rl.experience.rollout_manager.AsyncRolloutImpl method) _aggregate_train_results() (in module nemo_rl.models.policy.tq_policy) (in module nemo_rl.models.value.tq_value) _AIOHTTP_INFRA_TYPES (in module nemo_rl.experience.failures) _align_dp() (nemo_rl.algorithms.x_token.token_aligner.TokenAligner static method) _align_single() (nemo_rl.algorithms.x_token.token_aligner.TokenAligner method) _align_with_anchors() (nemo_rl.algorithms.x_token.token_aligner.TokenAligner method) _alignment_mask() (nemo_rl.algorithms.x_token.token_aligner.TokenAligner static method) _all_gather_tp_shards() (in module nemo_rl.models.megatron.draft.utils) _all_image_sizes_equal() (in module nemo_rl.models.automodel.data) _all_sleep_tags() (nemo_rl.models.generation.trtllm.trtllm_worker_async.TrtllmAsyncGenerationWorkerImpl class method) _allgather_destination() (in module nemo_rl.weight_sync.xferdtensor_python) _allocate_transfer_buffer() (nemo_rl.utils.checkpoint_engines.nixl.NIXLCheckpointEngine method) _allowed_new_tokens() (nemo_rl.models.generation.dynamo.dynamo_generation.DynamoGeneration method) _AllReduceSum (class in nemo_rl.distributed.model_utils) _always_export() (in module nemo_rl.telemetry.setup) _apply() (nemo_rl.models.dtensor.parallelize.ColwiseParallelWithGather method) _apply_chat_template() (in module nemo_rl.models.generation.dynamo.token_wrapper) _apply_configured_message_level_advantage_penalties() (in module nemo_rl.algorithms.grpo) _apply_decoded_items() (nemo_rl.models.generation.vllm.vllm_sparse_delta.VllmSparseDeltaApplier method) _apply_dynamic_sampling() (in module nemo_rl.algorithms.grpo_sync) _apply_effort_shaping() (in module nemo_rl.experience.rollouts) _apply_mask_sample_filter() (in module nemo_rl.algorithms.grpo) _apply_message_level_advantage_penalties() (in module nemo_rl.algorithms.grpo) _apply_moe_config() (in module nemo_rl.models.megatron.setup) _apply_mtp_config() (in module nemo_rl.models.megatron.setup) _apply_packing_prep() (nemo_rl.data_plane.worker_mixin.TQWorkerMixin method) _apply_parallelism_config() (in module nemo_rl.models.megatron.setup) _apply_performance_config() (in module nemo_rl.models.megatron.setup) _apply_ppo_seq_logprob_error_masking() (in module nemo_rl.algorithms.ppo) _apply_precision_config() (in module nemo_rl.models.megatron.setup) _apply_s3_manifest_payload() (nemo_rl.models.generation.vllm.vllm_sparse_refit.VllmSparseRefitReceiver method) _apply_state_dict_to_model() (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) _apply_temperature_scaling() (nemo_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) _apply_top_k_only_fn() (in module nemo_rl.algorithms.logits_sampling_utils) _apply_top_k_top_p_filtering() (nemo_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) _apply_top_k_top_p_fn() (in module nemo_rl.algorithms.logits_sampling_utils) _apply_vllm_patches() (in module nemo_rl.models.generation.vllm.patches) _apply_zmq_payload() (nemo_rl.models.generation.vllm.vllm_sparse_refit.VllmSparseRefitReceiver method) _ApplyTopKTopP (class in nemo_rl.algorithms.logits_sampling_utils) _ArgvBuilder (class in nemo_rl.models.generation.dynamo.arguments) _ARMED (in module nemo_rl.distributed.refit_watchdog) _ARMED_LOCK (in module nemo_rl.distributed.refit_watchdog) _as_routed_experts_tensor() (in module nemo_rl.models.generation.vllm.utils) _assemble_local_states() (nemo_rl.models.megatron.draft.hidden_capture.HiddenStateCapture method) _assert_no_key_loss() (in module nemo_rl.data_plane.adapters.transfer_queue) _assert_response_within_context() (nemo_rl.models.generation.dynamo.dynamo_generation.DynamoGeneration method) _assert_services_alive() (nemo_rl.models.generation.dynamo.managed_runtime.ManagedDynamoRuntime method) _assert_step_open() (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) _assign_optional_layer_weight() (in module nemo_rl.models.megatron.draft.utils) _async_checkpoint_cuda_cache_active (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl attribute) _async_generate_base() (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) _async_ppo_buffer_max_age() (in module nemo_rl.algorithms.ppo) _async_ppo_generation_lead_steps() (in module nemo_rl.algorithms.ppo) _attach_context_parallel_hooks() (in module nemo_rl.models.policy.workers.dtensor_policy_worker) _attach_input_quantizer_amax_loaders() (nemo_rl.modelopt.models.generation.vllm_quant_backend.VllmQuantInternalWorkerExtension method) _attach_multimodal_data_to_user_message() (in module nemo_rl.environments.nemo_gym) _attach_or_repack_pack_metadata() (nemo_rl.data_plane.worker_mixin.TQWorkerMixin method) _attach_routed_experts_to_message_log_prefix() (in module nemo_rl.experience.rollouts) _autocast_context() (nemo_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl method) (nemo_rl.models.value.workers.dtensor_value_worker_v2.DTensorValueWorkerV2Impl method) _AUX_LOSS_TRACK_NAMES (in module nemo_rl.models.megatron.common) _averaged_logits_kd() (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossFn method) _BACKEND_MODELS (in module nemo_rl.data_plane.interfaces) _baseline() (nemo_rl.utils.weight_transfer_sparse_codec.DeltaCompressionTracker method) _batch_fused_modelopt_moe_weights() (in module nemo_rl.modelopt.models.generation.vllm_quant_backend) _BEHAVIORAL_DATASET_CONFIG_KEYS (in module nemo_rl.data.datasets.utils) _bind_socket_in_range() (in module nemo_rl.distributed.virtual_cluster) _BRIDGE_SIGNAL_HANDLER_PATCHED (in module nemo_rl.models.megatron.setup) _broadcast_batched_data_dict() (in module nemo_rl.data_plane.worker_mixin) _broadcast_destination() (in module nemo_rl.weight_sync.xferdtensor_python) _broadcast_misc_params_packed() (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) _broadcast_weights_for_collective() (nemo_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) (nemo_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl method) (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) _BUCKET_OVERRIDE (in module nemo_rl.telemetry.instrumentation) _bucket_size_bytes (nemo_rl.weight_sync.checkpoint_engine_weight_synchronizer.CheckpointEngineWeightSynchronizer attribute) _build() (nemo_rl.weight_sync.nccl_reshard_weight_synchronizer.NcclReshardWeightSynchronizer method) _build_advantage_estimator() (in module nemo_rl.algorithms.single_controller_utils.setup) _build_async_grpo_train_data() (in module nemo_rl.algorithms.grpo) _build_clusters() (in module nemo_rl.algorithms.single_controller_utils.setup) _build_colocated_inference_model() (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) _build_completion_request() (nemo_rl.models.generation.dynamo.dynamo_generation.DynamoGeneration method) _build_exact_plan() (in module nemo_rl.weight_sync.xferdtensor_python) _build_expert_groups() (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) _build_generation() (in module nemo_rl.algorithms.single_controller_utils.setup) _build_hf_to_gen_backend_mapping() (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) _build_inputs() (nemo_rl.experience.rollout_manager.AsyncNemoGymRolloutImpl method) _build_layer_to_pp_stage() (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) _build_model_batch() (in module nemo_rl.models.automodel.train) _build_pair_strings() (nemo_rl.algorithms.x_token.token_aligner.TokenAligner static method) _build_prompt_token_ids() (in module nemo_rl.models.generation.trtllm.trtllm_http_server) _build_reasoning_parser() (in module nemo_rl.models.generation.trtllm.trtllm_http_server) _build_refit_conversion_tasks() (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) _build_resource_attributes() (in module nemo_rl.telemetry.setup) _build_retry_policy() (in module nemo_rl.algorithms.single_controller_utils.setup) _build_rollouts_state() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector method) _build_sampling_params() (in module nemo_rl.models.generation.trtllm.trtllm_http_server) (nemo_rl.models.generation.megatron.megatron_worker.MegatronGenerationMixin method) (nemo_rl.models.generation.trtllm.trtllm_worker_async.TrtllmAsyncGenerationWorkerImpl method) (nemo_rl.models.generation.vllm.vllm_worker.BaseVllmGenerationWorker method) _build_split_axis_by_parameter() (in module nemo_rl.models.megatron.draft.utils) _build_striped_targets() (in module nemo_rl.weight_sync.xferdtensor_python) _build_striped_transfers() (in module nemo_rl.weight_sync.xferdtensor_python) _build_task_index_map() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector static method) _build_token_level_rewards() (nemo_rl.algorithms.advantage_estimator.GeneralizedAdvantageEstimator method) _build_tool_parser() (in module nemo_rl.models.generation.trtllm.trtllm_http_server) _build_trainer() (in module nemo_rl.algorithms.single_controller_utils.setup) _build_value() (in module nemo_rl.algorithms.single_controller_utils.setup) _build_video_metadata() (in module nemo_rl.models.generation.vllm.video_utils) _CACHED_VIDEO_FRAME_MANIFEST_MAGIC (in module nemo_rl.models.generation.vllm.video_utils) _CACHED_VIDEO_FRAME_MANIFEST_MIME (in module nemo_rl.models.generation.vllm.video_utils) _calculate_preference_score() (nemo_rl.environments.code_jaccard_environment.CodeJaccardVerifyWorker method) _calculate_refit_param_info() (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) _calculate_target_weights() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector method) _call_dp() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _call_model_loader_hook_if_available() (in module nemo_rl.models.generation.trtllm.trtllm_backend) _canonical_manifest_value() (in module nemo_rl.algorithms.async_utils.replay_buffer) _canonicalize_hf_config_overrides() (in module nemo_rl.models.megatron.setup) _canonicalize_nvfp4_scale_() (in module nemo_rl.modelopt.models.generation.vllm_modelopt) _canonicalize_sequence() (in module nemo_rl.algorithms.x_token.token_aligner) _capture_dataloader_state() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector method) _causal_topk_pairs() (in module nemo_rl.utils.flops_formulas) _chat_template_kwargs() (in module nemo_rl.models.generation.dynamo.token_wrapper) _chat_template_kwargs_for_processor() (in module nemo_rl.environments.nemo_gym_video) _check_connect_timeout_fits() (nemo_rl.algorithms.single_controller_utils.config.GenerationRouterConfig method) _check_consistent() (nemo_rl.algorithms.single_controller_utils.config.FleetHealthConfig method) (nemo_rl.algorithms.single_controller_utils.config.RolloutFailureConfig method) (nemo_rl.algorithms.single_controller_utils.config.WatchdogConfig method) _check_container_fingerprint() (in module nemo_rl) _check_env_health() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _check_port_range() (nemo_rl.algorithms.single_controller_utils.config.GenerationRouterConfig method) _check_router_deadline_fits_inside_the_rollout() (nemo_rl.algorithms.single_controller_utils.config.AsyncRLConfig method) _check_stall_watchdog_outlasts_rollouts() (nemo_rl.algorithms.single_controller_utils.config.AsyncRLConfig method) _check_status_is_not_retried_by_gym() (nemo_rl.algorithms.single_controller_utils.config.GenerationRouterConfig method) _check_weight_sync_results() (in module nemo_rl.models.policy.utils) _CHECKPOINT_CANDIDATE_NAMES (in module nemo_rl.models.megatron.draft.utils) _checkpoint_engine_config (nemo_rl.weight_sync.checkpoint_engine_weight_synchronizer.CheckpointEngineWeightSynchronizer attribute) _checkpoint_engine_params() (nemo_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) (nemo_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl method) _checkpoint_engine_ready (nemo_rl.weight_sync.checkpoint_engine_weight_synchronizer.CheckpointEngineWeightSynchronizer attribute) _checkpoint_engine_weight_iterator() (nemo_rl.models.policy.workers.checkpoint_engine.DTensorCheckpointEngineSendMixin method) (nemo_rl.models.policy.workers.checkpoint_engine.MegatronCheckpointEngineSendMixin method) (nemo_rl.models.policy.workers.checkpoint_engine.PolicyCheckpointEngineMixin method) _checkpoint_engine_weight_layout() (nemo_rl.models.generation.vllm.refit_loader.VllmShardedExpertRefitMixin method) _CHECKPOINT_LAYER_KEY_PATTERN (in module nemo_rl.models.megatron.draft.utils) _CHECKPOINTABLE_BACKENDS (in module nemo_rl.data_plane.interfaces) _clamp_max_num_steps() (in module nemo_rl.algorithms.single_controller_utils.setup) _classify_generation_failure() (in module nemo_rl.experience.rollout_manager) _classify_items() (nemo_rl.data.packing.algorithms.ModifiedFirstFitDecreasingPacker method) _cleanup_finished_threads() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector method) _clear_data_plane_samples() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _clear_fp8_caches() (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) _clear_samples_unlocked() (nemo_rl.algorithms.async_utils.replay_buffer.TQReplayBuffer method) _CLIENT_RESPONSE_ERROR (in module nemo_rl.experience.failures) _clip_grpo_advantages() (in module nemo_rl.algorithms.grpo) _CODE_TO_DTYPE (in module nemo_rl.models.megatron.draft.hidden_capture) _coerce_logprob_list() (in module nemo_rl.models.generation.dynamo.token_wrapper) _coerce_to_scalar() (nemo_rl.utils.logger.TensorboardLogger static method) _coerce_token_id_list() (in module nemo_rl.models.generation.dynamo.token_wrapper) _collect() (nemo_rl.utils.logger.RayGpuMonitorLogger method) _collect_gpu_sku() (nemo_rl.utils.logger.RayGpuMonitorLogger method) _collect_metrics() (nemo_rl.utils.logger.RayGpuMonitorLogger method) _collect_mtp_hf_layer_names() (in module nemo_rl.models.policy.workers.megatron_policy_worker) _collect_mtp_metrics() (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) _collect_refit_apply_results() (nemo_rl.models.generation.vllm.vllm_sparse_refit.VllmSparseRefitReceiver method) _collect_reserved_urls() (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) _collect_rollout_batch() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector method) _collection_loop() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector method) (nemo_rl.utils.logger.RayGpuMonitorLogger method) _colocated_reshard_plan (nemo_rl.models.generation.megatron.megatron_worker.MegatronGenerationMixin attribute) _combine_consecutive_misaligned() (nemo_rl.algorithms.x_token.token_aligner.TokenAligner static method) _combine_or_shard_weight_parts() (in module nemo_rl.models.megatron.draft.utils) _commit_admission() (nemo_rl.algorithms.async_utils.staleness_sampler._GatedSampler method) _commit_baseline_updates() (nemo_rl.utils.weight_transfer_sparse_codec.DeltaCompressionTracker method) _CompletedNemoGymGroup (class in nemo_rl.experience.rollouts) _completion_url() (nemo_rl.models.generation.dynamo.dynamo_generation.DynamoGeneration method) _compute_buffer_size() (nemo_rl.weight_sync.ipc_weight_synchronizer.IPCWeightSynchronizer method) (nemo_rl.weight_sync.sglang_weight_synchronizer._SGLangWeightSynchronizer method) _compute_ce() (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossFn method) _compute_critic_metrics() (in module nemo_rl.algorithms.ppo) _compute_distributed_log_softmax() (in module nemo_rl.distributed.model_utils) _compute_distributed_log_softmax_with_grad() (in module nemo_rl.distributed.model_utils) _compute_distributed_softmax() (in module nemo_rl.distributed.model_utils) _compute_dynamic_prompt_length() (in module nemo_rl.environments.nemo_gym_video) _compute_dynamic_weights() (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossFn method) _compute_gae() (nemo_rl.algorithms.advantage_estimator.GeneralizedAdvantageEstimator method) _compute_gold() (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossFn method) _compute_layer_owner_map() (nemo_rl.models.megatron.draft.hidden_capture.HiddenStateCapture method) _compute_local_layer_mapping() (nemo_rl.models.megatron.draft.hidden_capture.HiddenStateCapture method) _compute_local_logprobs() (nemo_rl.models.automodel.train.LogprobsPostProcessor method) _compute_metrics() (nemo_rl.algorithms.advantage_estimator.OPDAdvantageEstimator method) _compute_moe_grad_scale() (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) _compute_p_kl() (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossFn method) _compute_rollout_metrics() (nemo_rl.experience.rollout_manager.AsyncNemoGymRolloutImpl method) _compute_same_vocab_kl() (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossFn method) _compute_seq_logprob_error_metrics() (in module nemo_rl.algorithms.grpo_sync) _compute_shard_slices() (in module nemo_rl.weight_sync.xferdtensor) (in module nemo_rl.weight_sync.xferdtensor_python) _compute_splice_inputs() (in module nemo_rl.models.generation.trtllm.trtllm_http_server) _compute_teacher_kd() (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossFn method) _compute_teacher_logprobs() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector method) _compute_video_timestamps() (in module nemo_rl.models.generation.vllm.video_utils) _config_to_env() (in module nemo_rl.telemetry.setup) _configure_quant_engine_kwargs() (in module nemo_rl.modelopt.models.generation.vllm_quant_worker) _configure_socket() (in module nemo_rl.utils.weight_transfer_zmq) _connect() (nemo_rl.weight_sync.sglang_weight_synchronizer._SGLangWeightSynchronizer method) (nemo_rl.weight_sync.sglang_weight_synchronizer.SGLangColocatedWeightSynchronizer method) (nemo_rl.weight_sync.sglang_weight_synchronizer.SGLangDisaggregatedWeightSynchronizer method) _connect_existing() (in module nemo_rl.data_plane.adapters.transfer_queue) _construct_multichoice_prompt() (in module nemo_rl.data.processors) _contains_post_write_enrichment_error() (in module nemo_rl.experience.rollout_manager) _context (in module nemo_rl.utils.r3_trace) _context_capped_max_new_tokens() (in module nemo_rl.models.generation.vllm.vllm_worker) _copy_main_params_to_param_buffer() (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) _count_for_target() (nemo_rl.algorithms.async_utils.replay_buffer.ReplayBufferImpl method) _COUPLED_MULTIMODAL_KEYS (in module nemo_rl.distributed.batched_data_dict) _cp_gather_logits() (in module nemo_rl.models.automodel.train) _create_advantage_estimator() (in module nemo_rl.algorithms.grpo) (in module nemo_rl.algorithms.ppo) _create_checkpoint_config() (in module nemo_rl.models.megatron.setup) _create_draft_pre_wrap_hook() (in module nemo_rl.models.megatron.setup) _create_engine() (nemo_rl.modelopt.models.generation.vllm_quant_worker.VllmQuantAsyncGenerationWorker method) (nemo_rl.modelopt.models.generation.vllm_quant_worker.VllmQuantGenerationWorker method) (nemo_rl.models.generation.vllm.vllm_worker.VllmGenerationWorkerImpl method) (nemo_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) _create_indexed_lengths() (nemo_rl.data.packing.algorithms.SequencePacker method) _create_megatron_config() (in module nemo_rl.models.megatron.setup) _create_nixl_agent() (in module nemo_rl.utils.checkpoint_engines.nixl) _create_placement_groups_internal() (nemo_rl.distributed.virtual_cluster.RayVirtualCluster method) _create_workers_from_bundle_indices() (nemo_rl.distributed.worker_groups.RayWorkerGroup method) _CREDENTIAL_FLAGS (in module nemo_rl.models.generation.dynamo.arguments) _credit_shortfall() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _current_context() (in module nemo_rl.utils.r3_trace) _custom_sampler_class() (in module nemo_rl.algorithms.async_utils.staleness_sampler) _DATA (in module nemo_rl.utils.weight_transfer_zmq) _datum_preprocessor() (nemo_rl.data.datasets.response_datasets.general_conversations_dataset.GeneralConversationsJsonlDataset class method) _Deadline (class in nemo_rl.experience.rollout_manager) _debug_payload_metrics (nemo_rl.models.generation.interfaces.GenerationConfig attribute) _decode_json_object() (in module nemo_rl.models.generation.dynamo.http_client) _decode_torchcodec_video() (in module nemo_rl.models.generation.vllm.video_utils) _deep_merge_dict() (in module nemo_rl.environments.nemo_gym_video) _DEFAULT_GROUP_BUCKET (in module nemo_rl.telemetry.instrumentation) _default_off_policy_distillation_save_state() (in module nemo_rl.algorithms.xtoken_off_policy_distillation) _default_ppo_save_state() (in module nemo_rl.algorithms.ppo) _DEFAULT_TRACE_DIR (in module nemo_rl.utils.r3_trace) _DEFAULT_TRACE_MICROBATCHES (in module nemo_rl.utils.r3_trace) _DEFAULT_TRACE_SAMPLES (in module nemo_rl.utils.r3_trace) _DEFAULT_TRACE_STEPS (in module nemo_rl.utils.r3_trace) _deinterleave_qkv() (in module nemo_rl.models.megatron.draft.utils) _derive_engine_gpu_offsets() (in module nemo_rl.models.policy.utils) _derive_required_prefix_token_ids() (in module nemo_rl.models.generation.dynamo.token_wrapper) _destination_buffer() (in module nemo_rl.weight_sync.xferdtensor_python) _destination_groups() (in module nemo_rl.weight_sync.xferdtensor_python) _destroy_subcommunicator() (in module nemo_rl.weight_sync.xferdtensor_python) _detach_pending_layerwise_weights() (in module nemo_rl.modelopt.models.generation.vllm_quant_backend) (in module nemo_rl.models.generation.vllm.vllm_backend) _detect_invalid_tool_call_and_malformed_thinking() (in module nemo_rl.environments.nemo_gym) _DictDataset (class in nemo_rl.modelopt.models.policy.workers.utils) _dig() (in module nemo_rl.telemetry.setup) _direct_full_vocab_kl() (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossFn method) _direct_topk_kl() (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossFn method) _disable_forward_pre_hook_until_next_train_step() (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) _disconnect_peers() (nemo_rl.utils.checkpoint_engines.nixl.NIXLCheckpointEngine method) _divert_batch_to_reserve() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _download_and_extract() (nemo_rl.data.datasets.response_datasets.intent.IntentDataset method) _dp_client (nemo_rl.data_plane.worker_mixin.TQWorkerMixin attribute) _dp_global_masked_mean() (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossFn method) _dpo_loss() (nemo_rl.algorithms.loss.loss_functions.DPOLossFn method) _drain_reserve_into_steps() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _drain_router_failures() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _drain_thread() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector method) _drop_nonclass_quant_registry_keys() (in module nemo_rl.modelopt.models.generation.vllm_quant_patch) _drop_padding() (nemo_rl.algorithms.x_token.token_aligner.TokenAligner static method) _DTYPE_TO_CODE (in module nemo_rl.models.megatron.draft.hidden_capture) _dummy_routed_experts_for_tokens() (in module nemo_rl.experience.rollouts) _eager_audio_probe() (nemo_rl.data.datasets.response_datasets.audiomcq.AudioMCQDataset method) _EagleLayerLayout (class in nemo_rl.models.megatron.draft.utils) _EagleModelLayout (class in nemo_rl.models.megatron.draft.utils) _EFFICIENCY_INSTRUMENTS (in module nemo_rl.telemetry.metrics) _EFFICIENCY_KEY_PREFIX (in module nemo_rl.telemetry.metrics) _EFFICIENCY_PCT_KEY (in module nemo_rl.telemetry.metrics) _EFFICIENCY_PCT_PER_STEP_KEY (in module nemo_rl.telemetry.metrics) _EFFICIENCY_SECONDS_SUFFIX (in module nemo_rl.telemetry.metrics) _effort_shaping_metrics() (in module nemo_rl.experience.rollouts) _EffortShapingMetrics (class in nemo_rl.experience.rollouts) _emit() (nemo_rl.data_plane.observability.MetricsDataPlaneClient method) _encode_explicit_locations() (in module nemo_rl.utils.weight_transfer_sparse_codec) _encode_single_image_source() (in module nemo_rl.data.multimodal_utils) _engine_already_imported() (in module nemo_rl.data_plane.adapters.transfer_queue_env) _ENGINE_MODULES (in module nemo_rl.data_plane.adapters.transfer_queue_env) _enqueue_rollout_group() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector method) _enqueue_sparse_payload_apply() (nemo_rl.models.generation.vllm.vllm_sparse_refit.VllmSparseRefitReceiver method) _enrich_sync() (nemo_rl.algorithms.opd.TQTeacherLogprobCoordinator method) _ensure_checkpoint_engine_ready() (nemo_rl.weight_sync.checkpoint_engine_weight_synchronizer.CheckpointEngineWeightSynchronizer method) _ensure_mmpr_cached() (in module nemo_rl.data.datasets.response_datasets.mmpr_tiny) _env_builder() (in module nemo_rl.utils.venvs) _ENV_FIELD_MAP (in module nemo_rl.telemetry.setup) _env_int() (in module nemo_rl.utils.r3_trace) _estimate_bins_needed() (nemo_rl.data.packing.algorithms.SequencePacker method) _estimate_refit_tensor_size_in_bytes() (in module nemo_rl.models.policy.workers.megatron_policy_worker) _event_counts (in module nemo_rl.utils.r3_trace) _evict_communicators() (in module nemo_rl.weight_sync.xferdtensor_python) _eviction_window() (nemo_rl.algorithms.async_utils.staleness_sampler._GatedSampler method) (nemo_rl.algorithms.async_utils.staleness_sampler.BaseSampler method) (nemo_rl.algorithms.async_utils.staleness_sampler.WindowedSampler method) _EXACT_TOKEN_MAP_CACHE (in module nemo_rl.algorithms.x_token.loss_utils) _exchange_exact_overlaps() (in module nemo_rl.weight_sync.xferdtensor_python) _executor() (in module nemo_rl.utils.weight_transfer_stream) _expand_nemotron_video_placeholders() (in module nemo_rl.environments.nemotron_utils) _expected_with_missing_route_fallback() (in module nemo_rl.utils.r3_trace) _export_layer_weights_to_hf() (in module nemo_rl.models.megatron.draft.utils) _EXTRA_ENV_VARS (in module nemo_rl.modelopt.models.generation.vllm_quant_worker) _extract_hash_answer() (in module nemo_rl.data.datasets.response_datasets.gsm8k) _extract_input_images_from_message() (in module nemo_rl.environments.nemo_gym) _extract_layer_name() (in module nemo_rl.weight_sync.nccl_reshard_utils) _extract_layer_prefix() (in module nemo_rl.weight_sync.nccl_reshard_utils) _extract_mask_sample_flags() (in module nemo_rl.experience.rollouts) _extract_static_video_messages() (in module nemo_rl.environments.nemo_gym_video) _extract_tensor_state_dict() (in module nemo_rl.models.megatron.draft.utils) _extract_videos_zip_once() (in module nemo_rl.data.datasets.response_datasets.intent) _EXTRACTION_SENTINEL (in module nemo_rl.data.datasets.response_datasets.intent) _fakequant_run_prolog_worker() (in module nemo_rl.modelopt.models.generation.vllm_quant_patch) _fanout() (nemo_rl.utils.weight_transfer_zmq.ZmqSparseRefitServer method) _fetch() (nemo_rl.data_plane.worker_mixin.TQWorkerMixin method) _fetch_and_parse_metrics() (nemo_rl.utils.logger.RayGpuMonitorLogger method) _fields_ (nemo_rl.distributed.stateless_process_group._VllmNcclUniqueId attribute) _fill_routed_experts_padding() (in module nemo_rl.models.megatron.data) _filter_covered_rows() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector method) _filter_records() (nemo_rl.data.datasets.response_datasets.intent.IntentDataset method) _finalize_checkpoint_engine_weight_send() (nemo_rl.models.policy.workers.checkpoint_engine.DTensorCheckpointEngineSendMixin method) (nemo_rl.models.policy.workers.checkpoint_engine.PolicyCheckpointEngineMixin method) _finalize_process_group() (in module nemo_rl.weight_sync.xferdtensor_python) _finalize_selection() (nemo_rl.algorithms.async_utils.staleness_sampler.BaseSampler method) _finalize_weight_update() (nemo_rl.models.generation.trtllm.trtllm_backend.NcclExtension method) _find_other_quant_checkpoint_caches() (in module nemo_rl.modelopt.models.policy.workers.megatron_quant_policy_worker) _find_routed_experts_template() (in module nemo_rl.experience.rollouts) _find_torchcodec_decodable_frame_count() (in module nemo_rl.models.generation.vllm.video_utils) _find_weight_quantizer() (nemo_rl.modelopt.models.policy.workers.megatron_quant_policy_worker.MegatronQuantPolicyWorker static method) _finish_deferred_generation() (in module nemo_rl.algorithms.single_controller_utils.setup) _finish_train_step_body() (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) _flag_for_key() (in module nemo_rl.models.generation.dynamo.arguments) _flatten_chunk() (nemo_rl.algorithms.x_token.token_aligner.TokenAligner static method) _flatten_mesh_ranks() (in module nemo_rl.weight_sync.xferdtensor) _flatten_metadata() (in module nemo_rl.weight_sync.checkpoint_engine_weight_synchronizer) _flatten_nemotron_video_frame_messages() (in module nemo_rl.environments.nemotron_utils) _flatten_rollout_message_log_for_tq() (in module nemo_rl.experience.sync_rollout_actor) _flush_queued_sparse_payloads() (nemo_rl.models.generation.vllm.vllm_sparse_refit.VllmSparseRefitReceiver method) _fmt() (nemo_rl.utils.timer.Timer method) _force_sync_optimizer_fp32_from_model() (in module nemo_rl.models.megatron.setup) _format_for_eval() (nemo_rl.data.datasets.eval_datasets.daily_omni.DailyOmniEvalDataset method) _format_options() (in module nemo_rl.data.datasets.response_datasets.intent) _format_refit_key_error() (in module nemo_rl.models.generation.vllm.vllm_backend) _forward() (nemo_rl.models.generation.generation_router.GenerationRouterImpl method) (nemo_rl.utils.weight_transfer_zmq.ZmqSparseRefitServer method) _forward_chat_completion() (nemo_rl.models.generation.dynamo.token_wrapper.DynamoTokenWrapperServer method) _forward_pad_seqlen() (nemo_rl.data_plane.worker_mixin.TQWorkerMixin method) _forward_pre_hook_enabled() (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) _from_wire() (in module nemo_rl.data_plane.adapters.transfer_queue) _frontend_env() (nemo_rl.models.generation.dynamo.managed_runtime.ManagedDynamoRuntime method) _FUSED_MODELOPT_MOE_SUFFIXES (in module nemo_rl.modelopt.models.generation.vllm_quant_backend) _GATE_POLL_SECONDS (in module nemo_rl.algorithms.async_utils.staleness_sampler) _gated_required_buffer_capacity() (in module nemo_rl.algorithms.async_utils.staleness_sampler) _GatedSampler (class in nemo_rl.algorithms.async_utils.staleness_sampler) _gather_cancelling_siblings() (in module nemo_rl.experience.rollout_manager) _gather_distributed() (nemo_rl.models.megatron.draft.hidden_capture.HiddenStateCapture method) _gather_tp_gate_up_weight() (in module nemo_rl.models.megatron.draft.utils) _gather_tp_qkv_weight() (in module nemo_rl.models.megatron.draft.utils) _gather_tp_weight_if_needed() (in module nemo_rl.models.megatron.draft.utils) _gen_fleet_probe_pump() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _gen_model() (nemo_rl.models.generation.megatron.megatron_worker.MegatronGenerationMixin method) _gen_parallelism() (nemo_rl.weight_sync.nccl_reshard_weight_synchronizer.NcclReshardWeightSynchronizer method) _generate_on_shard() (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) _generate_response() (nemo_rl.experience.rollout_manager.AsyncRolloutImpl method) _generate_texts() (in module nemo_rl.evals.eval) _generate_with_persistent_engine() (nemo_rl.models.generation.megatron.megatron_worker.MegatronGenerationMixin method) _generation (nemo_rl.weight_sync.checkpoint_engine_weight_synchronizer.CheckpointEngineWeightSynchronizer attribute) _generation_max_seq_len() (in module nemo_rl.algorithms.single_controller_utils.setup) _generation_rpc() (nemo_rl.weight_sync.checkpoint_engine_weight_synchronizer.CheckpointEngineWeightSynchronizer method) _get_byte_value() (in module nemo_rl.algorithms.x_token.token_aligner) _get_content_part_url() (in module nemo_rl.environments.nemo_gym_video) _get_distillation_save_state() (in module nemo_rl.algorithms.distillation) _get_dp_rank() (nemo_rl.models.automodel.checkpoint.AutomodelCheckpointManager method) _get_dpo_save_state() (in module nemo_rl.algorithms.dpo) _get_draft_output_layer() (in module nemo_rl.models.megatron.draft.utils) _get_draft_to_target_token_mapping() (in module nemo_rl.models.megatron.draft.utils) _get_drafter_model() (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) _get_efficiency_instruments() (in module nemo_rl.telemetry.metrics) _get_effort_config() (in module nemo_rl.algorithms.grpo) _get_enabled_input_amax() (nemo_rl.modelopt.models.policy.workers.megatron_quant_policy_worker.MegatronQuantPolicyWorker static method) _get_expert_tp_shard_dim() (in module nemo_rl.weight_sync.nccl_reshard_utils) _get_fp8_token_alignment() (in module nemo_rl.models.megatron.data) _get_free_consecutive_ports_local() (in module nemo_rl.distributed.virtual_cluster) _get_free_port_local() (in module nemo_rl.distributed.virtual_cluster) _get_glm_index_compute_layers() (in module nemo_rl.utils.flops_tracker) _get_glm_moe_layer_pattern() (in module nemo_rl.utils.flops_tracker) _get_gpu_id_info() (in module nemo_rl.distributed.virtual_cluster) _get_grpo_save_state() (in module nemo_rl.algorithms.grpo) _get_hf_config_overrides_hash() (in module nemo_rl.models.megatron.setup) _get_local_node_ip() (in module nemo_rl.data_plane.adapters.transfer_queue) _get_manifest_s3_store() (in module nemo_rl.utils.weight_transfer_stream) _get_mesh_coords() (in module nemo_rl.weight_sync.xferdtensor) _get_model_config() (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) _get_model_extra_state_dict() (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) _get_modelopt_reload_roots() (nemo_rl.modelopt.models.generation.vllm_quant_backend.VllmQuantInternalWorkerExtension method) _get_named_parameters() (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) _get_next_target_for_generation() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector method) _get_node_ip_and_free_port() (in module nemo_rl.distributed.virtual_cluster) _get_node_ip_local() (in module nemo_rl.distributed.virtual_cluster) _get_non_packed_sequence_pad_factor() (in module nemo_rl.models.megatron.data) _get_num_aux_hidden_states() (in module nemo_rl.models.megatron.draft.utils) _get_numa_node() (in module nemo_rl.distributed.numa_utils) _get_pack_sequence_parameters_for_megatron() (in module nemo_rl.models.megatron.data) _get_raw_spec_counters() (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) (nemo_rl.models.generation.vllm.vllm_worker.BaseVllmGenerationWorker method) _get_real_quant_mode() (nemo_rl.modelopt.models.policy.workers.megatron_quant_policy_worker.MegatronQuantPolicyWorker method) _get_replica_group() (nemo_rl.data_plane.worker_mixin.TQWorkerMixin method) (nemo_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) (nemo_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl method) (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) (nemo_rl.models.value.workers.megatron_value_worker.MegatronValueWorkerImpl method) _get_replica_subcommunicator() (in module nemo_rl.weight_sync.xferdtensor_python) _get_required_reward_penalty_token_id() (in module nemo_rl.experience.rollouts) _get_required_reward_penalty_token_ids() (in module nemo_rl.experience.rollouts) _get_reward_penalty_config_value() (in module nemo_rl.experience.rollouts) _get_reward_penalty_token_id() (in module nemo_rl.experience.rollouts) _get_reward_penalty_token_ids() (in module nemo_rl.experience.rollouts) _get_rm_save_state() (in module nemo_rl.algorithms.rm) _get_sft_save_state() (in module nemo_rl.algorithms.sft) _get_sorted_bundle_indices() (nemo_rl.distributed.virtual_cluster.RayVirtualCluster method) _get_sparse_delta_applier() (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) _get_tensor_meta() (in module nemo_rl.weight_sync.xferdtensor) _get_tensor_model_parallel_rank() (in module nemo_rl.models.megatron.router_replay) _get_tensor_model_parallel_world_size() (in module nemo_rl.models.megatron.router_replay) _get_tied_worker_bundle_indices() (nemo_rl.models.generation.trtllm.trtllm_generation.TrtllmGeneration method) (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) _get_tokens_on_this_cp_rank() (in module nemo_rl.distributed.model_utils) _get_tp_rank() (in module nemo_rl.models.megatron.draft.utils) (nemo_rl.models.automodel.checkpoint.AutomodelCheckpointManager method) _get_transformer_engine_file() (in module nemo_rl.models.policy.workers.patches) _get_vllm_file() (in module nemo_rl.models.generation.vllm.patches) _global_moe_layer_numbers() (in module nemo_rl.models.megatron.router_replay) _gpt_forward_with_linear_ce_fusion() (in module nemo_rl.distributed.model_utils) _group_experts() (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) _GYM_RETRY_STATUSES (in module nemo_rl.algorithms.single_controller_utils.config) _GYM_TOKEN_METADATA_FIELDS (in module nemo_rl.models.generation.dynamo.token_wrapper) _handle() (nemo_rl.models.generation.generation_router.GenerationRouterImpl method) _has_nan_generation_logprobs() (in module nemo_rl.environments.nemo_gym) _has_verifiable_answer() (nemo_rl.data.datasets.response_datasets.numinamath.NuminaMath15Dataset static method) _held_port_uds_name() (in module nemo_rl.distributed.held_port) _HF_CONFIG_PATCHED (in module nemo_rl.models.megatron.setup) _HF_EXPERT_WEIGHT_RE (in module nemo_rl.models.generation.vllm.refit_layout) _HF_PROJECTION_SHARDS (in module nemo_rl.models.generation.vllm.refit_layout) _HF_SNAPSHOT_ALLOW_PATTERNS (in module nemo_rl.models.megatron.draft.utils) _HF_SNAPSHOT_IGNORE_PATTERNS (in module nemo_rl.models.megatron.draft.utils) _hide_extra_state() (nemo_rl.modelopt.models.policy.workers.megatron_quant_policy_worker.MegatronQuantPolicyWorker method) _HTTP_ADAPTER (in module nemo_rl.utils.weight_transfer_http) _http_executor() (in module nemo_rl.utils.weight_transfer_http) _http_get_text() (in module nemo_rl.models.generation.dynamo.metrics) _HTTP_LOCAL (in module nemo_rl.utils.weight_transfer_http) _HTTP_MAX_ATTEMPTS (in module nemo_rl.models.generation.dynamo.dynamo_generation) _HTTP_RETRY_DELAY_S (in module nemo_rl.models.generation.dynamo.dynamo_generation) _http_status() (in module nemo_rl.experience.failures) _hybrid_model_flops() (in module nemo_rl.utils.flops_formulas) _IMAGE_PLACEHOLDER_RE (in module nemo_rl.data.datasets.response_datasets.mmpr_tiny) _image_sources_equal() (in module nemo_rl.environments.nemo_gym) _IMAGE_SRC_PREFIXES (in module nemo_rl.environments.nemo_gym) _INACTIVE_SUBCOMM_CACHE (in module nemo_rl.weight_sync.xferdtensor_python) _index_per_turn_images() (in module nemo_rl.environments.nemo_gym) _INDIVIDUAL_EXPERT_RE (in module nemo_rl.weight_sync.nccl_reshard_utils) _INF_GRAD_NORM_WARNING (in module nemo_rl.utils.grad_norm) _infer_checkpoint_root() (in module nemo_rl.models.automodel.checkpoint) _infer_single_token_id() (in module nemo_rl.experience.rollouts) _INFERENCE_MODEL_OFFLOAD_TAG (in module nemo_rl.models.megatron.memory_saver) _inflight (nemo_rl.models.generation.fleet_health.HealthyShardSelector attribute) _INFRA_TYPE_NAMES (in module nemo_rl.experience.failures) _INFRA_TYPES (in module nemo_rl.experience.failures) _init_checkpoint_manager() (nemo_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl method) (nemo_rl.models.value.workers.dtensor_value_worker_v2.DTensorValueWorkerV2Impl method) _init_config() (nemo_rl.models.generation.vllm.vllm_worker.BaseVllmGenerationWorker method) _init_inference_engine_state() (nemo_rl.models.generation.megatron.megatron_worker.MegatronGenerationMixin method) _init_placement_groups() (nemo_rl.distributed.virtual_cluster.RayVirtualCluster method) _init_tq() (in module nemo_rl.data_plane.adapters.transfer_queue) _initial_distillation_save_state() (in module nemo_rl.algorithms.distillation) _initial_dpo_save_state() (in module nemo_rl.algorithms.dpo) _initial_grpo_save_state() (in module nemo_rl.algorithms.grpo) _initial_policy_generation_stale() (in module nemo_rl.algorithms.grpo) _initial_rm_save_state() (in module nemo_rl.algorithms.rm) _initial_sft_save_state() (in module nemo_rl.algorithms.sft) _initialize_inference_engine() (nemo_rl.models.generation.megatron.megatron_worker.MegatronGenerationMixin method) _inject_gym_token_metadata() (in module nemo_rl.models.generation.dynamo.token_wrapper) _inject_vllm_mm_processor_kwargs() (in module nemo_rl.environments.nemo_gym_video) _install_engine_input_socket_lock() (nemo_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) _install_missing_route_fallback_patch() (in module nemo_rl.models.megatron.router_replay) _install_value_head_load_skip() (in module nemo_rl.models.value.workers.megatron_value_worker) _INTEGER_DTYPE_BY_SIZE (in module nemo_rl.utils.weight_transfer_sparse_codec) _interleave_qkv() (in module nemo_rl.models.megatron.draft.utils) _intersect() (in module nemo_rl.weight_sync.xferdtensor_python) _invalidate() (nemo_rl.algorithms.async_utils.replay_buffer.DataPlaneMutationCut method) _IPCWeightManifest (class in nemo_rl.models.generation.vllm.vllm_backend) _is_build_isolation() (in module nemo_rl) _is_env_truthy() (in module nemo_rl.telemetry.setup) _is_float_format() (in module nemo_rl.modelopt.utils) _is_fp8_export() (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) _is_infra() (in module nemo_rl.experience.failures) _is_marked_valid() (nemo_rl.data.datasets.response_datasets.numinamath.NuminaMath15Dataset static method) _is_multimodal_dataset() (in module nemo_rl.data.datasets.eval_datasets) _is_real_quant_model() (nemo_rl.modelopt.models.generation.vllm_quant_backend.VllmQuantInternalWorkerExtension method) _is_replica_leader() (nemo_rl.data_plane.worker_mixin.TQWorkerMixin method) _is_retryable_http_response() (in module nemo_rl.models.generation.dynamo.dynamo_generation) _is_sharded_refit_weight() (nemo_rl.models.generation.vllm.refit_loader.VllmShardedExpertRefitMixin method) _is_torchcodec_end_of_stream_error() (in module nemo_rl.models.generation.vllm.video_utils) _is_trainable_output_item() (in module nemo_rl.environments.nemo_gym) _is_valid_for_target() (nemo_rl.algorithms.async_utils.replay_buffer.ReplayBufferImpl static method) _isolated_meta() (nemo_rl.data_plane.driver_mixin.TQDriverMixin method) _iter_hf_input_amax_names() (nemo_rl.modelopt.models.policy.workers.megatron_quant_policy_worker.MegatronQuantPolicyWorker static method) _iter_input_quantizer_amax_params() (nemo_rl.modelopt.models.policy.workers.megatron_quant_policy_worker.MegatronQuantPolicyWorker method) _iter_local_hf_param_shards() (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) _iter_model_modules() (in module nemo_rl.models.megatron.router_replay) _iter_modelopt_quant_modules() (in module nemo_rl.modelopt.models.generation.vllm_quant_backend) _iter_params_with_optional_kv_scales() (nemo_rl.modelopt.models.policy.workers.megatron_quant_policy_worker.MegatronQuantPolicyWorker method) (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) _iter_quant_ignore_suffix_variants() (in module nemo_rl.modelopt.utils) _iter_real_quant_refit_params() (nemo_rl.modelopt.models.policy.workers.megatron_quant_policy_worker.MegatronQuantPolicyWorker method) _iter_rollout_groups() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector method) _iter_sglang_hf_weight_buckets() (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) _iter_sparse_payload() (nemo_rl.models.generation.vllm.vllm_sparse_delta.VllmSparseDeltaApplier method) _join_until() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector static method) _json_bytes() (in module nemo_rl.utils.weight_transfer_zmq) _json_mapping() (in module nemo_rl.environments.nemo_gym_video) _latest_tokenized_assistant_index() (in module nemo_rl.models.generation.dynamo.token_wrapper) _LAYER_RE (in module nemo_rl.weight_sync.nccl_reshard_utils) _length_at() (in module nemo_rl.utils.r3_trace) _load_cached_video_frame_manifest() (in module nemo_rl.models.generation.vllm.video_utils) _load_checkpoint_file() (in module nemo_rl.models.megatron.draft.utils) _load_checkpoint_from_directory() (in module nemo_rl.models.megatron.draft.utils) _load_checkpoint_history() (in module nemo_rl.utils.checkpoint) _load_checkpoint_state() (in module nemo_rl.models.megatron.draft.utils) _load_custom_dataloader_func() (nemo_rl.data.dataloader.MultipleDataloaderWrapper method) _load_destination_local_expert_group() (nemo_rl.models.generation.vllm.refit_loader.VllmShardedExpertRefitMixin method) _load_draft_weights() (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) _load_full_hf_weights() (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) _load_hf_weights() (nemo_rl.models.generation.vllm.checkpoint_engine.VllmCheckpointEngineMixin method) (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) _load_index_checkpoint() (in module nemo_rl.models.megatron.draft.utils) _load_libnuma() (in module nemo_rl.distributed.numa_utils) _load_megatron_common_state_dict() (in module nemo_rl.utils.checkpoint) _load_mmpr_tiny_from_cache() (in module nemo_rl.data.datasets.response_datasets.mmpr_tiny) _load_model() (nemo_rl.models.generation.vllm.vllm_worker.BaseVllmGenerationWorker method) _load_modelopt_moe_input_scale() (in module nemo_rl.modelopt.models.generation.vllm_modelopt) _load_records() (nemo_rl.data.datasets.response_datasets.intent.IntentDataset method) _load_safetensors_file() (in module nemo_rl.models.megatron.draft.utils) _load_sharded_expert_weight_groups() (nemo_rl.models.generation.vllm.refit_loader.VllmShardedExpertRefitMixin method) _load_sharded_expert_weights() (nemo_rl.models.generation.vllm.refit_loader.VllmShardedExpertRefitMixin method) _load_sparse_payloads() (nemo_rl.models.generation.vllm.vllm_sparse_delta.VllmSparseDeltaApplier method) _load_torch_file() (in module nemo_rl.models.megatron.draft.utils) _load_video_frames_torchcodec_with_metadata() (in module nemo_rl.models.generation.vllm.video_utils) _load_weights() (nemo_rl.modelopt.models.generation.vllm_quant_backend.VllmQuantInternalWorkerExtension method) (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) (nemo_rl.models.generation.vllm.vllm_sparse_delta.VllmSparseDeltaApplier method) _LoaderObservation (in module nemo_rl.models.generation.vllm.vllm_sparse_delta) _LoaderWeight (in module nemo_rl.models.generation.vllm.vllm_sparse_delta) _local_coords() (nemo_rl.data_plane.worker_mixin.TQWorkerMixin method) (nemo_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) (nemo_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl method) (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) (nemo_rl.models.value.workers.megatron_value_worker.MegatronValueWorkerImpl method) _local_expert_id() (nemo_rl.models.generation.vllm.refit_loader.VllmShardedExpertRefitMixin method) _local_layer_numbers_for_model() (in module nemo_rl.models.megatron.router_replay) _local_slices() (in module nemo_rl.weight_sync.xferdtensor_python) _local_tensor() (in module nemo_rl.weight_sync.xferdtensor_python) _LOCAL_VIDEO_METADATA_KEYS (in module nemo_rl.environments.nemo_gym_video) _locked_file_patch() (in module nemo_rl.models.generation.vllm.patches) _log_code() (nemo_rl.utils.logger.WandbLogger method) _log_diffs() (nemo_rl.utils.logger.WandbLogger method) _log_effective_quantization_ignore_patterns() (in module nemo_rl.models.generation.vllm.vllm_worker) _log_gpu_mem() (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) _log_mixed_rewards_and_advantages_information() (in module nemo_rl.algorithms.grpo) _logprob_dispatch() (nemo_rl.models.policy.tq_policy.TQPolicy method) _looks_like_image_src() (in module nemo_rl.environments.nemo_gym) _MAGIC (in module nemo_rl.utils.routed_experts_codec) _make_embedding_hook() (nemo_rl.models.megatron.draft.hidden_capture.HiddenStateCapture method) _make_layer_output_hook() (nemo_rl.models.megatron.draft.hidden_capture.HiddenStateCapture method) _make_overlength_filtered_video_example() (in module nemo_rl.environments.nemo_gym_video) _make_parse_tool_calls() (in module nemo_rl.models.generation.trtllm.trtllm_http_server) _make_r3_trace_token_identity() (in module nemo_rl.models.megatron.data) _mamba_layer_flops() (in module nemo_rl.utils.flops_formulas) _MANAGED_FLAGS (in module nemo_rl.models.generation.dynamo.arguments) _managed_namespace() (in module nemo_rl.models.generation.dynamo.managed_runtime) _map_hf_state_to_eagle_state() (in module nemo_rl.models.megatron.draft.utils) _map_layer_hf_weight() (in module nemo_rl.models.megatron.draft.utils) _mark_collection_failed() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector method) _mark_data_operation_started() (nemo_rl.data_plane.adapters.transfer_queue.TQDataPlaneClient method) _match_fused_modelopt_moe_weight() (in module nemo_rl.modelopt.models.generation.vllm_quant_backend) _materialize_ragged_pixel_values() (in module nemo_rl.data.multimodal_utils) _MAX_CAUSE_DEPTH (in module nemo_rl.experience.failures) _MAX_NEMO_GYM_STREAM_RETRIES (in module nemo_rl.algorithms.async_utils.trajectory_collector) _max_p2p_peer_degree() (in module nemo_rl.weight_sync.xferdtensor_python) _may_still_run() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector static method) _maybe_adapt_tensor_to_hf() (in module nemo_rl.models.policy.workers.dtensor_policy_worker_v2) _maybe_attach_fleet_health() (in module nemo_rl.algorithms.single_controller_utils.setup) _maybe_enable_vllm_native_tracing() (in module nemo_rl.models.generation.vllm.vllm_worker) _maybe_inject_megatron_train_iters() (in module nemo_rl.algorithms.single_controller_utils.setup) _maybe_merge_lora_weight() (in module nemo_rl.models.policy.workers.dtensor_policy_worker_v2) _maybe_process_fp8_kv_cache() (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) _maybe_process_mtp_drafter_after_loading() (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) _maybe_refit_mtp_drafter() (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) _maybe_restore_async_replay_buffer_checkpoint() (in module nemo_rl.algorithms.grpo) _maybe_restore_native_data_plane_checkpoint() (in module nemo_rl.algorithms.single_controller_utils.setup) _maybe_restore_replacement_reserve() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _maybe_restore_replay_buffer() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _maybe_restore_rollout_recovery() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _maybe_set_force_hf() (in module nemo_rl.models.automodel.setup) _maybe_start_generation_router() (in module nemo_rl.algorithms.single_controller_utils.setup) _MB_METRIC_MAX (in module nemo_rl.algorithms.single_controller_utils.utils) _MB_METRIC_MEAN (in module nemo_rl.algorithms.single_controller_utils.utils) _MB_METRIC_MIN (in module nemo_rl.algorithms.single_controller_utils.utils) _MEBIBYTE (in module nemo_rl.weight_sync.checkpoint_engine_weight_synchronizer) _memlock_limit() (in module nemo_rl.data_plane.adapters.transfer_queue) _merge_checkpoint_shards() (in module nemo_rl.models.megatron.draft.utils) _merge_consecutive_bytes() (in module nemo_rl.algorithms.x_token.token_aligner) _merge_encoding_artifacts() (in module nemo_rl.algorithms.x_token.token_aligner) _merge_fp8_kwargs() (in module nemo_rl.models.generation.vllm.vllm_worker) _merge_generation_logger_workers() (in module nemo_rl.utils.logger) _merge_model_override_value() (in module nemo_rl.models.megatron.setup) _merge_model_overrides() (in module nemo_rl.models.megatron.setup) _merge_refit_info() (nemo_rl.weight_sync.vllm_remote_sparse_weight_synchronizer.VllmRemoteSparseWeightSynchronizer static method) _merge_stop_strings() (nemo_rl.models.generation.dynamo.dynamo_generation.DynamoGeneration method) (nemo_rl.models.generation.megatron.megatron_worker.MegatronGenerationMixin method) (nemo_rl.models.generation.vllm.vllm_worker.BaseVllmGenerationWorker method) _mesh_coordinates() (in module nemo_rl.weight_sync.xferdtensor_python) _mesh_rank_tensor() (in module nemo_rl.weight_sync.xferdtensor_python) _mesh_ranks() (in module nemo_rl.weight_sync.xferdtensor_python) _mesh_signature() (in module nemo_rl.weight_sync.xferdtensor_python) _meta_tensor_alloc_context() (in module nemo_rl.models.policy.workers.megatron_policy_worker) _metadata_extra_body() (in module nemo_rl.environments.nemo_gym_video) _MISSING_ROUTE_FALLBACK_PATCH_ATTR (in module nemo_rl.models.megatron.router_replay) _MISSING_ROUTE_SENTINEL (in module nemo_rl.models.megatron.router_replay) _mla_moe_linear_flops() (in module nemo_rl.utils.flops_formulas) _mla_projection_params() (in module nemo_rl.utils.flops_formulas) _mlp_layer_flops() (in module nemo_rl.utils.flops_formulas) _model_accepts_media_token_validity_mask() (in module nemo_rl.models.policy.workers.megatron_policy_worker) _MODEL_LAYER_QKV_KEY_PATTERN (in module nemo_rl.models.megatron.draft.utils) _model_media_placeholder_token_id() (in module nemo_rl.models.policy.workers.megatron_policy_worker) _MODEL_PREFIX_RE (in module nemo_rl.weight_sync.nccl_reshard_utils) _model_self_packs_for_cp() (in module nemo_rl.models.policy.workers.megatron_policy_worker) _model_self_packs_mtp_loss_mask() (in module nemo_rl.models.policy.workers.megatron_policy_worker) _model_slices_context_parallel_inputs() (in module nemo_rl.models.policy.workers.megatron_policy_worker) _model_uses_unquantized_flashinfer_trtllm() (in module nemo_rl.models.generation.vllm.vllm_backend) _modelopt_layerwise_reload_roots() (in module nemo_rl.modelopt.models.generation.vllm_quant_backend) _moe_ffn_params() (in module nemo_rl.utils.flops_formulas) _mooncake_transport_config() (in module nemo_rl.data_plane.adapters.transfer_queue) _mtp_drafter_from_disk (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension attribute) _mtp_drafter_refit_enabled() (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) _mtp_weights_from_refit (nemo_rl.models.generation.interfaces.GenerationConfig attribute) _MULTI_TOKEN_ARTIFACT_FIXES (in module nemo_rl.algorithms.x_token.token_aligner) _mute_output() (in module nemo_rl.environments.math_environment) (in module nemo_rl.environments.vlm_environment) _NACK (in module nemo_rl.utils.weight_transfer_zmq) _nccl_reshard_refit() (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) _nccl_reshard_refit_guarded() (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) _need_top_k_filtering() (in module nemo_rl.algorithms.logits_sampling_utils) _need_top_p_filtering() (in module nemo_rl.algorithms.logits_sampling_utils) _needs_hf_refit_handshake() (in module nemo_rl.algorithms.grpo) _NEMO_GYM_RETRY_DELAY_BASE_SECONDS (in module nemo_rl.algorithms.async_utils.trajectory_collector) _NEMO_UNIQUE_ID_KEY (in module nemo_rl.distributed.stateless_process_group) _NemoGymStreamAccumulator (class in nemo_rl.experience.rollouts) _NEMOTRON_OMNI_EXPANDED_SEQUENCE_CONTRACT (in module nemo_rl.models.megatron.setup) _nemotron_video_target_resolution() (in module nemo_rl.environments.nemotron_utils) _next_count() (in module nemo_rl.utils.r3_trace) _next_dp_shard_idx() (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) _NIXL_CONFIG_KEY (in module nemo_rl.models.generation.vllm.checkpoint_engine) _non_colocated_teacher_node_count() (in module nemo_rl.algorithms.single_controller_utils.setup) _non_mla_attn_layer_flops() (in module nemo_rl.utils.flops_formulas) _normalise_flag() (in module nemo_rl.models.generation.dynamo.arguments) _normalize_hf_key() (in module nemo_rl.models.megatron.draft.utils) _normalize_routed_experts_for_mcore() (in module nemo_rl.models.megatron.router_replay) _normalize_shard_dim() (in module nemo_rl.weight_sync.xferdtensor_python) _normalize_supported_save_consolidated() (in module nemo_rl.models.automodel.checkpoint) _normalize_tool_arguments_for_template() (in module nemo_rl.models.generation.dynamo.token_wrapper) _notify_refit_apply_waiters() (nemo_rl.models.generation.vllm.vllm_sparse_refit.VllmSparseRefitReceiver method) _nrl_layerwise_reload_active (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension attribute) _nrl_layerwise_reload_failure (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension attribute) _nrl_modelopt_reload_roots (nemo_rl.modelopt.models.generation.vllm_quant_backend.VllmQuantInternalWorkerExtension attribute) _nrl_named_parameters (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension attribute) _nrl_w13_num_shards_by_prefix (nemo_rl.modelopt.models.generation.vllm_quant_backend.VllmQuantInternalWorkerExtension attribute) _NVFP4_REAL_QUANT_MODES (in module nemo_rl.modelopt.utils) _observe_xor_copy() (nemo_rl.models.generation.vllm.vllm_sparse_delta._SparseWeightLoadMode method) _offload_inference_model() (nemo_rl.models.generation.megatron.megatron_worker.MegatronGenerationRefitMixin method) _on_backend_error() (nemo_rl.models.generation.generation_router.GenerationRouterImpl method) _onload_inference_model() (nemo_rl.models.generation.megatron.megatron_worker.MegatronGenerationRefitMixin method) _opd_cfg() (in module nemo_rl.algorithms.opd) _ordered_generation_metadata() (in module nemo_rl.weight_sync.checkpoint_engine_weight_synchronizer) _ordered_replica_members() (in module nemo_rl.weight_sync.xferdtensor_python) _original_get_replay_topk (in module nemo_rl.utils.r3_trace) _OTEL_FALLBACK_PREFIX (in module nemo_rl.telemetry.setup) _OTEL_PREFIX (in module nemo_rl.telemetry.setup) _owner_rank_for_global_layer() (nemo_rl.models.megatron.draft.hidden_capture.HiddenStateCapture method) _pack_generator() (in module nemo_rl.data.datasets.response_datasets.arrow_text_dataset) _pack_implementation() (nemo_rl.data.packing.algorithms.ConcatenativePacker method) (nemo_rl.data.packing.algorithms.FirstFitPacker method) (nemo_rl.data.packing.algorithms.ModifiedFirstFitDecreasingPacker method) (nemo_rl.data.packing.algorithms.SequencePacker method) _pack_input_ids() (in module nemo_rl.algorithms.loss.utils) _pack_sequences_for_megatron() (in module nemo_rl.models.megatron.data) _packing_args() (nemo_rl.data_plane.driver_mixin.TQDriverMixin method) _pad_sequence_aligned_tensors() (in module nemo_rl.models.megatron.data) _pad_teacher_logprobs() (in module nemo_rl.algorithms.grpo) _pad_tensor() (in module nemo_rl.data.llm_message_utils) _pad_token_id (nemo_rl.models.generation.interfaces.GenerationConfig attribute) _pad_value_dict() (nemo_rl.data_plane.worker_mixin.TQWorkerMixin method) _pairs_to_batch() (nemo_rl.algorithms.x_token.token_aligner.TokenAligner static method) _parallelize_gemma3() (in module nemo_rl.models.dtensor.parallelize) _parallelize_llama() (in module nemo_rl.models.dtensor.parallelize) _parallelize_model() (in module nemo_rl.models.dtensor.parallelize) _parallelize_nm5_h() (in module nemo_rl.models.dtensor.parallelize) _parallelize_qwen() (in module nemo_rl.models.dtensor.parallelize) _parent_communicator_key() (in module nemo_rl.weight_sync.xferdtensor_python) _parse_cpulist() (in module nemo_rl.distributed.numa_utils) _parse_data_message() (nemo_rl.utils.weight_transfer_zmq.ZmqSparseRefitServer method) _parse_dynamo_completion_response() (in module nemo_rl.models.generation.dynamo.dynamo_generation) _parse_extra_options() (in module nemo_rl.utils.nsys) _parse_gpu_sku() (nemo_rl.utils.logger.RayGpuMonitorLogger method) _parse_layer_checkpoint_key() (in module nemo_rl.models.megatron.draft.utils) _parse_metric() (nemo_rl.utils.logger.RayGpuMonitorLogger method) _parse_question() (in module nemo_rl.data.datasets.response_datasets.avqa) _parse_result_to_batched_data_dict() (nemo_rl.models.generation.megatron.megatron_worker.MegatronGenerationMixin method) _Partition (class in nemo_rl.data_plane.adapters.noop) _PassThroughEnvConfig (class in nemo_rl.evals.eval) _patch_bridge_signal_handler_for_worker_threads() (in module nemo_rl.models.megatron.setup) _patch_hf_config_double_instantiation() (in module nemo_rl.models.megatron.setup) _patch_lock (in module nemo_rl.utils.r3_trace) _patch_mooncake_register_check() (in module nemo_rl.data_plane.adapters.transfer_queue) _patch_mooncake_staging_buffers() (in module nemo_rl.data_plane.adapters.transfer_queue) _patch_named_parameters_to_include_buffers() (nemo_rl.modelopt.models.generation.vllm_quant_backend.VllmQuantInternalWorkerExtension method) _patch_nsight_file() (in module nemo_rl) _patch_qwen_vl_vision_key_mapping() (in module nemo_rl.models.automodel.checkpoint) _patch_setup_model_and_optimizer() (nemo_rl.modelopt.models.policy.workers.megatron_quant_policy_worker.MegatronQuantPolicyWorker method) _patch_transformers_tokenizer_class_set() (in module nemo_rl.models.policy) _patch_validate_model_paths() (nemo_rl.modelopt.models.policy.workers.megatron_quant_policy_worker.MegatronQuantPolicyWorker method) _patch_vllm_glm_decoder_sequence_parallel_moe() (in module nemo_rl.models.generation.vllm.patches) _patch_vllm_init_workers_ray() (in module nemo_rl.models.generation.vllm.patches) _patch_vllm_llama_eagle3_own_lm_head() (in module nemo_rl.models.generation.vllm.patches) _patch_vllm_nsight_config() (nemo_rl.models.generation.vllm.vllm_worker.BaseVllmGenerationWorker static method) _patch_vllm_radio_layerscale_loader() (in module nemo_rl.models.generation.vllm.patches) _patch_vllm_ray_executor_v2_tcpstore_port() (in module nemo_rl.models.generation.vllm.patches) _patch_vllm_shm_broadcast_bind_retry() (in module nemo_rl.models.generation.vllm.patches) _patch_vllm_tool_parser_namespace_tool() (in module nemo_rl.models.generation.vllm.patches) _patched (in module nemo_rl.utils.fastokens) _payload_indices_for_moe_layers() (in module nemo_rl.models.megatron.router_replay) _PENDING_PAYLOAD_METRICS (in module nemo_rl.utils.multimodal_payload_metrics) _PENDING_PAYLOAD_METRICS_LOCK (in module nemo_rl.utils.multimodal_payload_metrics) _PendingLayerWeights (class in nemo_rl.models.megatron.draft.utils) _pick_backend() (nemo_rl.models.generation.generation_router.GenerationRouterImpl method) _pickle_vllm_unique_id() (in module nemo_rl.distributed.stateless_process_group) _PIXEL_DTYPE_CAST_KEYS (nemo_rl.distributed.batched_data_dict.BatchedDataDict attribute) _placeholder_seq_logprob_error_metrics() (in module nemo_rl.algorithms.grpo) _PLACEHOLDER_STYLE_PROCESSOR_NAMES (in module nemo_rl.data.multimodal_utils) _placement_signature() (in module nemo_rl.weight_sync.xferdtensor_python) _PLAN_CACHE (in module nemo_rl.weight_sync.xferdtensor_python) _PLAN_CACHE_MAX_SIZE (in module nemo_rl.weight_sync.xferdtensor_python) _plan_geometry() (in module nemo_rl.weight_sync.xferdtensor_python) _policy (nemo_rl.weight_sync.checkpoint_engine_weight_synchronizer.CheckpointEngineWeightSynchronizer attribute) _policy_dtype() (in module nemo_rl.algorithms.grpo) _pooled_opd_metrics() (in module nemo_rl.algorithms.single_controller) _post_completion_request() (nemo_rl.models.generation.dynamo.dynamo_generation.DynamoGeneration method) _post_init() (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) _post_process_alignment() (nemo_rl.algorithms.x_token.token_aligner.TokenAligner static method) _post_worker_route() (in module nemo_rl.models.generation.dynamo.refit) _postprocess_nemo_gym_to_nemo_rl_result() (nemo_rl.environments.nemo_gym.NemoGym method) _postprocess_single_nemo_gym_group() (in module nemo_rl.experience.rollouts) _prefer_nvrx_for_dist_ckpt_save() (in module nemo_rl.models.megatron.community_import) _preference_loss() (nemo_rl.algorithms.loss.loss_functions.PreferenceLossFn method) _prepare_checkpoint_engine_weight_send() (nemo_rl.models.policy.workers.checkpoint_engine.DTensorCheckpointEngineSendMixin method) (nemo_rl.models.policy.workers.checkpoint_engine.PolicyCheckpointEngineMixin method) _prepare_data_for_generation() (nemo_rl.models.generation.megatron.megatron_worker.MegatronGenerationMixin method) _prepare_for_training_step() (nemo_rl.algorithms.async_utils.replay_buffer.ReplayBufferImpl method) _prepare_input_fn() (nemo_rl.models.dtensor.parallelize.RotaryEmbedParallel static method) _prepare_loader_weight() (nemo_rl.models.generation.vllm.vllm_sparse_delta.VllmSparseDeltaApplier method) _prepare_multimodal_sharing() (in module nemo_rl.distributed.batched_data_dict) _prepare_nemo_gym_rows() (in module nemo_rl.experience.rollouts) _prepare_output_fn() (nemo_rl.models.dtensor.parallelize.ColwiseParallelWithGather static method) (nemo_rl.models.dtensor.parallelize.RotaryEmbedParallel static method) _prepare_sequences() (nemo_rl.data.packing.algorithms.FirstFitDecreasingPacker method) (nemo_rl.data.packing.algorithms.FirstFitPacker method) (nemo_rl.data.packing.algorithms.FirstFitShufflePacker method) _prepare_sparse_refit_info() (nemo_rl.models.generation.vllm.vllm_sparse_refit.VllmSparseRefitReceiver method) _prepare_vlm_batch_for_megatron() (in module nemo_rl.models.megatron.data) _preserve_router_replay_routed_experts() (in module nemo_rl.algorithms.grpo) _PRESETS (nemo_rl.telemetry.span_groups.RLSpanGroup attribute) _print_results() (in module nemo_rl.evals.eval) _probe_generation_fleet() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _process_batch() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector method) _PROCESS_GROUP_COMM_IDS (in module nemo_rl.weight_sync.xferdtensor_python) _PROCESS_GROUP_FINALIZERS (in module nemo_rl.weight_sync.xferdtensor_python) _promote_1d_leaves() (in module nemo_rl.data_plane.adapters.transfer_queue) _promote_into_step() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _promote_refit_shards() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _prompt_task_name() (in module nemo_rl.experience.rollout_recovery) _prompt_token_ids() (nemo_rl.models.generation.dynamo.dynamo_generation.DynamoGeneration method) _PROTOCOL (in module nemo_rl.utils.weight_transfer_zmq) _push_router_membership() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _qkv_head_dims() (in module nemo_rl.models.megatron.draft.utils) _QUANT_AMAX_SUFFIXES (nemo_rl.modelopt.models.generation.vllm_quant_backend.VllmQuantInternalWorkerExtension attribute) _quant_cfg_for_worker_env() (in module nemo_rl.modelopt.models.generation.vllm_quant_worker) _quant_checkpoint_cache_suffix() (in module nemo_rl.modelopt.models.policy.workers.megatron_quant_policy_worker) _QUANT_IGNORE_NAME_SUFFIXES (in module nemo_rl.modelopt.utils) _quantization_cfg() (nemo_rl.weight_sync.sglang_weight_synchronizer._SGLangWeightSynchronizer method) _quantize() (nemo_rl.modelopt.models.policy.workers.megatron_quant_policy_worker.MegatronQuantPolicyWorker method) _raise_if_message_level_advantage_penalties_enabled() (in module nemo_rl.algorithms.grpo_sync) _raise_if_reward_penalties_enabled_without_nemo_gym() (in module nemo_rl.algorithms.grpo) _rank_regions() (in module nemo_rl.weight_sync.xferdtensor_python) _ray_get() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _read_manifest() (in module nemo_rl.data.datasets.response_datasets.intent) _read_mtp_layer_weights_from_checkpoint() (in module nemo_rl.models.generation.vllm.vllm_backend) _reattach_original_multimodal_payloads() (in module nemo_rl.experience.rollouts) _reattach_static_multimodal_payloads_to_result() (in module nemo_rl.experience.rollouts) _rebuild_checkpointer_addons() (nemo_rl.models.automodel.checkpoint.AutomodelCheckpointManager method) _receive_and_load_misc_params() (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) _receive_weight_chunk_batches() (nemo_rl.utils.checkpoint_engines.nixl.NIXLCheckpointEngine method) _reconcile_refit_membership() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _record_clear() (nemo_rl.data_plane.observability.MetricsDataPlaneClient method) _record_put() (nemo_rl.data_plane.observability.MetricsDataPlaneClient method) _record_vllm_generation_metrics() (in module nemo_rl.models.generation.vllm.vllm_generation) _recover_from_failed_refit() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _recovering_from_refit (nemo_rl.algorithms.single_controller.SingleControllerActor attribute) _recovery_window() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _recv_tensor() (nemo_rl.models.megatron.draft.hidden_capture.HiddenStateCapture static method) _redispatch_restored_rollouts() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _reduce_seq_logprob_error_metrics() (in module nemo_rl.algorithms.single_controller_utils.utils) _REDUCTION_FUNCTIONS (nemo_rl.utils.timer.Timer attribute) _refit() (nemo_rl.weight_sync.sglang_weight_synchronizer._SGLangWeightSynchronizer method) _refit_await_budget_s() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _refit_collective_response() (nemo_rl.models.generation.vllm.vllm_sparse_refit.VllmSparseRefitReceiver static method) _refit_collective_rpc() (nemo_rl.models.generation.vllm.vllm_sparse_refit.VllmSparseRefitReceiver method) _refit_leader_workers() (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) _refit_tensor_dtype() (in module nemo_rl.models.policy.workers.dtensor_policy_worker_v2) _refit_transport_state() (nemo_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker method) _REFIT_UNWIND_GRACE_S (nemo_rl.algorithms.single_controller.SingleControllerActor attribute) _refresh_hpc_modules_after_layerwise_reload() (in module nemo_rl.models.generation.vllm.vllm_backend) _refresh_membership() (nemo_rl.models.generation.fleet_health.GenerationFleetHealth method) _region_numel() (in module nemo_rl.weight_sync.xferdtensor_python) _region_view_is_contiguous() (in module nemo_rl.weight_sync.xferdtensor_python) _register_checked() (in module nemo_rl.data_plane.adapters.transfer_queue) _register_routed_experts_quant_module() (in module nemo_rl.modelopt.models.generation.vllm_quant_patch) _registered (in module nemo_rl.modelopt.models.generation.vllm_modelopt) _rehydrate_rollout_recovery_prompts() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _reject_kv_scales() (nemo_rl.weight_sync.sglang_weight_synchronizer._SGLangWeightSynchronizer method) _reject_non_tensor_leaves() (in module nemo_rl.data_plane.adapters.noop) _reject_relocated_keys() (nemo_rl.algorithms.single_controller_utils.config.AsyncRLConfig method) _reject_renamed_blocks() (nemo_rl.algorithms.single_controller_utils.config.AsyncRLConfig method) _reject_renamed_keys() (nemo_rl.algorithms.single_controller_utils.config.RolloutFailureConfig method) _reject_unsupported_native_refit() (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) _rekey() (nemo_rl.data.datasets.eval_datasets.gpqa.GPQADataset method) (nemo_rl.data.datasets.eval_datasets.local_math_dataset.LocalMathDataset method) (nemo_rl.data.datasets.eval_datasets.math.MathDataset method) (nemo_rl.data.datasets.eval_datasets.mmlu.MMLUDataset method) (nemo_rl.data.datasets.eval_datasets.mmlu_pro.MMLUProDataset method) _RelayTransfer (class in nemo_rl.utils.weight_transfer_zmq) _release_after_refit() (nemo_rl.weight_sync.checkpoint_engine_weight_synchronizer.CheckpointEngineWeightSynchronizer method) _release_target() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector method) _release_transfer_buffers() (nemo_rl.utils.checkpoint_engines.nixl.NIXLCheckpointEngine method) _remap_pairs_to_original() (nemo_rl.algorithms.x_token.token_aligner.TokenAligner static method) _remember_changed() (nemo_rl.models.generation.vllm.vllm_sparse_delta._SparseWeightLoadMode method) _remote_sparse_refit (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl attribute) _REMOTE_SPARSE_TRANSPORTS (in module nemo_rl.weight_sync.vllm_remote_sparse_weight_synchronizer) _remote_transfers_avoid_double_staging() (in module nemo_rl.weight_sync.xferdtensor_python) _remove_incomplete_target_steps() (nemo_rl.algorithms.async_utils.replay_buffer.ReplayBufferImpl method) _remove_indices() (nemo_rl.algorithms.async_utils.replay_buffer.ReplayBufferImpl method) _remove_stale_trajectories() (nemo_rl.algorithms.async_utils.replay_buffer.ReplayBufferImpl method) _remove_unlocked() (nemo_rl.algorithms.async_utils.replay_buffer.TQReplayBuffer method) _remove_vllm_mm_processor_kwargs() (in module nemo_rl.environments.nemo_gym_video) _rename_checkpoint() (nemo_rl.utils.checkpoint.CheckpointManager method) _render_nemotron_video_prompt() (in module nemo_rl.environments.nemotron_utils) _render_prompt_token_ids() (in module nemo_rl.models.generation.dynamo.token_wrapper) _render_prompt_token_ids_with_optional_prefix() (in module nemo_rl.models.generation.dynamo.token_wrapper) _replace() (nemo_rl.data_plane.interfaces.KVBatchMeta method) _replace_cached_video_frames_with_native_video() (in module nemo_rl.environments.nemo_gym_video) _REPLAY_BUFFER_MAX_BACKOFF_SECONDS (in module nemo_rl.algorithms.async_utils.trajectory_collector) _REPLICATED_AXES (in module nemo_rl.models.value.tq_value) _report_device_id() (nemo_rl.models.generation.trtllm.trtllm_generation.TrtllmGeneration method) (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) _report_dp_openai_server_base_urls() (nemo_rl.models.generation.trtllm.trtllm_generation.TrtllmGeneration method) (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) _report_sharded_payload() (nemo_rl.models.policy.lm_policy.Policy method) _request() (nemo_rl.utils.weight_transfer_stream._S3ObjectStore method) _request_add_generation_prompt() (in module nemo_rl.models.generation.dynamo.token_wrapper) _request_max_new_tokens() (nemo_rl.models.generation.vllm.vllm_worker.BaseVllmGenerationWorker class method) _request_receivers() (nemo_rl.weight_sync.vllm_remote_sparse_weight_synchronizer.VllmRemoteSparseWeightSynchronizer method) _request_timeout_s() (nemo_rl.models.generation.dynamo.dynamo_generation.DynamoGeneration method) _require_checkpointing_support() (nemo_rl.data_plane.adapters.transfer_queue.TQDataPlaneClient method) _require_clean_for_load() (nemo_rl.data_plane.adapters.transfer_queue.TQDataPlaneClient method) _require_complete_modelopt_layerwise_reload() (in module nemo_rl.modelopt.models.generation.vllm_quant_backend) _require_dp_client() (nemo_rl.data_plane.worker_mixin.TQWorkerMixin method) _require_group() (nemo_rl.experience.rollout_recovery.RolloutRecoveryLedger method) _require_int() (in module nemo_rl.experience.rollout_recovery) _require_nonempty_vllm_config() (in module nemo_rl.models.generation.dynamo.config) _require_remote_sparse_refit() (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) _require_spinup() (nemo_rl.environments.nemo_gym.NemoGym method) _require_state_tensor() (in module nemo_rl.models.megatron.draft.utils) _require_unwanted_token_ids_when_penalized() (nemo_rl.algorithms.grpo.RewardPenaltyConfig method) _require_video_config_value() (in module nemo_rl.environments.nemo_gym_video) _required_config_value() (in module nemo_rl.environments.nemotron_utils) _requires_nvrx_cuda_cache_release() (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) _resample_audio() (in module nemo_rl.data.datasets.response_datasets.audiomcq) (in module nemo_rl.data.datasets.response_datasets.avqa) _reserve_port() (nemo_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) _reset_encoder_cache_after_weight_update() (nemo_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) _reshard_into_inference_model() (nemo_rl.models.generation.megatron.megatron_worker.MegatronGenerationRefitMixin method) _resize_and_normalize_nemotron_video_frame() (in module nemo_rl.environments.nemotron_utils) _resolve_bucket_size_bytes() (nemo_rl.weight_sync.checkpoint_engine_weight_synchronizer.CheckpointEngineWeightSynchronizer method) _resolve_cached_video_media_path() (in module nemo_rl.models.generation.vllm.video_utils) _resolve_device_id() (in module nemo_rl.utils.nvml) _resolve_effective_quantizer_formats() (in module nemo_rl.modelopt.utils) _resolve_enable_prefix_caching() (in module nemo_rl.models.generation.vllm.vllm_worker) _resolve_gold_xtoken() (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossFn method) _resolve_iter_dir_from_root() (in module nemo_rl.models.megatron.setup) _resolve_lambda_policy() (nemo_rl.algorithms.advantage_estimator.GeneralizedAdvantageEstimator method) _resolve_lambda_value() (nemo_rl.algorithms.advantage_estimator.GeneralizedAdvantageEstimator method) _resolve_local_video_path() (in module nemo_rl.environments.nemo_gym_video) _resolve_logprob_skip_flags() (in module nemo_rl.algorithms.grpo) _resolve_message_level_advantage_penalties() (in module nemo_rl.algorithms.grpo) _resolve_optional_key() (in module nemo_rl.models.megatron.draft.utils) _resolve_snapshot_root() (in module nemo_rl.data.datasets.response_datasets.audiomcq) _resolve_teacher() (nemo_rl.algorithms.opd.TQTeacherLogprobCoordinator method) _resolve_tool_parser_name() (in module nemo_rl.models.generation.trtllm.trtllm_http_server) _resolve_video_path() (in module nemo_rl.data.datasets.response_datasets.intent) _resolve_wake_tags() (nemo_rl.models.generation.trtllm.trtllm_worker_async.TrtllmAsyncGenerationWorkerImpl method) _restore_model_extra_state_dict() (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) _restore_modelopt_state_pre_load() (nemo_rl.modelopt.models.policy.workers.megatron_quant_policy_worker.MegatronQuantPolicyWorker method) _restore_placement() (in module nemo_rl.weight_sync.nccl_reshard_utils) _restore_saved_mcore_hooks() (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) _restore_tensors() (in module nemo_rl.data.multimodal_utils) _result_to_completion() (nemo_rl.experience.rollout_manager.AsyncNemoGymRolloutImpl method) _RETRIABLE_HTTP_STATUSES (in module nemo_rl.experience.failures) _RETRYABLE_HTTP_STATUS_CODES (in module nemo_rl.models.generation.dynamo.dynamo_generation) _retune_lookahead_versions() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _return_routed_experts_enabled() (nemo_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) _REWARD_PENALTY_FLAGS (in module nemo_rl.algorithms.grpo) _reward_whiten() (nemo_rl.algorithms.advantage_estimator.GeneralizedAdvantageEstimator method) _rich_logging_configured (in module nemo_rl.utils.logger) _rollout_metrics_turn_count_for_diagnostics() (nemo_rl.algorithms.async_utils.replay_buffer.ReplayBufferImpl static method) _rollout_pump() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _round_up_to_multiple() (in module nemo_rl.models.megatron.common) _round_video_frame_count() (in module nemo_rl.models.generation.vllm.video_utils) _ROUTED_EXPERTS_DTYPE_NAMES (in module nemo_rl.models.generation.interfaces) _ROUTED_EXPERTS_DTYPES (in module nemo_rl.environments.nemo_gym) _router_replay_action_name() (in module nemo_rl.utils.r3_trace) _router_replay_instances_for_model() (in module nemo_rl.models.megatron.router_replay) _router_replay_patch_depth (in module nemo_rl.utils.r3_trace) _ROUTER_REPLAY_VALIDATE_ENV (in module nemo_rl.models.megatron.router_replay) _router_replay_validation_enabled() (in module nemo_rl.models.megatron.router_replay) _row_segment_indices() (nemo_rl.data.multimodal_utils.PackedTensor method) _run() (nemo_rl.data_plane.observability.MetricsDataPlaneClient method) (nemo_rl.models.generation.dynamo.metrics.DynamoMetricsSampler method) (nemo_rl.utils.weight_transfer_zmq.ZmqSparseRefitServer method) _run_async_coordinator_start() (nemo_rl.models.generation.megatron.megatron_worker.MegatronGenerationMixin method) _run_collection_loop() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector method) _run_env_eval_impl() (in module nemo_rl.evals.eval) _run_generation() (nemo_rl.weight_sync.checkpoint_engine_weight_synchronizer.CheckpointEngineWeightSynchronizer method) _run_generation_workers() (nemo_rl.weight_sync.vllm_remote_sparse_weight_synchronizer.VllmRemoteSparseWeightSynchronizer method) _RUN_ID_ENV (in module nemo_rl.telemetry.setup) _run_multi_turn_rollout_async() (in module nemo_rl.experience.rollouts) _run_policy() (nemo_rl.weight_sync.checkpoint_engine_weight_synchronizer.CheckpointEngineWeightSynchronizer method) _run_policy_workers() (nemo_rl.weight_sync.vllm_remote_sparse_weight_synchronizer.VllmRemoteSparseWeightSynchronizer method) _run_rollout_batch_worker() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector method) _run_rollouts() (nemo_rl.experience.rollout_manager.AsyncNemoGymRolloutImpl method) _run_single_rollout() (nemo_rl.experience.rollout_manager.AsyncRolloutImpl method) _RUN_WINDOW_WALL_CLOCK_CATEGORIES (in module nemo_rl.telemetry.metrics) _s3_client() (in module nemo_rl.utils.weight_transfer_stream) _S3_MEMORY_LIMIT (in module nemo_rl.utils.weight_transfer_stream) _S3_PART_SIZE (in module nemo_rl.utils.weight_transfer_stream) _S3ObjectStore (class in nemo_rl.utils.weight_transfer_stream) _same_vocab_masked_kl() (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossFn method) _sampler_class_for_config() (in module nemo_rl.algorithms.async_utils.staleness_sampler) _save_async_replay_buffer_checkpoint() (in module nemo_rl.algorithms.grpo) _save_checkpoint() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _save_data_plane_checkpoint() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _save_evaluation_data_to_json() (in module nemo_rl.evals.eval) _scalar() (in module nemo_rl.telemetry.metrics) _scatter_values() (nemo_rl.models.generation.vllm.vllm_sparse_delta.VllmSparseDeltaApplier static method) _select_extracted_answer() (in module nemo_rl.environments.math_environment) _select_teacher_kd() (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossFn method) _select_video_frame_count() (in module nemo_rl.models.generation.vllm.video_utils) _send_buckets() (nemo_rl.weight_sync.sglang_weight_synchronizer._SGLangWeightSynchronizer method) (nemo_rl.weight_sync.sglang_weight_synchronizer.SGLangColocatedWeightSynchronizer method) (nemo_rl.weight_sync.sglang_weight_synchronizer.SGLangDisaggregatedWeightSynchronizer method) _send_reply() (nemo_rl.utils.weight_transfer_zmq.ZmqSparseRefitServer static method) _send_tensor() (nemo_rl.models.megatron.draft.hidden_capture.HiddenStateCapture static method) _serialise_value() (in module nemo_rl.models.generation.dynamo.arguments) _serve_fd_once() (nemo_rl.distributed.held_port.HeldPortReservation method) _service_env() (nemo_rl.models.generation.dynamo.managed_runtime.ManagedDynamoRuntime method) _SERVICE_NAME_ENV (in module nemo_rl.telemetry.setup) _SERVING_STATES (in module nemo_rl.models.generation.fleet_health) _set_moe_grad_scale_func() (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) _set_mtp_grad_scale_func() (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) _set_numa_membind() (in module nemo_rl.distributed.numa_utils) _set_quantization_model_specs() (in module nemo_rl.modelopt.models.policy.workers.megatron_quant_policy_worker) _settle_before_propagating() (in module nemo_rl.weight_sync.collective_weight_synchronizer) (in module nemo_rl.weight_sync.nccl_reshard_weight_synchronizer) _settle_budget_s() (nemo_rl.weight_sync.collective_weight_synchronizer.CollectiveWeightSynchronizer method) (nemo_rl.weight_sync.nccl_reshard_weight_synchronizer.NcclReshardWeightSynchronizer method) _setup_colocated_cuda_graph_managers() (nemo_rl.models.generation.megatron.megatron_worker.MegatronGenerationMixin method) _setup_openai_api_server() (nemo_rl.models.generation.megatron.megatron_worker.MegatronGenerationMixin method) _setup_vllm_openai_api_server() (nemo_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) _setup_vllm_refit_server() (nemo_rl.models.generation.vllm.vllm_sparse_refit.VllmSparseRefitReceiver method) _setup_vllm_server() (nemo_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) _SGLangWeightSynchronizer (class in nemo_rl.weight_sync.sglang_weight_synchronizer) _shape() (in module nemo_rl.utils.r3_trace) _shard_base_urls() (in module nemo_rl.algorithms.single_controller_utils.setup) _shard_for_logprob() (nemo_rl.models.policy.lm_policy.Policy method) _shard_for_train() (nemo_rl.models.policy.lm_policy.Policy method) _shard_routed_experts_for_cp() (in module nemo_rl.models.megatron.data) _shard_to_local_tp() (in module nemo_rl.models.megatron.draft.utils) _sharded_refit_param_names() (nemo_rl.models.generation.vllm.refit_loader.VllmShardedExpertRefitMixin method) _shift_pairs() (nemo_rl.algorithms.x_token.token_aligner.TokenAligner static method) _should_log_nemo_gym_responses() (in module nemo_rl.algorithms.grpo) _should_pause_for_generation_limits() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector method) _should_trace_step() (in module nemo_rl.utils.r3_trace) _should_use_router_replay() (in module nemo_rl.models.policy.workers.megatron_policy_worker) _SINGLE_LETTER_LINE (in module nemo_rl.data.datasets.eval_datasets.daily_omni) _single_sample_output() (nemo_rl.models.generation.dynamo.dynamo_generation.DynamoGeneration method) _skip_prev_logprobs() (in module nemo_rl.algorithms.opd) _SKIPPED_REQUEST_HEADERS (in module nemo_rl.models.generation.generation_router) _SKIPPED_RESPONSE_HEADERS (in module nemo_rl.models.generation.generation_router) _sleep() (nemo_rl.models.generation.megatron.megatron_worker.MegatronGenerationMixin method) _sleep_engine() (nemo_rl.models.generation.megatron.megatron_worker.MegatronGenerationMixin method) _sort_bundle_indices_by_topology() (in module nemo_rl.distributed.virtual_cluster) _sort_ranked_metadata() (in module nemo_rl.weight_sync.checkpoint_engine_weight_synchronizer) _source_rank_for_rollout() (in module nemo_rl.utils.checkpoint_engines.nixl) _source_tensor() (nemo_rl.models.generation.vllm.vllm_sparse_delta.VllmSparseDeltaApplier method) _sparse_delta_applier (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension attribute) _SPARSE_PROJECTION_CACHE (in module nemo_rl.algorithms.x_token.loss_utils) _SparsePayloadBucket (class in nemo_rl.utils.weight_transfer_stream) _SparseWeightLoadMode (class in nemo_rl.models.generation.vllm.vllm_sparse_delta) _spec_decode_max_tokens() (nemo_rl.models.generation.vllm.vllm_worker.BaseVllmGenerationWorker static method) _SPECIAL_TOKEN_MAP (in module nemo_rl.algorithms.x_token.token_aligner) _spinup() (nemo_rl.environments.nemo_gym.NemoGym method) _spinup_gym() (in module nemo_rl.algorithms.single_controller_utils.setup) _SPLIT_CONFIG (in module nemo_rl.data.datasets.response_datasets.intent) _split_for_sequence_parallel() (in module nemo_rl.models.megatron.router_replay) _split_policy_and_draft_weights() (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension static method) _split_step_state_init() (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) _stack_ragged_pixel_values() (in module nemo_rl.data.multimodal_utils) _stage_rank_operations() (in module nemo_rl.weight_sync.xferdtensor_python) _stage_sparse_payload() (in module nemo_rl.models.generation.vllm.vllm_sparse_refit) _stage_striped_operations() (in module nemo_rl.weight_sync.xferdtensor_python) _StagedSparsePayload (class in nemo_rl.models.generation.vllm.vllm_sparse_refit) _STAGING_SLOT_TIMEOUT_S (in module nemo_rl.data_plane.adapters.transfer_queue) _StagingPool (class in nemo_rl.data_plane.adapters.transfer_queue) _StagingPoolRegistry (class in nemo_rl.data_plane.adapters.transfer_queue) _stale (nemo_rl.weight_sync.checkpoint_engine_weight_synchronizer.CheckpointEngineWeightSynchronizer attribute) _stall_watchdog_pump() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _stamp() (nemo_rl.algorithms.async_utils.staleness_sampler._GatedSampler method) (nemo_rl.algorithms.async_utils.staleness_sampler.InOrderSampler method) _stamp_pad_seqlen() (nemo_rl.data_plane.driver_mixin.TQDriverMixin method) _stamp_task_indices() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector method) _stamped_task_indices() (in module nemo_rl.algorithms.async_utils.trajectory_collector) _stand_down_refit_deadline() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _start_etcd() (nemo_rl.models.generation.dynamo.managed_runtime.ManagedDynamoRuntime method) _start_frontend() (nemo_rl.models.generation.dynamo.managed_runtime.ManagedDynamoRuntime method) _start_inference_coordinator() (nemo_rl.models.generation.megatron.megatron_worker.MegatronGenerationMixin method) _start_inference_loop_thread() (nemo_rl.models.generation.megatron.megatron_worker.MegatronGenerationMixin method) _start_nats() (nemo_rl.models.generation.dynamo.managed_runtime.ManagedDynamoRuntime method) _start_vllm_metrics_logger() (nemo_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) _stop_process() (nemo_rl.models.generation.dynamo.dynamo_worker.DynamoVllmWorker method) (nemo_rl.models.generation.dynamo.managed_runtime.ManagedDynamoRuntime static method) _storage_key() (in module nemo_rl.models.generation.vllm.vllm_sparse_delta) _STR_TO_DTYPE (in module nemo_rl.weight_sync.nccl_reshard_utils) _STREAM_CHUNK_BYTES (in module nemo_rl.models.generation.generation_router) _STREAM_LOCAL (in module nemo_rl.utils.weight_transfer_stream) _stream_rows() (nemo_rl.experience.rollout_manager.AsyncNemoGymRolloutImpl method) _strings_equal_flexible() (in module nemo_rl.algorithms.x_token.token_aligner) _strip_gym_token_metadata() (in module nemo_rl.models.generation.dynamo.token_wrapper) _strip_local_media_metadata() (in module nemo_rl.environments.nemo_gym_video) _striped_geometry() (in module nemo_rl.weight_sync.xferdtensor_python) _STRIPED_PLAN_CACHE (in module nemo_rl.weight_sync.xferdtensor_python) _STRIPED_PLAN_CACHE_MAX_SIZE (in module nemo_rl.weight_sync.xferdtensor_python) _SUBCOMM_CACHE (in module nemo_rl.weight_sync.xferdtensor_python) _submit_pending_sparse_payloads() (nemo_rl.models.generation.vllm.vllm_sparse_refit.VllmSparseRefitReceiver method) _sum_kd() (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossFn method) _summarize_list() (in module nemo_rl.utils.logger) _supports_unquantized_flashinfer_trtllm_refit() (nemo_rl.modelopt.models.generation.vllm_quant_backend.VllmQuantInternalWorkerExtension method) (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) _sync_device() (in module nemo_rl.utils.checkpoint_engines.nixl) _sync_weights() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _sync_weights_within() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _synchronize_before_ipc_data_ack() (nemo_rl.modelopt.models.generation.vllm_quant_backend.VllmQuantInternalWorkerExtension method) (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) _take_replacement() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _target_ (nemo_rl.models.policy.AutomodelBackendConfig attribute) _target_groups_for_step() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _target_url() (nemo_rl.models.generation.generation_router.GenerationRouterImpl static method) _td_bytes() (in module nemo_rl.data_plane.observability) _teacher_is_same_vocab() (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossFn method) _teacher_score_inputs() (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossFn method) _teacher_weight_score() (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossFn method) _tee_efficiency_metrics() (in module nemo_rl.telemetry.metrics) _tee_rl_metrics_to_otel() (in module nemo_rl.telemetry.metrics) _TELEMETRY_HANDLE (in module nemo_rl.telemetry.setup) _TELEMETRY_INITIALISED (in module nemo_rl.telemetry.setup) _tensor_metadata() (in module nemo_rl.weight_sync.xferdtensor_python) _tensor_nbytes() (in module nemo_rl.utils.multimodal_payload_metrics) _tensor_preview() (in module nemo_rl.utils.r3_trace) _tensor_record() (in module nemo_rl.utils.r3_trace) _tensor_sha256() (in module nemo_rl.utils.r3_trace) _tensorize_by_key() (in module nemo_rl.experience.rollouts) _tensorize_nemo_gym_result() (in module nemo_rl.experience.rollouts) _TensorPayloadBuilder (class in nemo_rl.utils.weight_transfer_sparse_codec) _tensors_equal() (in module nemo_rl.utils.r3_trace) _TensorViewKey (in module nemo_rl.models.generation.vllm.vllm_sparse_delta) _timed_refit() (nemo_rl.weight_sync.sglang_weight_synchronizer._SGLangWeightSynchronizer method) _timestamp_to_video_frame_index() (in module nemo_rl.models.generation.vllm.video_utils) _to_int_ids() (in module nemo_rl.models.generation.trtllm.trtllm_http_server) _tokenize_batch() (nemo_rl.data.cross_tokenizer_collate.CrossTokenizerCollator static method) _tolerate_dummy_weight_nan_amax() (in module nemo_rl.modelopt.models.generation.vllm_quant_patch) _TOOL_ARGUMENT_MAPPING_ERROR (in module nemo_rl.models.generation.dynamo.token_wrapper) _TOPK_PROJECTION_CACHE (in module nemo_rl.algorithms.x_token.loss_utils) _TORCH_DTYPE_NAMES (in module nemo_rl.utils.routed_experts_codec) _torch_rank_info() (in module nemo_rl.utils.r3_trace) _TORCHCODEC_END_OF_STREAM_ERROR (in module nemo_rl.models.generation.vllm.video_utils) _torchcodec_sample_indices() (in module nemo_rl.models.generation.vllm.video_utils) _tp_target_logprobs() (in module nemo_rl.distributed.model_utils) _tq_shape_drift_error() (in module nemo_rl.data_plane.adapters.transfer_queue) _TRACE_DIR_ENV (in module nemo_rl.utils.r3_trace) _TRACE_ENV (in module nemo_rl.utils.r3_trace) _trace_microbatches() (in module nemo_rl.utils.r3_trace) _TRACE_MICROBATCHES_ENV (in module nemo_rl.utils.r3_trace) _trace_path() (in module nemo_rl.utils.r3_trace) _trace_router_replay_topk_use() (in module nemo_rl.utils.r3_trace) _trace_samples() (in module nemo_rl.utils.r3_trace) _TRACE_SAMPLES_ENV (in module nemo_rl.utils.r3_trace) _trace_steps() (in module nemo_rl.utils.r3_trace) _TRACE_STEPS_ENV (in module nemo_rl.utils.r3_trace) _TRACE_VERIFY_FORWARD_ENV (in module nemo_rl.utils.r3_trace) _train_fields_for_step() (in module nemo_rl.algorithms.grpo_sync) _train_microbatch_body() (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) _train_parallelism() (nemo_rl.weight_sync.nccl_reshard_weight_synchronizer.NcclReshardWeightSynchronizer method) _TRAIN_PREFIXES (in module nemo_rl.telemetry.metrics) _train_pump() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _train_step_state (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl attribute) _transition() (nemo_rl.models.generation.fleet_health.GenerationFleetHealth method) _trim_vocab_padding() (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension static method) _truncate_to_max_size() (nemo_rl.algorithms.async_utils.replay_buffer.ReplayBufferImpl method) _try_merge_byte_buffer() (in module nemo_rl.algorithms.x_token.token_aligner) _try_zero_copy_teacher_logits() (in module nemo_rl.algorithms.x_token.loss_utils) _TYPE_TEMPLATE (in module nemo_rl.data.datasets.response_datasets.intent) _typed_content_media_nbytes() (in module nemo_rl.utils.multimodal_payload_metrics) _typed_content_media_segment_count() (in module nemo_rl.utils.multimodal_payload_metrics) _typed_gym_failure() (in module nemo_rl.environments.nemo_gym) _unanimous_task_index() (in module nemo_rl.algorithms.async_utils.trajectory_collector) _UNICODE_FIXES (in module nemo_rl.algorithms.x_token.token_aligner) _unpack_sequences_from_megatron() (in module nemo_rl.models.megatron.data) _unpack_value_sequences() (in module nemo_rl.models.value.workers.megatron_value_worker) _unrank() (in module nemo_rl.telemetry.setup) _unwrap_model_config() (in module nemo_rl.models.megatron.router_replay) _unwrapped_chunks() (in module nemo_rl.models.policy.workers.megatron_policy_worker) _update_moe_gate_bias_if_supported() (nemo_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl method) _update_weights_from_checkpoint_engine_async() (nemo_rl.models.generation.vllm.checkpoint_engine.VllmCheckpointEngineMixin method) _update_weights_from_collective() (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) _update_worker_weights() (in module nemo_rl.models.generation.dynamo.refit) _use_golden_api() (in module nemo_rl.weight_sync.xferdtensor) _use_python_api() (in module nemo_rl.weight_sync.xferdtensor) _use_real_quant_refit() (nemo_rl.modelopt.models.policy.workers.megatron_quant_policy_worker.MegatronQuantPolicyWorker method) _uses_fp8_kv_cache() (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) _uses_mxfp8_overlap_shared_param_buffer() (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) _uses_native_layerwise_refit() (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) _uses_unquantized_flashinfer_trtllm() (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) _v2 (nemo_rl.models.policy.DTensorConfig attribute) _valid_sample_record() (in module nemo_rl.utils.r3_trace) _validate_algo_settings() (in module nemo_rl.algorithms.single_controller_utils.config) _validate_argv() (nemo_rl.models.generation.dynamo.dynamo_worker.DynamoVllmWorker static method) _validate_backend_boundary() (nemo_rl.models.generation.dynamo.config.DynamoConfig method) _validate_batch_shortfall() (in module nemo_rl.experience.rollout_recovery) _validate_checkpoint_engine_weight_update() (nemo_rl.models.generation.vllm.checkpoint_engine.VllmCheckpointEngineMixin method) (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtensionWithCheckpointEngine method) _validate_chunking_config() (in module nemo_rl.models.megatron.setup) _validate_default_teacher_alias() (in module nemo_rl.algorithms.opd) _validate_dtype_config() (in module nemo_rl.models.megatron.setup) _validate_endpoint_types() (nemo_rl.models.generation.dynamo.config.DynamoWorkerArgs method) _validate_engine_data() (in module nemo_rl.models.generation.dynamo.token_wrapper) _validate_expert_storage() (nemo_rl.models.generation.vllm.refit_loader.VllmShardedExpertRefitMixin static method) _validate_failure_settings() (in module nemo_rl.algorithms.single_controller_utils.config) _validate_group_bounds() (nemo_rl.algorithms.async_utils.staleness_sampler.BaseSampler static method) _validate_init_params() (nemo_rl.experience.rollout_manager.AsyncNemoGymRolloutImpl method) _validate_layout() (in module nemo_rl.weight_sync.xferdtensor_python) _validate_loader_report() (nemo_rl.models.generation.vllm.vllm_sparse_delta.VllmSparseDeltaApplier static method) _validate_local_inputs() (in module nemo_rl.weight_sync.xferdtensor_python) _validate_model_override_conflicts() (in module nemo_rl.models.megatron.setup) _validate_multimodal_dedup_capability() (in module nemo_rl.algorithms.grpo) _validate_native_layerwise_refit() (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) _validate_nvfp4_quantizer_format() (in module nemo_rl.modelopt.utils) _validate_optimizer_config() (in module nemo_rl.models.megatron.setup) _validate_parallelism_and_precision() (nemo_rl.models.generation.dynamo.config.DynamoVllmConfig method) _validate_prompt_identity() (in module nemo_rl.experience.rollout_recovery) _validate_replay_inventory() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _validate_replay_tensor() (in module nemo_rl.models.megatron.router_replay) _validate_sequence_lengths() (nemo_rl.data.packing.algorithms.SequencePacker method) _validate_tensor_consistency() (in module nemo_rl.data.llm_message_utils) _validate_training_config() (in module nemo_rl.models.megatron.setup) _validate_use_kl_in_reward_compat() (in module nemo_rl.algorithms.grpo) _validated_packed_values() (in module nemo_rl.data.llm_message_utils) _validated_w4a16_config() (in module nemo_rl.modelopt.models.generation.vllm_modelopt) _validated_workers() (nemo_rl.models.generation.dynamo.refit.DynamoRefitChannel method) _validation_early_stop_message() (in module nemo_rl.algorithms.grpo) _validation_stop_value() (in module nemo_rl.algorithms.grpo) _value_loss_prepare_fn() (in module nemo_rl.models.value.workers.megatron_value_worker) _value_nbytes() (in module nemo_rl.utils.multimodal_payload_metrics) _value_segment_count() (in module nemo_rl.utils.multimodal_payload_metrics) _value_stage() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _value_train() (nemo_rl.algorithms.single_controller.SingleControllerActor method) _verify_r3_trace_cp_token_alignment() (in module nemo_rl.models.megatron.data) _verify_router_replay_forward_context() (in module nemo_rl.utils.r3_trace) _video_to_image_content() (in module nemo_rl.environments.nemo_gym_video) _VideoConfigValue (in module nemo_rl.environments.nemo_gym_video) _view_key() (in module nemo_rl.models.generation.vllm.vllm_sparse_delta) _violation_counts() (in module nemo_rl.experience.payload) _VIOLATION_COUNTS_KEY (in module nemo_rl.experience.payload) _VLLM_CFG_INAPPLICABLE (in module nemo_rl.models.generation.dynamo.config) _VLLM_CFG_MANAGED_RUNTIME (in module nemo_rl.models.generation.dynamo.config) _VLLM_CFG_MOVED (in module nemo_rl.models.generation.dynamo.config) _VLLM_CFG_STRUCTURAL (in module nemo_rl.models.generation.dynamo.config) _VLLM_CFG_UNSUPPORTED (in module nemo_rl.models.generation.dynamo.config) _VLLM_NCCL_MODULE (in module nemo_rl.distributed.stateless_process_group) _VLLM_PICKLE_LOCK (in module nemo_rl.distributed.stateless_process_group) _vllm_port_for_node_slot() (in module nemo_rl.models.generation.dynamo.worker_pool) _VLLM_SINGLE_RANK_ONLY_FIELDS (in module nemo_rl.models.generation.dynamo.config) _VLLM_UNIQUE_ID_KEY (in module nemo_rl.distributed.stateless_process_group) _VllmNcclUniqueId (class in nemo_rl.distributed.stateless_process_group) _vocab_and_mtp_flops() (in module nemo_rl.utils.flops_formulas) _w13_num_shards_from_state_dict_info() (in module nemo_rl.modelopt.models.generation.vllm_quant_backend) _W4A16_ALGO (in module nemo_rl.modelopt.models.generation.vllm_modelopt) _W4A4_ALGO (in module nemo_rl.modelopt.models.generation.vllm_modelopt) _wait_for_baseline_commits() (nemo_rl.utils.weight_transfer_sparse_codec.DeltaCompressionTracker method) _wait_for_etcd() (nemo_rl.models.generation.dynamo.managed_runtime.ManagedDynamoRuntime method) _wait_for_frontend() (nemo_rl.models.generation.dynamo.managed_runtime.ManagedDynamoRuntime method) _wait_for_port() (nemo_rl.models.generation.dynamo.managed_runtime.ManagedDynamoRuntime method) _wait_for_system_port() (nemo_rl.models.generation.dynamo.dynamo_worker.DynamoVllmWorker method) _wait_read() (nemo_rl.utils.checkpoint_engines.nixl.NIXLCheckpointEngine method) _wake() (nemo_rl.models.generation.megatron.megatron_worker.MegatronGenerationMixin method) _wake_engine() (nemo_rl.models.generation.megatron.megatron_worker.MegatronGenerationMixin method) _WAKE_RETRY_INTERVAL_S (in module nemo_rl.algorithms.async_utils.trajectory_collector) _wake_waits() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector method) _wanted_engine_env() (in module nemo_rl.data_plane.adapters.transfer_queue_env) _warn_if_other_quant_checkpoint_caches() (in module nemo_rl.modelopt.models.policy.workers.megatron_quant_policy_worker) _warn_on_delete_failure() (nemo_rl.utils.checkpoint.CheckpointManager static method) _warn_unauthenticated_refit_server() (in module nemo_rl.models.generation.vllm.vllm_sparse_refit) _warn_unsupported_in_flight_refit_pause_once() (in module nemo_rl.models.generation.interfaces) _WARNED (in module nemo_rl.telemetry.metrics) _watch() (nemo_rl.distributed.refit_watchdog.RefitAbortWatchdog method) _weight_update_errors_are_fatal() (nemo_rl.modelopt.models.generation.vllm_quant_backend.VllmQuantInternalWorkerExtension method) (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) _weight_update_lifecycle() (nemo_rl.modelopt.models.generation.vllm_quant_backend.VllmQuantInternalWorkerExtension method) (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) _weights_tags() (nemo_rl.models.generation.trtllm.trtllm_worker_async.TrtllmAsyncGenerationWorkerImpl class method) _WIRE_TORCH_DTYPES (in module nemo_rl.utils.routed_experts_codec) _without_initial_image_sources() (in module nemo_rl.environments.nemo_gym) _WORKER_FQN (in module nemo_rl.models.generation.dynamo.worker_pool) _WORKER_GROUP_ENV (in module nemo_rl.telemetry.setup) _worker_resource_attributes() (in module nemo_rl.telemetry.setup) _write_back() (nemo_rl.data_plane.worker_mixin.TQWorkerMixin method) _write_back_result_field() (nemo_rl.data_plane.worker_mixin.TQWorkerMixin method) _write_latest_checkpoint_status() (in module nemo_rl.algorithms.grpo) _write_lock (in module nemo_rl.utils.r3_trace) _write_record() (in module nemo_rl.utils.r3_trace) _XFERDTENSOR_PATH_LOGGED (in module nemo_rl.weight_sync.xferdtensor) _xferdtensor_python_impl_v1() (in module nemo_rl.weight_sync.xferdtensor_python) A a2a_experimental (nemo_rl.models.policy.MegatronPeftConfig attribute) abort() (nemo_rl.distributed.refit_watchdog._Abortable method) (nemo_rl.distributed.stateless_process_group.StatelessProcessGroup method) abort_train_step() (nemo_rl.models.policy.tq_policy.TQPolicy method) (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) abort_train_step_presharded() (nemo_rl.data_plane.worker_mixin.TQWorkerMixin method) abort_xferdtensor_python_subcommunicators() (in module nemo_rl.weight_sync.xferdtensor_python) absent_shards() (nemo_rl.models.generation.fleet_health.GenerationFleetHealth method) AbstractPolicyWorker (class in nemo_rl.models.policy.workers.base_policy_worker) accuracy (nemo_rl.algorithms.dpo.DPOValMetrics attribute) (nemo_rl.algorithms.rm.RMValMetrics attribute) ACK (nemo_rl.models.policy.utils.IPCProtocol attribute) ack_timeout_ms (nemo_rl.data_plane.interfaces.DataPlaneConfig attribute) acquire() (nemo_rl.models.generation.fleet_health.HealthyShardSelector method) activation_checkpointing (nemo_rl.models.policy.DTensorConfig attribute) (nemo_rl.models.policy.MegatronConfig attribute) actor_class_fqn (nemo_rl.environments.utils.EnvRegistryEntry attribute) ACTOR_ENVIRONMENT_REGISTRY (in module nemo_rl.distributed.ray_actor_environment_registry) adam_beta1 (nemo_rl.models.policy.MegatronOptimizerConfig attribute) adam_beta2 (nemo_rl.models.policy.MegatronOptimizerConfig attribute) adam_eps (nemo_rl.models.policy.MegatronOptimizerConfig attribute) add() (nemo_rl.algorithms.async_utils.interfaces.ReplayBufferProtocol method) (nemo_rl.algorithms.async_utils.replay_buffer.ReplayBufferImpl method) (nemo_rl.experience.rollouts._NemoGymStreamAccumulator method) (nemo_rl.models.generation.dynamo.arguments._ArgvBuilder method) add_bos (nemo_rl.data.DataConfig attribute) add_eos (nemo_rl.data.DataConfig attribute) add_generation_prompt (nemo_rl.data.DataConfig attribute) add_grpo_token_loss_masks_and_generation_logprobs() (in module nemo_rl.algorithms.grpo) add_locations() (nemo_rl.utils.weight_transfer_sparse_codec._TensorPayloadBuilder method) add_loss_mask_to_message_log() (in module nemo_rl.data.llm_message_utils) add_raw() (nemo_rl.models.generation.dynamo.arguments._ArgvBuilder method) add_ref_logprobs_to_data() (in module nemo_rl.algorithms.dpo) add_remote_agent() (nemo_rl.utils.checkpoint_engines.nixl.NixlAgent method) add_system_prompt (nemo_rl.data.DataConfig attribute) add_values() (nemo_rl.utils.weight_transfer_sparse_codec._TensorPayloadBuilder method) ADDITIONAL_OPTIONAL_KEY_TENSORS (nemo_rl.distributed.batched_data_dict.BatchedDataDict attribute) address() (nemo_rl.distributed.held_port.HeldPortReservation method) admission_id (nemo_rl.experience.rollout_recovery.PromptGroupRecoveryRecord attribute) (nemo_rl.experience.rollout_recovery.PromptGroupRecoveryState attribute) admit() (nemo_rl.algorithms.async_utils.staleness_sampler._GatedSampler method) (nemo_rl.algorithms.async_utils.staleness_sampler.BaseSampler method) (nemo_rl.algorithms.async_utils.staleness_sampler.PromptGroupSampler method) (nemo_rl.algorithms.async_utils.staleness_sampler.WindowedSampler method) ADMITTED (nemo_rl.experience.rollout_recovery.PromptGroupPhase attribute) adv_estimator (nemo_rl.algorithms.grpo.GRPOConfig attribute) (nemo_rl.algorithms.ppo.PPOConfig attribute) ADVANTAGE (nemo_rl.telemetry.span_groups.RLSpanGroup attribute) advantage_clip_high (nemo_rl.algorithms.grpo.GRPOConfig attribute) advantage_clip_low (nemo_rl.algorithms.grpo.GRPOConfig attribute) advantage_estimator (nemo_rl.algorithms.single_controller_utils.setup.SingleControllerActorArgs attribute) AdvantageConfig (class in nemo_rl.algorithms.single_controller_utils.config) advantages (nemo_rl.algorithms.loss.loss_functions.ClippedPGLossDataDict attribute) AdvEstimatorConfig (class in nemo_rl.algorithms.advantage_estimator) aggregate_per_sample_handles() (in module nemo_rl.models.policy.utils) aggregate_rollout_metrics() (in module nemo_rl.algorithms.grpo) aggregate_spec_decode_counters() (in module nemo_rl.models.generation.vllm.utils) aggregate_step_metrics() (in module nemo_rl.algorithms.single_controller_utils.utils) aggregate_training_statistics() (in module nemo_rl.models.automodel.train) (in module nemo_rl.models.megatron.train) AIMEDataset (class in nemo_rl.data.datasets.response_datasets.aime) AIMEEvalDataConfig (class in nemo_rl.data) AIMEVariant (in module nemo_rl.data.datasets.response_datasets.aime) algo_config() (in module nemo_rl.algorithms.single_controller_utils.config) algorithm (nemo_rl.distributed.batched_data_dict.SequencePackingArgs attribute) (nemo_rl.models.policy.SequencePackingConfig attribute) alias (nemo_rl.models.policy.teacher_worker_group.TeacherConfig attribute) alias_to_group_alias (nemo_rl.algorithms.single_controller_utils.setup.SingleControllerActorArgs attribute) align() (nemo_rl.algorithms.x_token.token_aligner.TokenAligner method) alignment_from_flat_batch() (in module nemo_rl.algorithms.x_token.loss_utils) AlignmentBatch (class in nemo_rl.algorithms.x_token.token_aligner) AlignmentPair (class in nemo_rl.algorithms.x_token.token_aligner) all_gather() (nemo_rl.distributed.batched_data_dict.BatchedDataDict method) ALL_GROUPS (nemo_rl.telemetry.span_groups.RLSpanGroup attribute) all_to_all_sq2vp() (in module nemo_rl.distributed.model_utils) all_to_all_vp2sq() (in module nemo_rl.distributed.model_utils) allgather_cp_sharded_tensor() (in module nemo_rl.distributed.model_utils) AllGatherCPTensor (class in nemo_rl.distributed.model_utils) allow_flash_attn_args (nemo_rl.models.automodel.config.RuntimeConfig attribute) AllTaskProcessedDataset (class in nemo_rl.data.datasets.processed_dataset) alpha (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute) (nemo_rl.models.policy.LoRAConfig attribute) (nemo_rl.models.policy.MegatronPeftConfig attribute) answers (nemo_rl.environments.interfaces.EnvironmentReturn attribute) apply_batch_size (nemo_rl.models.generation.vllm.config.VllmRefitTuningConfig attribute) apply_queue_depth (nemo_rl.models.generation.vllm.config.VllmRefitTuningConfig attribute) apply_reward_penalties() (in module nemo_rl.experience.rollouts) apply_reward_shaping() (in module nemo_rl.algorithms.reward_functions) apply_rope_fusion (nemo_rl.models.policy.MegatronConfig attribute) apply_temperature_scaling() (in module nemo_rl.models.automodel.train) (in module nemo_rl.models.megatron.train) apply_to() (nemo_rl.models.megatron.draft.utils._PendingLayerWeights method) apply_top_k_top_p() (in module nemo_rl.algorithms.logits_sampling_utils) apply_top_k_top_p_filtering_for_local_logits() (in module nemo_rl.models.automodel.train) apply_transformer_engine_patch() (in module nemo_rl.models.policy.workers.patches) armed (nemo_rl.distributed.refit_watchdog.RefitAbortWatchdog property) ArrowTextDataset (class in nemo_rl.data.datasets.response_datasets.arrow_text_dataset) artifact_location (nemo_rl.utils.logger.MLflowConfig attribute) as_metrics() (nemo_rl.experience.rollout_manager.RolloutStats method) (nemo_rl.models.generation.fleet_health.GenerationFleetHealth method) as_tensor() (nemo_rl.data.multimodal_utils.PackedTensor method) assemble_teacher_logits_from_shards() (in module nemo_rl.algorithms.x_token.loss_utils) assert_no_double_bos() (in module nemo_rl.data.datasets.utils) assert_prev_logprobs_available() (in module nemo_rl.algorithms.opd) assert_teacher_student_batch_grid() (in module nemo_rl.algorithms.x_token.utils) assert_xtoken_ipc_node_local() (in module nemo_rl.algorithms.x_token.utils) async_engine (nemo_rl.models.generation.dynamo.config.DynamoVllmConfig attribute) (nemo_rl.models.generation.trtllm.config.TrtllmSpecificArgs attribute) (nemo_rl.models.generation.vllm.config.VllmSpecificArgs attribute) async_generate_response_for_sample_turn() (in module nemo_rl.experience.rollouts) async_grpo (nemo_rl.algorithms.grpo.GRPOConfig attribute) async_grpo_train() (in module nemo_rl.algorithms.grpo) async_http_post_json() (in module nemo_rl.models.generation.dynamo.http_client) async_ppo (nemo_rl.algorithms.ppo.PPOConfig attribute) async_ppo_train() (in module nemo_rl.algorithms.ppo) async_rl (nemo_rl.algorithms.single_controller_utils.config.MasterConfig attribute) async_save (nemo_rl.models.policy.MegatronCheckpointConfig attribute) AsyncGRPOConfig (class in nemo_rl.algorithms.grpo) AsyncNemoGymRolloutImpl (class in nemo_rl.experience.rollout_manager) AsyncPPOConfig (class in nemo_rl.algorithms.ppo) AsyncRLConfig (class in nemo_rl.algorithms.single_controller_utils.config) AsyncRolloutImpl (class in nemo_rl.experience.rollout_manager) AsyncTrajectoryCollector (class in nemo_rl.algorithms.async_utils.trajectory_collector) attach_fleet_health() (nemo_rl.models.generation.interfaces.GenerationInterface method) (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) attach_image_model_inputs_to_message() (in module nemo_rl.data.multimodal_utils) attach_initial_nemo_gym_image_payloads() (in module nemo_rl.experience.rollouts) attach_media_token_validity_mask() (in module nemo_rl.data.multimodal_utils) attach_message_log_view() (in module nemo_rl.data.llm_message_utils) attach_routed_experts_to_chat_response_choices() (in module nemo_rl.models.generation.vllm.utils) attach_static_multimodal_payload() (in module nemo_rl.experience.rollouts) attach_token_information_to_chat_response_choices() (in module nemo_rl.models.generation.vllm.utils) attention_backend (nemo_rl.models.policy.MegatronConfig attribute) attention_heads (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) attention_mask (nemo_rl.models.automodel.data.ProcessedInputs attribute) (nemo_rl.models.megatron.data.ProcessedInputs attribute) (nemo_rl.models.megatron.data.ProcessedMicrobatch attribute) attn (nemo_rl.models.policy.AutomodelBackendConfig attribute) attn_impl (nemo_rl.models.automodel.config.RuntimeConfig attribute) audio (nemo_rl.models.policy.TokenizerConfig attribute) AUDIO_CONTENT_TYPES (in module nemo_rl.data.multimodal_utils) AUDIOMCQ_MANIFEST (in module nemo_rl.data.datasets.response_datasets.audiomcq) AUDIOMCQ_REPO_ID (in module nemo_rl.data.datasets.response_datasets.audiomcq) AudioMCQDataset (class in nemo_rl.data.datasets.response_datasets.audiomcq) autocast_enabled (nemo_rl.models.automodel.config.ModelAndOptimizerState attribute) AUTOMODEL (nemo_rl.distributed.virtual_cluster.PY_EXECUTABLES attribute) AUTOMODEL_FACTORY (in module nemo_rl.models.policy.utils) automodel_forward_backward() (in module nemo_rl.models.automodel.train) automodel_kwargs (nemo_rl.models.policy.DTensorConfig attribute) AutomodelBackendConfig (class in nemo_rl.models.policy) AutomodelCheckpointManager (class in nemo_rl.models.automodel.checkpoint) AutomodelFreezeConfig (class in nemo_rl.models.policy) AutomodelKwargs (class in nemo_rl.models.policy) aux_layer_indices (nemo_rl.models.policy.DraftConfig attribute) AVQADataset (class in nemo_rl.data.datasets.response_datasets.avqa) await_off_loop() (in module nemo_rl.distributed.refit_watchdog) B backend (nemo_rl.data_plane.interfaces.DataPlaneConfig attribute) (nemo_rl.models.generation.dynamo.config.DynamoConfig attribute) (nemo_rl.models.generation.interfaces.CheckpointEngineConfig attribute) (nemo_rl.models.generation.interfaces.GenerationConfig attribute) (nemo_rl.models.policy.AutomodelKwargs attribute) backend_config() (in module nemo_rl.data_plane.interfaces) backend_init_params (nemo_rl.models.generation.vllm.config.VllmNixlRefitConfig attribute) backend_name (nemo_rl.models.generation.vllm.config.VllmNixlRefitConfig attribute) backend_timeout_s (nemo_rl.algorithms.single_controller_utils.config.GenerationRouterConfig attribute) backfill_missing_routed_experts() (in module nemo_rl.experience.rollouts) backoff_base_s (nemo_rl.algorithms.single_controller_utils.config.RolloutFailureConfig attribute) (nemo_rl.experience.rollout_manager.RolloutRetryPolicy attribute) backoff_for() (nemo_rl.experience.rollout_manager.RolloutRetryPolicy method) backward() (nemo_rl.algorithms.logits_sampling_utils._ApplyTopKTopP static method) (nemo_rl.algorithms.x_token.loss_utils.Fp32SparseMM static method) (nemo_rl.distributed.model_utils._AllReduceSum static method) (nemo_rl.distributed.model_utils.AllGatherCPTensor method) (nemo_rl.distributed.model_utils.ChunkedDistributedEntropy static method) (nemo_rl.distributed.model_utils.ChunkedDistributedGatherLogprob static method) (nemo_rl.distributed.model_utils.ChunkedDistributedHiddenStatesToLogprobs static method) (nemo_rl.distributed.model_utils.ChunkedDistributedLogprob static method) (nemo_rl.distributed.model_utils.ChunkedDistributedLogprobWithSampling static method) (nemo_rl.distributed.model_utils.DistributedCrossEntropy static method) (nemo_rl.distributed.model_utils.DistributedLogprob static method) (nemo_rl.distributed.model_utils.DistributedLogprobWithSampling static method) bad_words (nemo_rl.models.generation.interfaces.GenerationConfig attribute) BASE (nemo_rl.distributed.virtual_cluster.PY_EXECUTABLES attribute) base (nemo_rl.weight_sync.nccl_reshard_utils.LocalParamSpec attribute) base_url (nemo_rl.models.generation.fleet_health.ShardHealth attribute) base_url() (nemo_rl.models.generation.generation_router.GenerationRouterImpl method) base_urls (nemo_rl.environments.nemo_gym.NemoGymConfig attribute) baseline (nemo_rl.models.generation.vllm.config.VllmSparseRefitConfig attribute) BaseMathEnvironment (class in nemo_rl.environments.math_environment) BaseSampler (class in nemo_rl.algorithms.async_utils.staleness_sampler) BaseVllmGenerationWorker (class in nemo_rl.models.generation.vllm.vllm_worker) batch_multiplier (nemo_rl.algorithms.grpo.GRPOConfig attribute) (nemo_rl.algorithms.ppo.PPOConfig attribute) batch_shortfall (nemo_rl.experience.rollout_recovery.ParsedRolloutRecoveryState attribute) (nemo_rl.experience.rollout_recovery.RolloutRecoveryState attribute) batch_size (nemo_rl.environments.reward_model_environment.RewardModelEnvironmentConfig attribute) batched_message_log_to_flat_message() (in module nemo_rl.data.llm_message_utils) BatchedDataDict (class in nemo_rl.distributed.batched_data_dict) bbox_giou_reward() (in module nemo_rl.environments.rewards) begin_finalization() (nemo_rl.utils.checkpoint.CheckpointManager method) begin_train_step() (nemo_rl.models.policy.tq_policy.TQPolicy method) (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) begin_train_step_presharded() (nemo_rl.data_plane.worker_mixin.TQWorkerMixin method) bert() (in module nemo_rl.utils.flops_formulas) bf16 (nemo_rl.models.policy.MegatronOptimizerConfig attribute) bias_activation_fusion (nemo_rl.models.policy.MegatronConfig attribute) BinaryPreferenceDataset (class in nemo_rl.data.datasets.preference_datasets.binary_preference_dataset) bind_numa() (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) bind_runtime_prompt() (nemo_rl.experience.rollout_recovery.RolloutRecoveryLedger method) bind_to_gpu_numa() (in module nemo_rl.distributed.numa_utils) block_size_tokens (nemo_rl.models.generation.megatron.config.MCoreGenerationSpecificArgs attribute) blocks_training() (nemo_rl.models.generation.interfaces.GenerationInterface method) (nemo_rl.models.generation.megatron.megatron_generation.MegatronGeneration method) boxed (in module nemo_rl.environments.rewards) broadcast() (nemo_rl.distributed.stateless_process_group.StatelessProcessGroup method) broadcast_hf_buckets_via_distributed_impl() (in module nemo_rl.models.policy.utils) broadcast_loss_metrics_from_last_stage() (in module nemo_rl.models.megatron.pipeline_parallel) broadcast_obj_from_pp_rank() (in module nemo_rl.models.megatron.pipeline_parallel) broadcast_tensor() (in module nemo_rl.models.megatron.common) broadcast_tensors_from_last_stage() (in module nemo_rl.models.megatron.pipeline_parallel) broadcast_weights_for_collective() (nemo_rl.models.policy.interfaces.ColocatablePolicyInterface method) (nemo_rl.models.policy.lm_policy.Policy method) (nemo_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) (nemo_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl method) (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) Bucket (class in nemo_rl.telemetry.instrumentation) bucket_for_efficiency_category() (in module nemo_rl.telemetry.instrumentation) bucket_for_span_group() (in module nemo_rl.telemetry.instrumentation) bucket_scope() (in module nemo_rl.telemetry.instrumentation) buf (nemo_rl.weight_sync.nccl_reshard_utils.RefitCtx attribute) buffer() (nemo_rl.data_plane.adapters.transfer_queue._StagingPool method) buffer_size_bytes (nemo_rl.models.generation.interfaces.CollectiveSenderSpec attribute) buffer_size_gb (nemo_rl.models.generation.megatron.config.MCoreGenerationSpecificArgs attribute) build_app() (nemo_rl.models.generation.generation_router.GenerationRouterImpl method) build_cached_video_frame_data_url() (in module nemo_rl.models.generation.vllm.video_utils) build_cached_video_frame_metadata() (in module nemo_rl.models.generation.vllm.video_utils) build_data_plane_client() (in module nemo_rl.data_plane.factory) build_draft_model() (in module nemo_rl.models.megatron.draft.utils) build_dynamo_frontend_argv() (in module nemo_rl.models.generation.dynamo.arguments) build_dynamo_vllm_argv() (in module nemo_rl.models.generation.dynamo.arguments) build_exact_token_map() (in module nemo_rl.algorithms.x_token.loss_utils) build_hf_to_local_param_map() (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) (nemo_rl.weight_sync.nccl_reshard_utils.RefitBuilderInterface method) build_inference_model() (in module nemo_rl.models.megatron.setup) build_managed_worker_env() (in module nemo_rl.models.generation.dynamo.arguments) build_media_token_validity_mask() (in module nemo_rl.data.multimodal_utils) build_mesh_info() (in module nemo_rl.weight_sync.nccl_reshard_utils) build_nccl_reshard_refit_info() (in module nemo_rl.weight_sync.nccl_reshard_utils) build_reward_component_columns() (in module nemo_rl.environments.nemo_gym) build_rollout_recovery_state() (in module nemo_rl.experience.rollout_recovery) build_router_replay_assignments() (in module nemo_rl.models.megatron.router_replay) build_router_replay_tensors() (in module nemo_rl.models.megatron.router_replay) build_vllm_modelopt_nvfp4_config() (in module nemo_rl.modelopt.utils) bytes_outstanding (nemo_rl.data_plane.observability.DataPlaneStats attribute) bytes_outstanding_by_partition() (nemo_rl.data_plane.observability.MetricsDataPlaneClient method) C calculate_advantages_on_gpu (nemo_rl.algorithms.grpo.GRPOConfig attribute) calculate_aligned_size() (in module nemo_rl.models.policy.utils) calculate_baseline_and_std_per_prompt() (in module nemo_rl.algorithms.utils) calculate_kl() (in module nemo_rl.algorithms.utils) calculate_pass_rate_per_prompt() (in module nemo_rl.environments.metrics) calculate_rewards() (in module nemo_rl.experience.rollouts) calculate_single_metric() (in module nemo_rl.experience.metric_utils) calculate_stats_only() (nemo_rl.data.packing.metrics.PackingMetrics method) calibrate_qkv_fp8_scales() (nemo_rl.models.policy.interfaces.PolicyInterface method) (nemo_rl.models.policy.lm_policy.Policy method) (nemo_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) (nemo_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl method) (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) call_data_plane() (in module nemo_rl.data_plane.async_utils) callback (nemo_rl.data_plane.interfaces.ObservabilityConfig attribute) called_workers (nemo_rl.distributed.worker_groups.MultiWorkerFuture attribute) CANONICAL_LOGGER_ALIASES (in module nemo_rl.models.generation.dynamo.metrics) canonical_token() (in module nemo_rl.algorithms.x_token.token_aligner) cap_max_tokens_to_context (nemo_rl.models.generation.vllm.config.VllmSpecificArgs attribute) capture_context() (nemo_rl.models.megatron.draft.hidden_capture.HiddenStateCapture method) CapturedStates (class in nemo_rl.models.megatron.draft.hidden_capture) causal_self_attn (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) ce_label_mask() (in module nemo_rl.algorithms.x_token.loss_utils) ce_loss_scale (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute) chat_template (nemo_rl.models.policy.TokenizerConfig attribute) chat_template_kwargs (nemo_rl.models.policy.TokenizerConfig attribute) chdir() (nemo_rl.environments.code_environment.CodeExecutionWorker method) check_consumption_status() (nemo_rl.data_plane.adapters.noop.NoOpDataPlaneClient method) (nemo_rl.data_plane.adapters.transfer_queue.TQDataPlaneClient method) (nemo_rl.data_plane.interfaces.DataPlaneClient method) (nemo_rl.data_plane.observability.MetricsDataPlaneClient method) check_health() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector method) check_nccl_reshard_refit_support() (in module nemo_rl.weight_sync.nccl_reshard_utils) check_save() (nemo_rl.utils.timer.TimeoutChecker method) check_sequence_dim() (in module nemo_rl.models.automodel.data) check_tensor_parallel_attributes() (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) check_vocab_equality() (in module nemo_rl.algorithms.distillation) checkpoint (nemo_rl.models.policy.MegatronConfig attribute) checkpoint() (nemo_rl.algorithms.async_utils.replay_buffer.DataPlaneCheckpointBarrier method) checkpoint_dir (nemo_rl.utils.checkpoint.CheckpointingConfig attribute) checkpoint_engine (nemo_rl.models.generation.vllm.checkpoint_engine.VllmCheckpointEngineMixin attribute) (nemo_rl.models.policy.workers.checkpoint_engine.PolicyCheckpointEngineMixin attribute) checkpoint_engine_refit_config() (in module nemo_rl.weight_sync.checkpoint_engine_config) checkpoint_engine_rpc() (nemo_rl.models.generation.vllm.checkpoint_engine.VllmCheckpointEngineRpcMixin method) (nemo_rl.models.policy.workers.checkpoint_engine.PolicyCheckpointEngineMixin method) checkpoint_engine_rpc_async() (nemo_rl.models.generation.vllm.checkpoint_engine.VllmAsyncCheckpointEngineRpcMixin method) checkpoint_engine_total_memory_bytes() (nemo_rl.models.generation.vllm.checkpoint_engine.VllmCheckpointEngineMixin method) checkpoint_must_save_by (nemo_rl.utils.checkpoint.CheckpointingConfig attribute) checkpoint_path (nemo_rl.environments.reward_model_environment.RewardModelEnvironmentConfig attribute) checkpoint_prefix (nemo_rl.models.megatron.draft.utils._EagleLayerLayout attribute) CheckpointEngine (class in nemo_rl.utils.checkpoint_engines.base) CheckpointEngineConfig (class in nemo_rl.models.generation.interfaces) CheckpointEngineWeightSynchronizer (class in nemo_rl.weight_sync.checkpoint_engine_weight_synchronizer) checkpointing (nemo_rl.algorithms.distillation.MasterConfig attribute) (nemo_rl.algorithms.dpo.MasterConfig attribute) (nemo_rl.algorithms.grpo.MasterConfig attribute) (nemo_rl.algorithms.ppo.MasterConfig attribute) (nemo_rl.algorithms.rm.MasterConfig attribute) (nemo_rl.algorithms.sft.MasterConfig attribute) (nemo_rl.algorithms.single_controller_utils.config.MasterConfig attribute) (nemo_rl.algorithms.xtoken_off_policy_distillation.MasterConfig attribute) checkpointing_context (nemo_rl.models.megatron.config.ModelAndOptimizerState attribute) CheckpointingConfig (class in nemo_rl.utils.checkpoint) CheckpointLoader (in module nemo_rl.models.megatron.draft.utils) CheckpointManager (class in nemo_rl.utils.checkpoint) checksums (nemo_rl.utils.weight_transfer_zmq._RelayTransfer attribute) chosen_key (nemo_rl.data.PreferenceDatasetConfig attribute) chunk() (nemo_rl.distributed.batched_data_dict.BatchedDataDict method) chunk_average_finalize() (in module nemo_rl.algorithms.x_token.loss_utils) chunk_average_log_probs() (in module nemo_rl.algorithms.x_token.loss_utils) chunk_list_to_workers() (in module nemo_rl.environments.utils) chunk_log_prob_sums() (in module nemo_rl.algorithms.x_token.loss_utils) chunk_offset (nemo_rl.utils.checkpoint_engines.base.TensorMeta attribute) chunk_size (nemo_rl.utils.checkpoint_engines.base.TensorMeta attribute) ChunkedDistributedEntropy (class in nemo_rl.distributed.model_utils) ChunkedDistributedGatherLogprob (class in nemo_rl.distributed.model_utils) ChunkedDistributedHiddenStatesToLogprobs (class in nemo_rl.distributed.model_utils) ChunkedDistributedLogprob (class in nemo_rl.distributed.model_utils) ChunkedDistributedLogprobWithSampling (class in nemo_rl.distributed.model_utils) chunks_accept_media_token_validity_mask() (in module nemo_rl.data.multimodal_utils) ckpt_assume_constant_structure (nemo_rl.models.policy.MegatronCheckpointConfig attribute) ckpt_fully_parallel_load_exchange_algo (nemo_rl.models.policy.MegatronCheckpointConfig attribute) ckpt_fully_parallel_load_process_group (nemo_rl.models.policy.MegatronCheckpointConfig attribute) ckpt_fully_parallel_save_process_group (nemo_rl.models.policy.MegatronCheckpointConfig attribute) claim_meta() (nemo_rl.data_plane.adapters.noop.NoOpDataPlaneClient method) (nemo_rl.data_plane.adapters.transfer_queue.TQDataPlaneClient method) (nemo_rl.data_plane.interfaces.DataPlaneClient method) (nemo_rl.data_plane.observability.MetricsDataPlaneClient method) claim_meta_poll_interval_s (nemo_rl.data_plane.interfaces.DataPlaneConfig attribute) class_token_len (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) classify_rollout_failure() (in module nemo_rl.experience.failures) cleanup (nemo_rl.utils.weight_transfer_stream.SparseRefitTransport attribute) cleanup() (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) cleanup_process_group() (nemo_rl.models.generation.dynamo.dynamo_worker.DynamoGpuReservation method) cleanup_zmq() (nemo_rl.models.generation.trtllm.trtllm_backend.NcclExtension method) clear() (nemo_rl.algorithms.async_utils.interfaces.ReplayBufferProtocol method) (nemo_rl.algorithms.async_utils.replay_buffer.ReplayBufferImpl method) (nemo_rl.models.generation.dynamo.metrics.DynamoMetricsSampler method) clear_cache_every_n_steps (nemo_rl.models.policy.DTensorConfig attribute) clear_global_router_replay_instances() (in module nemo_rl.models.megatron.router_replay) clear_hooks() (nemo_rl.models.megatron.draft.hidden_capture.HiddenStateCapture method) clear_logger_metrics() (nemo_rl.models.generation.dynamo.dynamo_generation.DynamoGeneration method) (nemo_rl.models.generation.interfaces.GenerationInterface method) (nemo_rl.models.generation.trtllm.trtllm_generation.TrtllmGeneration method) (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) clear_memory_caches_before_refit (nemo_rl.models.policy.MegatronConfig attribute) clear_router_replay() (in module nemo_rl.models.megatron.router_replay) clear_samples() (nemo_rl.data_plane.adapters.noop.NoOpDataPlaneClient method) (nemo_rl.data_plane.adapters.transfer_queue.TQDataPlaneClient method) (nemo_rl.data_plane.interfaces.DataPlaneClient method) (nemo_rl.data_plane.observability.MetricsDataPlaneClient method) clear_vllm_logger_metrics() (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) (nemo_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) clear_xferdtensor_python_caches() (in module nemo_rl.weight_sync.xferdtensor_python) CLEVRCoGenTDataset (class in nemo_rl.data.datasets.response_datasets.clevr) client_copy() (nemo_rl.models.generation.dynamo.refit.DynamoRefitChannel method) clip_grad (nemo_rl.models.policy.MegatronOptimizerConfig attribute) clip_grad_by_total_norm_() (in module nemo_rl.models.dtensor.parallelize) ClippedPGLossConfig (class in nemo_rl.algorithms.loss.loss_functions) ClippedPGLossDataDict (class in nemo_rl.algorithms.loss.loss_functions) ClippedPGLossFn (class in nemo_rl.algorithms.loss.loss_functions) cliprange (nemo_rl.algorithms.loss.loss_functions.MseValueLossConfig attribute) close() (nemo_rl.data_plane.adapters.noop.NoOpDataPlaneClient method) (nemo_rl.data_plane.adapters.transfer_queue.TQDataPlaneClient method) (nemo_rl.data_plane.interfaces.DataPlaneClient method) (nemo_rl.data_plane.observability.MetricsDataPlaneClient method) (nemo_rl.utils.weight_transfer_zmq.ZmqSparseRefitClient method) (nemo_rl.utils.weight_transfer_zmq.ZmqSparseRefitServer method) cluster (nemo_rl.algorithms.distillation.MasterConfig attribute) (nemo_rl.algorithms.dpo.MasterConfig attribute) (nemo_rl.algorithms.grpo.MasterConfig attribute) (nemo_rl.algorithms.ppo.MasterConfig attribute) (nemo_rl.algorithms.rm.MasterConfig attribute) (nemo_rl.algorithms.sft.MasterConfig attribute) (nemo_rl.algorithms.single_controller_utils.config.MasterConfig attribute) (nemo_rl.algorithms.xtoken_off_policy_distillation.MasterConfig attribute) (nemo_rl.evals.eval.MasterConfig attribute) ClusterConfig (class in nemo_rl.distributed.virtual_cluster) CodeEnvConfig (class in nemo_rl.environments.code_environment) CodeEnvironment (class in nemo_rl.environments.code_environment) CodeEnvMetadata (class in nemo_rl.environments.code_environment) CodeExecutionWorker (class in nemo_rl.environments.code_environment) CodeJaccardEnvConfig (class in nemo_rl.environments.code_jaccard_environment) CodeJaccardEnvironment (class in nemo_rl.environments.code_jaccard_environment) CodeJaccardEnvironmentMetadata (class in nemo_rl.environments.code_jaccard_environment) CodeJaccardVerifyWorker (class in nemo_rl.environments.code_jaccard_environment) collect_multimodal_payload_metrics() (in module nemo_rl.utils.multimodal_payload_metrics) collect_overlapping_teacher_shards() (in module nemo_rl.algorithms.x_token.loss_utils) collect_sharded_multimodal_payload_metrics() (in module nemo_rl.utils.multimodal_payload_metrics) collection_interval (nemo_rl.utils.logger.GPUMonitoringConfig attribute) collective_init_time_s (nemo_rl.algorithms.metric_utils.SetupTimingMetrics attribute) CollectiveSenderSpec (class in nemo_rl.models.generation.interfaces) CollectiveWeightSynchronizer (class in nemo_rl.weight_sync.collective_weight_synchronizer) COLLECTOR_LOOP_CATEGORIES (in module nemo_rl.telemetry.instrumentation) COLLECTOR_WALL_CLOCK_MEASUREMENT (in module nemo_rl.telemetry.metrics) ColocatablePolicyInterface (class in nemo_rl.models.policy.interfaces) colocated (nemo_rl.models.generation.interfaces.GenerationConfig attribute) colocated_reshard_plan (nemo_rl.models.megatron.config.ModelAndOptimizerState attribute) ColocatedReshardPlan (class in nemo_rl.models.megatron.config) ColocationConfig (class in nemo_rl.models.generation.interfaces) COLUMN_PARALLEL_SUFFIXES (in module nemo_rl.weight_sync.nccl_reshard_utils) ColwiseParallelWithGather (class in nemo_rl.models.dtensor.parallelize) combine_reward_functions() (in module nemo_rl.environments.rewards) commit() (nemo_rl.algorithms.async_utils.replay_buffer.TQReplayBuffer method) commit_admission() (nemo_rl.algorithms.async_utils.staleness_sampler._GatedSampler method) (nemo_rl.algorithms.async_utils.staleness_sampler.TransactionalAdmissionSampler method) (nemo_rl.algorithms.async_utils.staleness_sampler.WindowedSampler method) COMMITTED (nemo_rl.experience.rollout_manager.RolloutOutcome attribute) committed (nemo_rl.experience.rollout_manager.RolloutStats attribute) COMMON_CHAT_TEMPLATES (class in nemo_rl.data.chat_templates) compile_or_warm_up_model() (nemo_rl.modelopt.models.generation.vllm_quant_patch.FakeQuantWorker method) COMPLETE (nemo_rl.models.policy.utils.IPCProtocol attribute) Completion (class in nemo_rl.experience.interfaces) completions (nemo_rl.experience.interfaces.PromptGroupRecord attribute) compute_advantage() (nemo_rl.algorithms.advantage_estimator.GDPOAdvantageEstimator method) (nemo_rl.algorithms.advantage_estimator.GeneralizedAdvantageEstimator method) (nemo_rl.algorithms.advantage_estimator.GRPOAdvantageEstimator method) (nemo_rl.algorithms.advantage_estimator.OPDAdvantageEstimator method) (nemo_rl.algorithms.advantage_estimator.RawRewardAdvantageEstimator method) (nemo_rl.algorithms.advantage_estimator.ReinforcePlusPlusAdvantageEstimator method) compute_and_apply_seq_logprob_error_masking() (in module nemo_rl.algorithms.grpo) compute_metrics() (nemo_rl.data.packing.algorithms.SequencePacker method) compute_score() (in module nemo_rl.environments.dapo_math_verifier) compute_spec_decode_metrics() (in module nemo_rl.models.generation.vllm.utils) concat() (nemo_rl.data.multimodal_utils.PackedTensor class method) (nemo_rl.data_plane.interfaces.KVBatchMeta method) CONCATENATIVE (nemo_rl.data.packing.algorithms.PackingAlgorithm attribute) ConcatenativePacker (class in nemo_rl.data.packing.algorithms) condemn_silent_participant() (nemo_rl.models.generation.fleet_health.GenerationFleetHealth method) configure_dynamo_cache() (in module nemo_rl.models.policy.utils) configure_engine_env() (in module nemo_rl.data_plane.adapters.transfer_queue_env) configure_generation_config() (in module nemo_rl.models.generation) configure_nixl_worker() (in module nemo_rl.models.generation.vllm.checkpoint_engine) configure_refit_environment() (in module nemo_rl.models.megatron.setup) configure_rich_logging() (in module nemo_rl.utils.logger) configure_tree() (nemo_rl.utils.weight_transfer_zmq.ZmqSparseRefitServer method) configure_vllm_for_router_replay() (in module nemo_rl.models.megatron.router_replay) configure_worker() (nemo_rl.models.generation.trtllm.trtllm_worker_async.TrtllmAsyncGenerationWorkerImpl static method) (nemo_rl.models.generation.vllm.vllm_worker.BaseVllmGenerationWorker static method) (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl static method) (nemo_rl.models.value.workers.megatron_value_worker.MegatronValueWorkerImpl static method) configure_zmq_sparse_refit_relay() (nemo_rl.models.generation.vllm.vllm_sparse_refit.VllmSparseRefitReceiver method) (nemo_rl.models.generation.vllm.vllm_worker.BaseVllmGenerationWorker method) connect_colocate_topology() (in module nemo_rl.models.policy.utils) connect_rollout_engines_from_distributed() (in module nemo_rl.models.policy.utils) connect_sglang_rollout_engines() (nemo_rl.models.policy.interfaces.ColocatablePolicyInterface method) (nemo_rl.models.policy.lm_policy.Policy method) (nemo_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker method) connect_sglang_rollout_engines_distributed() (nemo_rl.models.policy.interfaces.ColocatablePolicyInterface method) (nemo_rl.models.policy.lm_policy.Policy method) (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) connect_timeout_s (nemo_rl.algorithms.single_controller_utils.config.GenerationRouterConfig attribute) consecutive_probe_failures (nemo_rl.models.generation.fleet_health.ShardHealth attribute) consecutive_probe_successes (nemo_rl.models.generation.fleet_health.ShardHealth attribute) consecutive_reported_failures (nemo_rl.models.generation.fleet_health.ShardHealth attribute) consumed (nemo_rl.data_plane.adapters.noop._Partition attribute) consumed_samples (nemo_rl.algorithms.distillation.DistillationSaveState attribute) (nemo_rl.algorithms.dpo.DPOSaveState attribute) (nemo_rl.algorithms.grpo.GRPOSaveState attribute) (nemo_rl.algorithms.ppo.PPOSaveState attribute) (nemo_rl.algorithms.rm.RMSaveState attribute) (nemo_rl.algorithms.sft.SFTSaveState attribute) (nemo_rl.algorithms.xtoken_off_policy_distillation.OffPolicyDistillationSaveState attribute) consumer_tasks (nemo_rl.data_plane.adapters.noop._Partition attribute) context (nemo_rl.environments.code_environment.CodeEnvMetadata attribute) context_parallel_size (nemo_rl.algorithms.opd.TeacherResourceConfig attribute) (nemo_rl.models.policy.DTensorConfig attribute) (nemo_rl.models.policy.MegatronConfig attribute) (nemo_rl.models.policy.teacher_worker_group.TeacherConfig attribute) continue_generation() (nemo_rl.models.generation.interfaces.GenerationInterface method) control_timeout_s (nemo_rl.models.generation.dynamo.config.DynamoCfg attribute) controller_address (nemo_rl.data_plane.interfaces.DataPlaneConfig attribute) conversation_process_message() (in module nemo_rl.data.datasets.response_datasets.general_conversations_dataset) conversation_sender_mapping_sample_to_allowed (in module nemo_rl.data.datasets.response_datasets.general_conversations_dataset) convert_config_to_flops_config() (in module nemo_rl.utils.flops_tracker) convert_dcp_to_hf() (in module nemo_rl.utils.native_checkpoint) convert_metadata() (in module nemo_rl.data.datasets.response_datasets.general_conversations_dataset) convert_to_seconds() (in module nemo_rl.utils.timer) copy_defaults() (nemo_rl.data.interfaces.TaskDataSpec method) copy_policy_lm_head_to_draft() (in module nemo_rl.models.megatron.draft.utils) count_for_target_step() (nemo_rl.algorithms.async_utils.replay_buffer.TQReplayBuffer method) counts_by_state() (nemo_rl.models.generation.fleet_health.GenerationFleetHealth method) cp_gradient_fanout (nemo_rl.models.automodel.train.LossPostProcessor property) cp_load_balanced_to_contiguous() (in module nemo_rl.distributed.model_utils) cp_sharder (nemo_rl.models.automodel.train.PreparedModelForward attribute) cp_shift_next() (in module nemo_rl.distributed.model_utils) cp_size (nemo_rl.models.automodel.config.DistributedContext attribute) (nemo_rl.models.automodel.train.PreparedModelForward attribute) cpu_offload (nemo_rl.models.automodel.config.RuntimeConfig attribute) (nemo_rl.models.policy.DTensorConfig attribute) create_app() (in module nemo_rl.models.generation.trtllm.trtllm_http_server) create_checkpoint_engine() (in module nemo_rl.utils.checkpoint_engines.base) create_context_parallel_ctx() (nemo_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl static method) create_env() (in module nemo_rl.environments.utils) create_frozen_environment_symlinks() (in module nemo_rl.utils.prefetch_venvs) create_local_venv() (in module nemo_rl.utils.venvs) create_local_venv_on_each_node() (in module nemo_rl.utils.venvs) create_sampler() (in module nemo_rl.algorithms.async_utils.staleness_sampler) create_teacher_configs_from_opd_config() (in module nemo_rl.models.policy.teacher_worker_group) create_teacher_worker_groups() (in module nemo_rl.algorithms.opd) create_weight_synchronizer() (in module nemo_rl.weight_sync.factory) create_worker() (nemo_rl.distributed.worker_groups.RayWorkerBuilder.IsolatedWorkerInitializer method) create_worker_async() (nemo_rl.distributed.worker_groups.RayWorkerBuilder method) CrossTokenizerCollator (class in nemo_rl.data.cross_tokenizer_collate) CrossTokenizerDistillationLossConfig (class in nemo_rl.algorithms.loss.loss_functions) CrossTokenizerDistillationLossDataDict (class in nemo_rl.algorithms.loss.loss_functions) CrossTokenizerDistillationLossFn (class in nemo_rl.algorithms.loss.loss_functions) cu_seqlens_k (nemo_rl.models.huggingface.common.FlashAttentionKwargs attribute) cu_seqlens_padded (nemo_rl.models.megatron.data.ProcessedInputs attribute) (nemo_rl.models.megatron.data.ProcessedMicrobatch attribute) cu_seqlens_q (nemo_rl.models.huggingface.common.FlashAttentionKwargs attribute) cuda_graph_impl (nemo_rl.models.generation.megatron.config.MCoreGenerationSpecificArgs attribute) (nemo_rl.models.policy.MegatronConfig attribute) cuda_graph_modules (nemo_rl.models.policy.MegatronConfig attribute) cuda_graph_warmup_steps (nemo_rl.models.policy.MegatronConfig attribute) CURATED_METRICS_INCLUDE_PREFIXES (in module nemo_rl.models.generation.dynamo.metrics) current_epoch (nemo_rl.algorithms.distillation.DistillationSaveState attribute) (nemo_rl.algorithms.grpo.GRPOSaveState attribute) (nemo_rl.algorithms.ppo.PPOSaveState attribute) (nemo_rl.algorithms.xtoken_off_policy_distillation.OffPolicyDistillationSaveState attribute) current_step (nemo_rl.algorithms.distillation.DistillationSaveState attribute) (nemo_rl.algorithms.grpo.GRPOSaveState attribute) (nemo_rl.algorithms.ppo.PPOSaveState attribute) (nemo_rl.algorithms.xtoken_off_policy_distillation.OffPolicyDistillationSaveState attribute) current_trace_carrier() (in module nemo_rl.telemetry.instrumentation) custom_dataloader (nemo_rl.data.DataConfig attribute) custom_jinja_template (nemo_rl.models.generation.dynamo.config.DynamoWorkerArgs attribute) custom_parallel_plan (nemo_rl.models.policy.DTensorConfig attribute) CustomSamplerConfig (class in nemo_rl.algorithms.async_utils.staleness_sampler) CyclingDataLoader (class in nemo_rl.data.dataloader) D DailyOmniDataset (class in nemo_rl.data.datasets.response_datasets.daily_omni) DailyOmniEvalDataConfig (class in nemo_rl.data) DailyOmniEvalDataset (class in nemo_rl.data.datasets.eval_datasets.daily_omni) DAPOMath17KDataset (class in nemo_rl.data.datasets.response_datasets.dapo_math) DAPOMathAIME2024Dataset (class in nemo_rl.data.datasets.response_datasets.dapo_math) data (nemo_rl.algorithms.distillation.MasterConfig attribute) (nemo_rl.algorithms.dpo.MasterConfig attribute) (nemo_rl.algorithms.grpo.MasterConfig attribute) (nemo_rl.algorithms.ppo.MasterConfig attribute) (nemo_rl.algorithms.rm.MasterConfig attribute) (nemo_rl.algorithms.sft.MasterConfig attribute) (nemo_rl.algorithms.single_controller_utils.config.MasterConfig attribute) (nemo_rl.algorithms.xtoken_off_policy_distillation.MasterConfig attribute) (nemo_rl.evals.eval.MasterConfig attribute) DATA (nemo_rl.experience.failures.FailureClass attribute) data_config (nemo_rl.data.datasets.raw_dataset.RawDataset attribute) data_dict (nemo_rl.models.automodel.data.ProcessedMicrobatch attribute) (nemo_rl.models.megatron.data.ProcessedMicrobatch attribute) data_failures_by_reason (nemo_rl.experience.rollout_manager.RolloutStats attribute) data_parallel_sharding_strategy (nemo_rl.models.policy.MegatronDDPConfig attribute) data_parallel_size (nemo_rl.models.policy.lm_policy.Policy property) data_path (nemo_rl.data.PreferenceDatasetConfig attribute) (nemo_rl.data.ResponseDatasetConfig attribute) data_plane (nemo_rl.algorithms.grpo.MasterConfig attribute) (nemo_rl.algorithms.single_controller_utils.config.MasterConfig attribute) data_plane_checkpoint_barrier (nemo_rl.algorithms.async_utils.replay_buffer.TQReplayBuffer property) DATA_PLANE_CHECKPOINT_DIR (in module nemo_rl.algorithms.async_utils.replay_buffer) data_plane_checkpoint_metadata (nemo_rl.algorithms.single_controller_utils.setup.SingleControllerActorArgs attribute) DATA_PLANE_CHECKPOINT_SCHEMA_VERSION (in module nemo_rl.data_plane.interfaces) data_plane_checkpoint_schema_version (nemo_rl.algorithms.async_utils.replay_buffer.DataPlaneCheckpointMetadata attribute) data_plane_supports_checkpointing() (in module nemo_rl.data_plane.interfaces) data_points (nemo_rl.utils.memory_tracker.MemoryTracker attribute) DATA_PROCESSING (nemo_rl.telemetry.span_groups.RLSpanGroup attribute) data_retries_by_reason (nemo_rl.experience.rollout_manager.RolloutStats attribute) DataConfig (class in nemo_rl.data) dataloader (nemo_rl.algorithms.single_controller_utils.setup.SingleControllerActorArgs attribute) DataPlaneCheckpointBarrier (class in nemo_rl.algorithms.async_utils.replay_buffer) DataPlaneCheckpointMetadata (class in nemo_rl.algorithms.async_utils.replay_buffer) DataPlaneClient (class in nemo_rl.data_plane.interfaces) DataPlaneConfig (class in nemo_rl.data_plane.interfaces) DataPlaneEvent (class in nemo_rl.data_plane.observability) DataPlaneMutationCut (class in nemo_rl.algorithms.async_utils.replay_buffer) DataPlaneStats (class in nemo_rl.data_plane.observability) dataset (nemo_rl.data.datasets.raw_dataset.RawDataset attribute) dataset_name (nemo_rl.data.AIMEEvalDataConfig attribute) (nemo_rl.data.DailyOmniEvalDataConfig attribute) (nemo_rl.data.GPQAEvalDataConfig attribute) (nemo_rl.data.LocalMathEvalDataConfig attribute) (nemo_rl.data.MathEvalDataConfig attribute) (nemo_rl.data.MMAUEvalDataConfig attribute) (nemo_rl.data.MMLUEvalDataConfig attribute) (nemo_rl.data.MMLUProEvalDataConfig attribute) (nemo_rl.data.PreferenceDatasetConfig attribute) (nemo_rl.data.ResponseDatasetConfig attribute) DATASET_REGISTRY (in module nemo_rl.data.datasets.preference_datasets) (in module nemo_rl.data.datasets.response_datasets) DatumSpec (class in nemo_rl.data.interfaces) DEAD (nemo_rl.models.generation.fleet_health.ShardState attribute) debug_payload_metrics (nemo_rl.algorithms.grpo.GRPOConfig attribute) decode_routed_experts() (in module nemo_rl.utils.routed_experts_codec) decode_sparse_payload() (in module nemo_rl.utils.weight_transfer_stream) decompose_message_log() (in module nemo_rl.data.llm_message_utils) dedicated_inference_megatron_cfg() (in module nemo_rl.models.generation.megatron.config) deduplicate_multimodal_data (nemo_rl.algorithms.grpo.GRPOConfig attribute) deduplicate_shared_teacher_checkpoints (nemo_rl.algorithms.opd.OnPolicyDistillationConfig attribute) deduplication_enabled (nemo_rl.data.multimodal_utils.PackedTensor property) DeepScalerDataset (class in nemo_rl.data.datasets.response_datasets.deepscaler) deepseekv3() (in module nemo_rl.utils.flops_formulas) default (nemo_rl.data.DataConfig attribute) DEFAULT_CALIB_BATCH_SIZE (in module nemo_rl.modelopt.models.policy.workers.utils) DEFAULT_CALIB_SAMPLE_LENGTH (in module nemo_rl.modelopt.models.policy.workers.utils) default_chat_template_kwargs (nemo_rl.models.generation.trtllm.config.TrtllmSpecificArgs attribute) DEFAULT_DYNAMO_CONTROL_PORT_RANGE_HIGH (in module nemo_rl.distributed.virtual_cluster) DEFAULT_DYNAMO_CONTROL_PORT_RANGE_LOW (in module nemo_rl.distributed.virtual_cluster) DEFAULT_DYNAMO_HTTP_PORT_RANGE_HIGH (in module nemo_rl.distributed.virtual_cluster) DEFAULT_DYNAMO_HTTP_PORT_RANGE_LOW (in module nemo_rl.distributed.virtual_cluster) DEFAULT_DYNAMO_SYSTEM_PORT_RANGE_HIGH (in module nemo_rl.distributed.virtual_cluster) DEFAULT_DYNAMO_SYSTEM_PORT_RANGE_LOW (in module nemo_rl.distributed.virtual_cluster) DEFAULT_GENERATION_PORT_RANGE_HIGH (in module nemo_rl.distributed.virtual_cluster) DEFAULT_GENERATION_PORT_RANGE_LOW (in module nemo_rl.distributed.virtual_cluster) DEFAULT_GYM_PORT_RANGE_HIGH (in module nemo_rl.distributed.virtual_cluster) DEFAULT_GYM_PORT_RANGE_LOW (in module nemo_rl.distributed.virtual_cluster) DEFAULT_INVALID_TOOL_CALL_PATTERNS (in module nemo_rl.environments.nemo_gym) DEFAULT_MASTER_PORT_RANGE_HIGH (in module nemo_rl.distributed.virtual_cluster) DEFAULT_MASTER_PORT_RANGE_LOW (in module nemo_rl.distributed.virtual_cluster) DEFAULT_MEDIA_EXTENSIONS (in module nemo_rl.data.multimodal_utils) DEFAULT_METRICS_EXCLUDE_PREFIXES (in module nemo_rl.models.generation.dynamo.metrics) DEFAULT_NVFP4_IGNORE (in module nemo_rl.modelopt.utils) default_processor (nemo_rl.environments.utils.EnvRegistryEntry attribute) DEFAULT_PY_EXECUTABLE (nemo_rl.environments.reward_model_environment.RewardModelEnvironment attribute) DEFAULT_SGLANG_PROMETHEUS_PORT_RANGE_HIGH (in module nemo_rl.distributed.virtual_cluster) DEFAULT_SGLANG_PROMETHEUS_PORT_RANGE_LOW (in module nemo_rl.distributed.virtual_cluster) DEFAULT_SGLANG_ROUTER_PORT_RANGE_HIGH (in module nemo_rl.distributed.virtual_cluster) DEFAULT_SGLANG_ROUTER_PORT_RANGE_LOW (in module nemo_rl.distributed.virtual_cluster) default_teacher_alias (nemo_rl.algorithms.opd.OnPolicyDistillationConfig attribute) default_teacher_cfg (nemo_rl.algorithms.opd.NonColocatedTeachersConfig attribute) DEFAULT_TEMPLATE (in module nemo_rl.data.datasets.eval_datasets.mmau) (in module nemo_rl.data.datasets.response_datasets.audiomcq) (in module nemo_rl.data.datasets.response_datasets.avqa) DEFAULT_THINKING_TAGS (in module nemo_rl.environments.nemo_gym) DEFAULT_VENV_DIR (in module nemo_rl.utils.venvs) DEFAULT_VLLM_PORT_RANGE_LOW (in module nemo_rl.distributed.virtual_cluster) DEFAULT_VLLM_PORTS_PER_ENGINE (in module nemo_rl.distributed.virtual_cluster) defer_fp32_logits (nemo_rl.models.policy.MegatronConfig attribute) defer_fsdp_grad_sync (nemo_rl.models.policy.DTensorConfig attribute) define_metric() (nemo_rl.utils.logger.WandbLogger method) delete() (nemo_rl.utils.weight_transfer_stream._S3ObjectStore method) delta_compression (nemo_rl.models.generation.vllm.config.VllmSparseRefitConfig attribute) DeltaCompressionTracker (class in nemo_rl.utils.weight_transfer_sparse_codec) dense_bytes (nemo_rl.utils.weight_transfer_stream._SparsePayloadBucket attribute) dequantize_base_checkpoint (nemo_rl.models.value.config.ValueConfig attribute) destroy_parallel_state() (in module nemo_rl.models.megatron.setup) detect() (nemo_rl.models.megatron.draft.utils._EagleModelLayout class method) detect_checkpoint_format() (in module nemo_rl.models.automodel.checkpoint) determine_available_memory() (nemo_rl.modelopt.models.generation.vllm_quant_patch.FakeQuantWorker method) device (nemo_rl.models.generation.vllm.config.VllmNixlRefitConfig attribute) device_id_to_physical_device_id() (in module nemo_rl.utils.nvml) device_mesh (nemo_rl.models.automodel.config.DistributedContext attribute) diagnostics (nemo_rl.algorithms.single_controller_utils.config.AsyncRLConfig attribute) DictT (in module nemo_rl.distributed.batched_data_dict) dim (nemo_rl.models.policy.LoRAConfig attribute) (nemo_rl.models.policy.MegatronPeftConfig attribute) dir_path (in module nemo_rl.distributed.virtual_cluster) (in module nemo_rl.utils.venvs) disable_forward_pre_hook() (nemo_rl.modelopt.models.policy.workers.megatron_quant_policy_worker.MegatronQuantPolicyWorker method) (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) (nemo_rl.models.value.workers.megatron_value_worker.MegatronValueWorkerImpl method) disable_modelopt_layer_spec (nemo_rl.models.policy.PolicyConfig attribute) disable_ppo_ratio (nemo_rl.algorithms.loss.loss_functions.ClippedPGLossConfig attribute) disable_quantization() (nemo_rl.modelopt.models.policy.workers.megatron_quant_policy_worker.MegatronQuantPolicyWorker method) discard_canonical_groups() (nemo_rl.experience.rollout_recovery.RolloutRecoveryLedger method) discard_group() (nemo_rl.experience.rollout_recovery.RolloutRecoveryLedger method) discard_prompt_group() (nemo_rl.experience.rollout_manager.RolloutManager method) discard_samples() (nemo_rl.models.policy.tq_policy.TQPolicy method) disconnect_rollout_engines_from_distributed() (in module nemo_rl.models.policy.utils) discover_native_skips() (nemo_rl.models.generation.vllm.vllm_sparse_delta.VllmSparseDeltaApplier method) dispatch_index (nemo_rl.algorithms.async_utils.staleness_sampler.BaseSampler property) (nemo_rl.algorithms.async_utils.staleness_sampler.PromptGroupSampler property) dispatcher (nemo_rl.models.policy.AutomodelBackendConfig attribute) distillation (nemo_rl.algorithms.distillation.MasterConfig attribute) DISTILLATION (nemo_rl.algorithms.loss.interfaces.LossInputType attribute) distillation (nemo_rl.algorithms.xtoken_off_policy_distillation.MasterConfig attribute) DISTILLATION_CROSS_TOKENIZER (nemo_rl.algorithms.loss.interfaces.LossInputType attribute) distillation_train() (in module nemo_rl.algorithms.distillation) DistillationConfig (class in nemo_rl.algorithms.distillation) DistillationLossConfig (class in nemo_rl.algorithms.loss.loss_functions) DistillationLossDataDict (class in nemo_rl.algorithms.loss.loss_functions) DistillationLossFn (class in nemo_rl.algorithms.loss.loss_functions) DistillationSaveState (class in nemo_rl.algorithms.distillation) distributed_data_parallel_config (nemo_rl.models.policy.MegatronConfig attribute) distributed_vocab_topk() (in module nemo_rl.distributed.model_utils) DistributedContext (class in nemo_rl.models.automodel.config) DistributedCrossEntropy (class in nemo_rl.distributed.model_utils) DistributedLogprob (class in nemo_rl.distributed.model_utils) DistributedLogprobWithSampling (class in nemo_rl.distributed.model_utils) download_and_unzip() (in module nemo_rl.data.datasets.response_datasets.refcoco) download_dir (nemo_rl.data.ResponseDatasetConfig attribute) download_s3_refit_payload() (in module nemo_rl.utils.weight_transfer_stream) DP_CALIB_INPUT_FIELDS (in module nemo_rl.data_plane.schema) dp_client (nemo_rl.algorithms.single_controller_utils.setup.SingleControllerActorArgs attribute) dp_replicate_size (nemo_rl.models.policy.DTensorConfig attribute) dp_shard_idx (nemo_rl.models.generation.fleet_health.ShardHealth attribute) dp_size (nemo_rl.distributed.worker_groups.RayWorkerGroup property) (nemo_rl.models.automodel.config.DistributedContext attribute) DP_TRAIN_FIELDS (in module nemo_rl.data_plane.schema) DP_VALUE_TRAIN_FIELDS (in module nemo_rl.data_plane.schema) dpo (nemo_rl.algorithms.dpo.MasterConfig attribute) dpo_train() (in module nemo_rl.algorithms.dpo) DPOConfig (class in nemo_rl.algorithms.dpo) DPOLossConfig (class in nemo_rl.algorithms.loss.loss_functions) DPOLossDataDict (class in nemo_rl.algorithms.loss.loss_functions) DPOLossFn (class in nemo_rl.algorithms.loss.loss_functions) DPOSaveState (class in nemo_rl.algorithms.dpo) DPOValMetrics (class in nemo_rl.algorithms.dpo) DRAFT (nemo_rl.algorithms.loss.interfaces.LossInputType attribute) draft (nemo_rl.models.policy.PolicyConfig attribute) DRAFT_GRAD_NORM_GROUP (in module nemo_rl.models.megatron.draft.utils) draft_model (nemo_rl.models.megatron.config.ModelAndOptimizerState attribute) draft_model_detached() (in module nemo_rl.models.megatron.draft.utils) DraftConfig (class in nemo_rl.models.policy) DraftConfigDisabled (class in nemo_rl.models.policy) DraftCrossEntropyLossConfig (class in nemo_rl.algorithms.loss.loss_functions) DraftCrossEntropyLossDataDict (class in nemo_rl.algorithms.loss.loss_functions) DraftCrossEntropyLossFn (class in nemo_rl.algorithms.loss.loss_functions) DraftLossWrapper (class in nemo_rl.algorithms.loss.wrapper) drain_backend_outcomes() (nemo_rl.models.generation.generation_router.GenerationRouterImpl method) drain_metrics() (nemo_rl.algorithms.opd.TQTeacherLogprobCoordinator method) drain_multimodal_payload_metrics() (in module nemo_rl.utils.multimodal_payload_metrics) drain_payload_metrics() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector method) drop() (nemo_rl.data_plane.interfaces.KVBatchMeta method) drop_incomplete_targets_on_restore (nemo_rl.algorithms.ppo.AsyncPPOConfig attribute) dropout (nemo_rl.models.policy.LoRAConfig attribute) (nemo_rl.models.policy.MegatronPeftConfig attribute) dropout_position (nemo_rl.models.policy.LoRAConfig attribute) (nemo_rl.models.policy.MegatronPeftConfig attribute) dsa_indexer_compute_layers (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) dsa_indexer_head_dim (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) dsa_indexer_n_heads (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) dsa_indexer_topk (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) dtensor_cfg (nemo_rl.environments.reward_model_environment.RewardModelEnvironmentConfig attribute) (nemo_rl.models.policy.PolicyConfig attribute) (nemo_rl.models.value.config.ValueConfig attribute) dtensor_from_parallel_logits_to_logprobs() (in module nemo_rl.distributed.model_utils) dtensor_params_generator() (in module nemo_rl.models.policy.workers.dtensor_policy_worker) (in module nemo_rl.models.policy.workers.dtensor_policy_worker_v2) DTensorCheckpointEngineSendMixin (class in nemo_rl.models.policy.workers.checkpoint_engine) DTensorConfig (class in nemo_rl.models.policy) DTensorConfigDisabled (class in nemo_rl.models.policy) DTensorPolicyWorker (class in nemo_rl.models.policy.workers.dtensor_policy_worker) DTensorPolicyWorkerImpl (class in nemo_rl.models.policy.workers.dtensor_policy_worker) DTensorPolicyWorkerV2 (class in nemo_rl.models.policy.workers.dtensor_policy_worker_v2) DTensorPolicyWorkerV2Impl (class in nemo_rl.models.policy.workers.dtensor_policy_worker_v2) DTensorQuantPolicyWorker (class in nemo_rl.modelopt.models.policy.workers.dtensor_quant_policy_worker) DTensorQuantPolicyWorkerV2 (class in nemo_rl.modelopt.models.policy.workers.dtensor_quant_policy_worker_v2) DTensorRef (class in nemo_rl.weight_sync.xferdtensor) DTensorValueWorkerV2 (class in nemo_rl.models.value.workers.dtensor_value_worker_v2) DTensorValueWorkerV2Impl (class in nemo_rl.models.value.workers.dtensor_value_worker_v2) dtype (nemo_rl.models.automodel.config.RuntimeConfig attribute) (nemo_rl.models.megatron.config.RuntimeConfig attribute) (nemo_rl.utils.checkpoint_engines.base.TensorMeta attribute) dtype_from_name() (in module nemo_rl.utils.weight_transfer_sparse_codec) dynamic_batching (nemo_rl.environments.reward_model_environment.RewardModelEnvironmentConfig attribute) (nemo_rl.models.policy.PolicyConfig attribute) (nemo_rl.models.value.config.ValueConfig attribute) dynamic_loss_scaling (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute) dynamic_sampling() (in module nemo_rl.algorithms.grpo) (in module nemo_rl.algorithms.ppo) dynamic_sampling_max_gen_batches (nemo_rl.algorithms.grpo.GRPOConfig attribute) (nemo_rl.algorithms.ppo.PPOConfig attribute) DynamicBatchingArgs (class in nemo_rl.distributed.batched_data_dict) DynamicBatchingConfig (class in nemo_rl.models.policy) DynamicBatchingConfigDisabled (class in nemo_rl.models.policy) DYNAMO_BACKEND (in module nemo_rl.models.generation.constants) dynamo_cfg (nemo_rl.models.generation.dynamo.config.DynamoConfig attribute) DYNAMO_VLLM_FLAGS (in module nemo_rl.models.generation.dynamo.config) DynamoCfg (class in nemo_rl.models.generation.dynamo.config) DynamoConfig (class in nemo_rl.models.generation.dynamo.config) DynamoFrontendArgs (class in nemo_rl.models.generation.dynamo.config) DynamoGeneration (class in nemo_rl.models.generation.dynamo.dynamo_generation) DynamoGpuReservation (class in nemo_rl.models.generation.dynamo.dynamo_worker) DynamoMetricsSampler (class in nemo_rl.models.generation.dynamo.metrics) DynamoRefitChannel (class in nemo_rl.models.generation.dynamo.refit) DynamoTokenWrapperServer (class in nemo_rl.models.generation.dynamo.token_wrapper) DynamoVllmConfig (class in nemo_rl.models.generation.dynamo.config) DynamoVllmWorker (class in nemo_rl.models.generation.dynamo.dynamo_worker) DynamoWorkerArgs (class in nemo_rl.models.generation.dynamo.config) DynamoWorkerEndpoint (class in nemo_rl.models.generation.dynamo.refit) E EagleModel (class in nemo_rl.models.megatron.draft.eagle) effective_megatron_cfg() (nemo_rl.models.generation.megatron.megatron_generation.MegatronGeneration static method) EFFICIENCY (nemo_rl.telemetry.span_groups.RLSpanGroup attribute) EFFICIENCY_CATEGORIES (in module nemo_rl.algorithms.utils) EFFICIENCY_CATEGORY_BUCKET (in module nemo_rl.telemetry.instrumentation) efficiency_measurements() (in module nemo_rl.telemetry.metrics) efficiency_span() (in module nemo_rl.telemetry.instrumentation) efficiency_window() (in module nemo_rl.telemetry.metrics) EffortLevelsConfig (class in nemo_rl.experience.rollouts) ELEM_COUNTS_PER_GB (in module nemo_rl.data_plane.schema) empty_like() (nemo_rl.data.multimodal_utils.PackedTensor class method) empty_rows_like() (nemo_rl.data.multimodal_utils.PackedTensor class method) empty_unused_memory_level (nemo_rl.models.policy.MegatronConfig attribute) enable_chunked_prefill (nemo_rl.models.generation.megatron.config.MCoreGenerationSpecificArgs attribute) enable_deduplication() (nemo_rl.data.multimodal_utils.PackedTensor method) enable_deepep (nemo_rl.models.policy.AutomodelBackendConfig attribute) enable_forward_pre_hook() (nemo_rl.modelopt.models.policy.workers.megatron_quant_policy_worker.MegatronQuantPolicyWorker method) (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) (nemo_rl.models.value.workers.megatron_value_worker.MegatronValueWorkerImpl method) enable_fsdp_optimizations (nemo_rl.models.policy.AutomodelBackendConfig attribute) enable_hf_state_dict_adapter (nemo_rl.models.policy.AutomodelBackendConfig attribute) enable_prefix_caching (nemo_rl.models.generation.megatron.config.MCoreGenerationSpecificArgs attribute) enable_return_routed_experts (nemo_rl.models.generation.vllm.config.VllmSpecificArgs attribute) enable_seq_packing (nemo_rl.models.automodel.config.RuntimeConfig attribute) enable_structural_tag (nemo_rl.models.generation.dynamo.config.DynamoWorkerArgs attribute) enable_vllm_metrics_logger (nemo_rl.models.generation.dynamo.config.DynamoVllmConfig attribute) enabled (nemo_rl.algorithms.grpo.AsyncGRPOConfig attribute) (nemo_rl.algorithms.grpo.RewardScalingConfig attribute) (nemo_rl.algorithms.opd.NonColocatedTeachersConfig attribute) (nemo_rl.algorithms.opd.OnPolicyDistillationConfig attribute) (nemo_rl.algorithms.ppo.AsyncPPOConfig attribute) (nemo_rl.algorithms.reward_functions.RewardShapingConfig attribute) (nemo_rl.algorithms.single_controller_utils.config.FleetHealthConfig attribute) (nemo_rl.algorithms.single_controller_utils.config.GenerationRouterConfig attribute) (nemo_rl.data_plane.interfaces.DataPlaneConfig attribute) (nemo_rl.data_plane.interfaces.ObservabilityConfig attribute) (nemo_rl.environments.reward_model_environment.RewardModelEnvironmentConfig attribute) (nemo_rl.models.generation.interfaces.ColocationConfig attribute) (nemo_rl.models.policy.DraftConfig attribute) (nemo_rl.models.policy.DraftConfigDisabled attribute) (nemo_rl.models.policy.DTensorConfig attribute) (nemo_rl.models.policy.DTensorConfigDisabled attribute) (nemo_rl.models.policy.DynamicBatchingConfig attribute) (nemo_rl.models.policy.DynamicBatchingConfigDisabled attribute) (nemo_rl.models.policy.Fp8Config attribute) (nemo_rl.models.policy.LoRAConfig attribute) (nemo_rl.models.policy.LoRAConfigDisabled attribute) (nemo_rl.models.policy.MegatronConfig attribute) (nemo_rl.models.policy.MegatronConfigDisabled attribute) (nemo_rl.models.policy.MegatronPeftConfig attribute) (nemo_rl.models.policy.MegatronPeftConfigDisabled attribute) (nemo_rl.models.policy.RewardModelConfig attribute) (nemo_rl.models.policy.RouterReplayConfig attribute) (nemo_rl.models.policy.RouterReplayConfigDisabled attribute) (nemo_rl.models.policy.SequencePackingConfig attribute) (nemo_rl.models.policy.SequencePackingConfigDisabled attribute) (nemo_rl.telemetry.config.TelemetryConfig attribute) (nemo_rl.utils.checkpoint.CheckpointingConfig attribute) enc_seq_len (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) encode_images_in_examples() (in module nemo_rl.data.multimodal_utils) encode_routed_experts() (in module nemo_rl.utils.routed_experts_codec) encode_s (nemo_rl.utils.weight_transfer_stream._SparsePayloadBucket attribute) encode_single() (nemo_rl.data.datasets.processed_dataset.AllTaskProcessedDataset method) encode_sparse_infos() (in module nemo_rl.utils.weight_transfer_sparse_codec) encode_workers (nemo_rl.models.generation.vllm.config.VllmRefitTuningConfig attribute) encoding (nemo_rl.models.generation.vllm.config.VllmDeltaCompressionConfig attribute) end_weight (nemo_rl.algorithms.async_utils.replay_buffer.TQReplayGroupMetadata attribute) end_weight_decay (nemo_rl.models.policy.MegatronSchedulerConfig attribute) endpoint_types (nemo_rl.models.generation.dynamo.config.DynamoWorkerArgs attribute) enforce_eager (nemo_rl.models.generation.dynamo.config.DynamoVllmConfig attribute) (nemo_rl.models.generation.vllm.config.VllmSpecificArgs attribute) engine (nemo_rl.models.generation.dynamo.config.DynamoCfg attribute) engine_kwargs (nemo_rl.models.generation.interfaces.CheckpointEngineConfig attribute) engine_world_size (nemo_rl.models.generation.dynamo.config.DynamoConfig property) EnglishMultichoiceVerifyWorker (class in nemo_rl.environments.math_environment) enrich() (nemo_rl.algorithms.opd.TQTeacherLogprobCoordinator method) ensure_teacher_ipc_buffer() (in module nemo_rl.models.policy.utils) ensure_vllm_source_compat() (in module nemo_rl.models.generation.vllm.patches) entity (nemo_rl.utils.logger.WandbConfig attribute) enums (nemo_rl.data_plane.adapters.noop._Partition attribute) env (nemo_rl.algorithms.distillation.MasterConfig attribute) (nemo_rl.algorithms.grpo.MasterConfig attribute) (nemo_rl.algorithms.ppo.MasterConfig attribute) (nemo_rl.algorithms.single_controller_utils.config.MasterConfig attribute) (nemo_rl.environments.nemo_gym.NemoGymCompatibleConfig property) (nemo_rl.evals.eval.MasterConfig attribute) env_extras (nemo_rl.experience.interfaces.Completion attribute) env_handles (nemo_rl.algorithms.single_controller_utils.setup.SingleControllerActorArgs attribute) env_name (nemo_rl.data.DailyOmniEvalDataConfig attribute) (nemo_rl.data.MMAUEvalDataConfig attribute) (nemo_rl.data.ResponseDatasetConfig attribute) ENV_REGISTRY (in module nemo_rl.environments.utils) env_s (nemo_rl.experience.rollout_manager.RolloutTimeouts attribute) env_timeout_s (nemo_rl.algorithms.single_controller_utils.config.NativeRolloutFTConfig attribute) env_vars (nemo_rl.models.generation.dynamo.config.DynamoVllmConfig attribute) (nemo_rl.models.generation.vllm.config.VllmSpecificArgs attribute) (nemo_rl.models.policy.DTensorConfig attribute) (nemo_rl.models.policy.MegatronConfig attribute) EnvironmentInterface (class in nemo_rl.environments.interfaces) EnvironmentReturn (class in nemo_rl.environments.interfaces) EnvRegistryEntry (class in nemo_rl.environments.utils) epoch (nemo_rl.algorithms.dpo.DPOSaveState attribute) (nemo_rl.algorithms.rm.RMSaveState attribute) (nemo_rl.algorithms.sft.SFTSaveState attribute) eval (nemo_rl.evals.eval.MasterConfig attribute) eval_collate_fn() (in module nemo_rl.data.collate_fn) eval_cons_k() (in module nemo_rl.evals.eval) eval_pass_k() (in module nemo_rl.evals.eval) EvalConfig (class in nemo_rl.evals.eval) EvalDataConfigType (in module nemo_rl.data) EventStatus (in module nemo_rl.data_plane.observability) evict() (nemo_rl.algorithms.async_utils.staleness_sampler.BaseSampler method) (nemo_rl.algorithms.async_utils.staleness_sampler.InOrderSampler method) (nemo_rl.algorithms.async_utils.staleness_sampler.PromptGroupSampler method) (nemo_rl.algorithms.async_utils.staleness_sampler.ReadyFirstSampler method) exact_answer_alphanumeric_reward() (in module nemo_rl.environments.rewards) exact_token_match_only (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute) exclude_modules (nemo_rl.models.policy.LoRAConfig attribute) (nemo_rl.models.policy.MegatronPeftConfig attribute) exclude_tools_when_tool_choice_none (nemo_rl.models.generation.dynamo.config.DynamoWorkerArgs attribute) execute() (nemo_rl.environments.code_environment.CodeExecutionWorker method) expected_generations (nemo_rl.experience.rollout_recovery.PromptGroupRecoveryRecord attribute) (nemo_rl.experience.rollout_recovery.PromptGroupRecoveryState attribute) experiment_name (nemo_rl.utils.logger.MLflowConfig attribute) expert_id (nemo_rl.models.generation.vllm.refit_layout.HfExpertWeight attribute) expert_model_parallel_size (nemo_rl.algorithms.opd.TeacherResourceConfig attribute) (nemo_rl.models.policy.MegatronConfig attribute) (nemo_rl.models.policy.teacher_worker_group.TeacherConfig attribute) expert_parallel_size (nemo_rl.models.generation.dynamo.config.DynamoVllmConfig attribute) (nemo_rl.models.generation.vllm.config.VllmSpecificArgs attribute) (nemo_rl.models.policy.DTensorConfig attribute) expert_params (nemo_rl.models.generation.vllm.refit_layout.VllmWeightLayout attribute) expert_tensor_parallel_size (nemo_rl.models.policy.MegatronConfig attribute) experts (nemo_rl.models.policy.AutomodelBackendConfig attribute) export_chunk_bytes (nemo_rl.models.generation.vllm.config.VllmDeltaCompressionConfig attribute) export_eagle_weights_to_hf() (in module nemo_rl.models.megatron.draft.utils) export_model_from_megatron() (in module nemo_rl.models.megatron.community_import) export_rank (nemo_rl.telemetry.config.TelemetryConfig attribute) export_sample_rate (nemo_rl.telemetry.config.TelemetryConfig attribute) export_strategy (nemo_rl.telemetry.config.TelemetryConfig attribute) export_teacher_logits_and_pack() (in module nemo_rl.algorithms.xtoken_off_policy_distillation) exporter (nemo_rl.telemetry.config.TelemetryConfig attribute) expose_http_server (nemo_rl.models.generation.dynamo.config.DynamoVllmConfig attribute) (nemo_rl.models.generation.megatron.config.MCoreGenerationSpecificArgs attribute) (nemo_rl.models.generation.trtllm.config.TrtllmSpecificArgs attribute) (nemo_rl.models.generation.vllm.config.VllmSpecificArgs attribute) extra (nemo_rl.weight_sync.nccl_reshard_utils.RefitCtx attribute) extra_cli_args (nemo_rl.models.generation.dynamo.config.DynamoFrontendArgs attribute) (nemo_rl.models.generation.dynamo.config.DynamoWorkerArgs attribute) extra_env_info (nemo_rl.data.interfaces.DatumSpec attribute) (nemo_rl.experience.interfaces.PromptGroupRecord attribute) extra_info (nemo_rl.data_plane.interfaces.KVBatchMeta attribute) extract_initial_prompt_messages() (in module nemo_rl.algorithms.grpo) extract_input_image_sources_from_responses_messages() (in module nemo_rl.data.multimodal_utils) extract_input_images_from_responses_messages() (in module nemo_rl.data.multimodal_utils) extract_logits() (in module nemo_rl.models.automodel.train) extract_multimodal_model_inputs() (in module nemo_rl.data.multimodal_utils) extract_necessary_env_names() (in module nemo_rl.data.datasets.utils) extract_reward_components() (in module nemo_rl.environments.nemo_gym) extracted_answer (nemo_rl.environments.math_environment.MathEnvironmentMetadata attribute) extras (nemo_rl.algorithms.metric_utils.SetupTimingMetrics attribute) F fabric_is_roce_only() (in module nemo_rl.data_plane.adapters.transfer_queue_env) FailureClass (class in nemo_rl.experience.failures) fake_balanced_gate (nemo_rl.models.policy.AutomodelBackendConfig attribute) FakeQuantWorker (class in nemo_rl.modelopt.models.generation.vllm_quant_patch) fc1_weight (nemo_rl.models.megatron.draft.utils._PendingLayerWeights attribute) fc1_weight_key (nemo_rl.models.megatron.draft.utils._EagleLayerLayout property) fc2_weight_key (nemo_rl.models.megatron.draft.utils._EagleLayerLayout property) FetchPolicy (in module nemo_rl.data_plane.worker_mixin) FFN_GROUPED_EXPERT_SUFFIXES (in module nemo_rl.weight_sync.nccl_reshard_utils) ffn_hs (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) FFN_PROJ_WEIGHT_SUFFIXES (in module nemo_rl.weight_sync.nccl_reshard_utils) fields (nemo_rl.data_plane.adapters.noop._Partition attribute) (nemo_rl.data_plane.interfaces.KVBatchMeta attribute) fields_for_put() (in module nemo_rl.algorithms.single_controller_utils.utils) fields_with_optional_routed_experts() (in module nemo_rl.data_plane.schema) file_format (nemo_rl.data.LocalMathEvalDataConfig attribute) filter_multimodal_kwargs_for_model() (in module nemo_rl.models.automodel.data) final_batch (nemo_rl.experience.rollouts.NemoGymRolloutResult attribute) (nemo_rl.experience.rollouts.RolloutGroupResult attribute) final_norm_key (nemo_rl.models.megatron.draft.utils._EagleModelLayout attribute) final_padded_vocab_size (nemo_rl.models.megatron.config.RuntimeConfig attribute) finalize() (nemo_rl.utils.checkpoint_engines.base.CheckpointEngine method) (nemo_rl.utils.checkpoint_engines.nixl.NIXLCheckpointEngine method) finalize_async_save() (nemo_rl.models.automodel.checkpoint.AutomodelCheckpointManager method) (nemo_rl.models.policy.lm_policy.Policy method) (nemo_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker method) (nemo_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl method) (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) finalize_checkpoint() (nemo_rl.utils.checkpoint.CheckpointManager method) finalize_checkpoint_engine() (nemo_rl.models.generation.vllm.checkpoint_engine.VllmCheckpointEngineMixin method) finalize_megatron_setup() (in module nemo_rl.models.megatron.setup) finalize_pending() (nemo_rl.utils.checkpoint.CheckpointManager method) find_draft_owner_chunk() (in module nemo_rl.models.megatron.draft.utils) fine_grained_activation_offloading (nemo_rl.models.policy.MegatronConfig attribute) finish() (nemo_rl.experience.rollouts._NemoGymStreamAccumulator method) (nemo_rl.models.generation.vllm.vllm_sparse_delta._SparseWeightLoadMode method) (nemo_rl.models.policy.workers.megatron_remote_sparse_refit.MegatronRemoteSparseRefit method) (nemo_rl.utils.logger.Logger method) (nemo_rl.utils.logger.WandbLogger method) (nemo_rl.utils.weight_transfer_sparse_codec._TensorPayloadBuilder method) finish_generation() (nemo_rl.models.generation.dynamo.dynamo_generation.DynamoGeneration method) (nemo_rl.models.generation.interfaces.GenerationInterface method) (nemo_rl.models.generation.megatron.megatron_generation.MegatronGeneration method) (nemo_rl.models.generation.megatron.megatron_worker.MegatronGenerationMixin method) (nemo_rl.models.generation.trtllm.trtllm_generation.TrtllmGeneration method) (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) (nemo_rl.models.policy.lm_policy.Policy method) finish_inference() (nemo_rl.models.policy.lm_policy.Policy method) (nemo_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl method) (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) (nemo_rl.models.value.lm_value.Value method) (nemo_rl.models.value.workers.dtensor_value_worker_v2.DTensorValueWorkerV2Impl method) (nemo_rl.models.value.workers.megatron_value_worker.MegatronValueWorkerImpl method) finish_remote_sparse_delta_sync() (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) finish_sparse_delta_refit() (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) (nemo_rl.models.generation.vllm.vllm_sparse_delta.VllmSparseDeltaApplier method) finish_step() (nemo_rl.models.policy.tq_policy.TQPolicy method) finish_train_step() (nemo_rl.models.policy.tq_policy.TQPolicy method) (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) finish_train_step_presharded() (nemo_rl.data_plane.worker_mixin.TQWorkerMixin method) finish_training() (nemo_rl.models.policy.interfaces.PolicyInterface method) (nemo_rl.models.policy.lm_policy.Policy method) (nemo_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker method) (nemo_rl.models.value.interfaces.ValueInterface method) (nemo_rl.models.value.lm_value.Value method) (nemo_rl.models.value.workers.dtensor_value_worker_v2.DTensorValueWorkerV2Impl method) (nemo_rl.models.value.workers.megatron_value_worker.MegatronValueWorkerImpl method) finished_at (nemo_rl.models.generation.vllm.vllm_sparse_refit._StagedSparsePayload attribute) fired (nemo_rl.distributed.refit_watchdog.RefitAbortWatchdog property) FIRST_FIT_DECREASING (nemo_rl.data.packing.algorithms.PackingAlgorithm attribute) FIRST_FIT_SHUFFLE (nemo_rl.data.packing.algorithms.PackingAlgorithm attribute) FirstFitDecreasingPacker (class in nemo_rl.data.packing.algorithms) FirstFitPacker (class in nemo_rl.data.packing.algorithms) FirstFitShufflePacker (class in nemo_rl.data.packing.algorithms) fix_gemma3_vision_weight_name() (in module nemo_rl.models.generation.vllm.vllm_backend) FixedDynamoWorkerPool (class in nemo_rl.models.generation.dynamo.worker_pool) flash_attn_kwargs (nemo_rl.models.automodel.data.ProcessedInputs attribute) FlashAttentionKwargs (class in nemo_rl.models.huggingface.common) FlatMessagesType (in module nemo_rl.data.interfaces) flatten_dict() (in module nemo_rl.utils.logger) flattened_concat() (nemo_rl.data.multimodal_utils.PackedTensor class method) fleet_monitor (nemo_rl.algorithms.single_controller_utils.setup.SingleControllerActorArgs attribute) FleetHealthConfig (class in nemo_rl.algorithms.single_controller_utils.config) FleetHealthPolicy (class in nemo_rl.models.generation.fleet_health) FLOPSConfig (class in nemo_rl.utils.flops_formulas) FLOPTracker (class in nemo_rl.utils.flops_tracker) flush() (nemo_rl.utils.logger.RayGpuMonitorLogger method) (nemo_rl.utils.weight_transfer_zmq.ZmqSparseRefitServer method) flush_cache() (nemo_rl.models.generation.dynamo.refit.DynamoRefitChannel method) flush_interval (nemo_rl.utils.logger.GPUMonitoringConfig attribute) flush_telemetry() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector method) flush_zmq_sparse_refit_relay() (nemo_rl.models.generation.vllm.vllm_sparse_refit.VllmSparseRefitReceiver method) flux() (in module nemo_rl.utils.flops_formulas) force_clear_fp8_caches (nemo_rl.models.policy.Fp8Config attribute) force_hf (nemo_rl.models.policy.AutomodelKwargs attribute) force_on_policy_ratio (nemo_rl.algorithms.loss.loss_functions.ClippedPGLossConfig attribute) force_reconvert_from_hf (nemo_rl.models.policy.MegatronConfig attribute) format (nemo_rl.utils.checkpoint.PretrainedCheckpointConfig attribute) format_answer_fromtags() (in module nemo_rl.data.datasets.response_datasets.clevr) format_clevr_cogent_dataset() (in module nemo_rl.data.datasets.response_datasets.clevr) format_data() (nemo_rl.data.datasets.eval_datasets.mmau.MMAUDataset method) (nemo_rl.data.datasets.preference_datasets.binary_preference_dataset.BinaryPreferenceDataset method) (nemo_rl.data.datasets.preference_datasets.helpsteer3.HelpSteer3Dataset method) (nemo_rl.data.datasets.preference_datasets.tulu3.Tulu3PreferenceDataset method) (nemo_rl.data.datasets.response_datasets.aime.AIMEDataset method) (nemo_rl.data.datasets.response_datasets.arrow_text_dataset.ArrowTextDataset method) (nemo_rl.data.datasets.response_datasets.audiomcq.AudioMCQDataset method) (nemo_rl.data.datasets.response_datasets.avqa.AVQADataset method) (nemo_rl.data.datasets.response_datasets.daily_omni.DailyOmniDataset method) (nemo_rl.data.datasets.response_datasets.dapo_math.DAPOMath17KDataset method) (nemo_rl.data.datasets.response_datasets.deepscaler.DeepScalerDataset method) (nemo_rl.data.datasets.response_datasets.gsm8k.GSM8KDataset method) (nemo_rl.data.datasets.response_datasets.helpsteer3.HelpSteer3Dataset method) (nemo_rl.data.datasets.response_datasets.intent.IntentDataset method) (nemo_rl.data.datasets.response_datasets.nemotron_cascade2_sft.NemotronCascade2SFTMathDataset method) (nemo_rl.data.datasets.response_datasets.numinamath.NuminaMath15Dataset method) (nemo_rl.data.datasets.response_datasets.oai_format_dataset.OpenAIFormatDataset method) (nemo_rl.data.datasets.response_datasets.openmathinstruct2.OpenMathInstruct2Dataset method) (nemo_rl.data.datasets.response_datasets.openr1_math.OpenR1Math220KDataset method) (nemo_rl.data.datasets.response_datasets.refcoco.RefCOCODataset method) (nemo_rl.data.datasets.response_datasets.response_dataset.ResponseDataset method) (nemo_rl.data.datasets.response_datasets.squad.SquadDataset method) (nemo_rl.data.datasets.response_datasets.tulu3.Tulu3SftMixtureDataset method) format_dynamo_error() (in module nemo_rl.models.generation.dynamo.http_client) format_geometry3k_dataset() (in module nemo_rl.data.datasets.response_datasets.geometry3k) format_mmpr_tiny_dataset() (in module nemo_rl.data.datasets.response_datasets.mmpr_tiny) format_prompt_for_vllm_generation() (in module nemo_rl.models.generation.vllm.utils) format_refcoco_dataset() (in module nemo_rl.data.datasets.response_datasets.refcoco) format_result() (nemo_rl.environments.code_environment.CodeExecutionWorker method) format_reward() (in module nemo_rl.environments.rewards) forward() (nemo_rl.algorithms.logits_sampling_utils._ApplyTopKTopP static method) (nemo_rl.algorithms.x_token.loss_utils.Fp32SparseMM static method) (nemo_rl.distributed.model_utils._AllReduceSum static method) (nemo_rl.distributed.model_utils.AllGatherCPTensor method) (nemo_rl.distributed.model_utils.ChunkedDistributedEntropy static method) (nemo_rl.distributed.model_utils.ChunkedDistributedGatherLogprob static method) (nemo_rl.distributed.model_utils.ChunkedDistributedHiddenStatesToLogprobs static method) (nemo_rl.distributed.model_utils.ChunkedDistributedLogprob static method) (nemo_rl.distributed.model_utils.ChunkedDistributedLogprobWithSampling static method) (nemo_rl.distributed.model_utils.DistributedCrossEntropy static method) (nemo_rl.distributed.model_utils.DistributedLogprob static method) (nemo_rl.distributed.model_utils.DistributedLogprobWithSampling static method) (nemo_rl.models.megatron.draft.eagle.EagleModel method) forward_with_post_processing_fn() (in module nemo_rl.models.automodel.train) (in module nemo_rl.models.megatron.train) fp16 (nemo_rl.models.policy.MegatronOptimizerConfig attribute) Fp32SparseMM (class in nemo_rl.algorithms.x_token.loss_utils) fp8 (nemo_rl.models.policy.Fp8Config attribute) fp8_cfg (nemo_rl.models.generation.megatron.config.MCoreGenerationSpecificArgs attribute) (nemo_rl.models.policy.MegatronConfig attribute) fp8_param (nemo_rl.models.policy.Fp8Config attribute) fp8_recipe (nemo_rl.models.policy.Fp8Config attribute) Fp8Config (class in nemo_rl.models.policy) freeze_audio_tower (nemo_rl.models.policy.AutomodelFreezeConfig attribute) freeze_config (nemo_rl.models.policy.AutomodelKwargs attribute) (nemo_rl.models.policy.MegatronConfig attribute) freeze_language_model (nemo_rl.models.policy.AutomodelFreezeConfig attribute) freeze_moe_router (nemo_rl.models.policy.MegatronConfig attribute) freeze_sound_encoder (nemo_rl.models.policy.MegatronConfig attribute) freeze_sound_projection (nemo_rl.models.policy.MegatronConfig attribute) freeze_vision_model (nemo_rl.models.policy.MegatronConfig attribute) freeze_vision_projection (nemo_rl.models.policy.MegatronConfig attribute) freeze_vision_tower (nemo_rl.models.policy.AutomodelFreezeConfig attribute) from_batches() (nemo_rl.distributed.batched_data_dict.BatchedDataDict class method) from_config() (nemo_rl.utils.flops_tracker.FLOPTracker class method) from_generation_config() (nemo_rl.models.generation.interfaces.GenerationSamplingParams class method) from_metadata() (nemo_rl.models.generation.dynamo.refit.DynamoWorkerEndpoint class method) from_parallel_hidden_states_to_logprobs() (in module nemo_rl.distributed.model_utils) from_parallel_logits_to_logprobs() (in module nemo_rl.distributed.model_utils) from_parallel_logits_to_logprobs_packed_sequences() (in module nemo_rl.distributed.model_utils) frontend_args (nemo_rl.models.generation.dynamo.config.DynamoCfg attribute) frontend_url (nemo_rl.models.generation.dynamo.dynamo_generation.DynamoGeneration property) (nemo_rl.models.generation.dynamo.managed_runtime.ManagedDynamoRuntime property) FRONTIER_ORDINAL_KEY (in module nemo_rl.experience.interfaces) FSDP (nemo_rl.distributed.virtual_cluster.PY_EXECUTABLES attribute) fsdp2_config (nemo_rl.models.automodel.config.DistributedContext attribute) ft_keep_latest_k (nemo_rl.utils.checkpoint.CheckpointingConfig attribute) ft_save_period (nemo_rl.utils.checkpoint.CheckpointingConfig attribute) FullLogitsPostProcessor (class in nemo_rl.models.automodel.train) fused_linear_logprobs_chunk_size (nemo_rl.models.policy.MegatronConfig attribute) futures (nemo_rl.distributed.worker_groups.MultiWorkerFuture attribute) (nemo_rl.utils.weight_transfer_zmq._RelayTransfer attribute) G G_ROUTED_EXPERTS_RANGE_CHECKED (in module nemo_rl.models.generation.vllm.utils) G_VLLM_REFIT_API_KEY_HEADER (in module nemo_rl.utils.weight_transfer_http) G_VLLM_REFIT_FLUSH_PATH (in module nemo_rl.utils.weight_transfer_http) G_VLLM_REFIT_PREPARE_PATH (in module nemo_rl.utils.weight_transfer_http) G_VLLM_REFIT_S3_MANIFEST_PATH (in module nemo_rl.utils.weight_transfer_http) G_VLLM_REFIT_ZMQ_FLUSH_PATH (in module nemo_rl.utils.weight_transfer_http) gae_gamma (nemo_rl.algorithms.advantage_estimator.GAEConfig attribute) gae_lambda (nemo_rl.algorithms.advantage_estimator.GAEConfig attribute) gae_lambda_policy (nemo_rl.algorithms.advantage_estimator.GAEConfig attribute) gae_lambda_value (nemo_rl.algorithms.advantage_estimator.GAEConfig attribute) GAEConfig (class in nemo_rl.algorithms.advantage_estimator) gate_precision (nemo_rl.models.policy.AutomodelBackendConfig attribute) gate_weight (nemo_rl.models.megatron.draft.utils._PendingLayerWeights attribute) gather_jagged_object_lists() (in module nemo_rl.distributed.collectives) gather_logits_at_global_indices() (in module nemo_rl.distributed.model_utils) gbs (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) GDPOAdvantageEstimator (class in nemo_rl.algorithms.advantage_estimator) gen_handle (nemo_rl.algorithms.single_controller_utils.setup.SingleControllerActorArgs attribute) GeneralConversationsJsonlDataset (class in nemo_rl.data.datasets.response_datasets.general_conversations_dataset) GeneralizedAdvantageEstimator (class in nemo_rl.algorithms.advantage_estimator) generate() (nemo_rl.modelopt.models.policy.workers.megatron_quant_policy_worker.MegatronQuantPolicyWorker method) (nemo_rl.models.generation.dynamo.dynamo_generation.DynamoGeneration method) (nemo_rl.models.generation.interfaces.GenerationInterface method) (nemo_rl.models.generation.megatron.megatron_generation.MegatronGeneration method) (nemo_rl.models.generation.megatron.megatron_worker.MegatronGenerationMixin method) (nemo_rl.models.generation.trtllm.trtllm_generation.TrtllmGeneration method) (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) (nemo_rl.models.generation.vllm.vllm_worker.VllmGenerationWorkerImpl method) (nemo_rl.models.policy.lm_policy.Policy method) generate_and_push() (nemo_rl.experience.rollout_manager.RolloutManager method) generate_async() (nemo_rl.models.generation.dynamo.dynamo_generation.DynamoGeneration method) (nemo_rl.models.generation.megatron.megatron_generation.MegatronGeneration method) (nemo_rl.models.generation.megatron.megatron_worker.MegatronGenerationMixin method) (nemo_rl.models.generation.trtllm.trtllm_generation.TrtllmGeneration method) (nemo_rl.models.generation.trtllm.trtllm_worker_async.TrtllmAsyncGenerationWorkerImpl method) (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) (nemo_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) generate_responses() (in module nemo_rl.experience.rollouts) generate_responses_async() (in module nemo_rl.experience.rollouts) generate_text() (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) (nemo_rl.models.generation.vllm.vllm_worker.VllmGenerationWorkerImpl method) generate_text_async() (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) (nemo_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) Generation (in module nemo_rl.algorithms.single_controller) generation (nemo_rl.environments.reward_model_environment.RewardModelEnvironmentConfig attribute) (nemo_rl.evals.eval.MasterConfig attribute) (nemo_rl.models.policy.PolicyConfig attribute) GENERATION (nemo_rl.telemetry.span_groups.RLSpanGroup attribute) generation_batch_size (nemo_rl.models.policy.PolicyConfig attribute) generation_fleet_health (nemo_rl.algorithms.single_controller_utils.config.AsyncRLConfig attribute) generation_init_load_time_s (nemo_rl.algorithms.metric_utils.SetupTimingMetrics attribute) generation_init_reserve_time_s (nemo_rl.algorithms.metric_utils.SetupTimingMetrics attribute) generation_init_time_s (nemo_rl.algorithms.metric_utils.SetupTimingMetrics attribute) generation_lengths (nemo_rl.models.generation.interfaces.GenerationOutputSpec attribute) generation_logprobs (nemo_rl.algorithms.loss.loss_functions.ClippedPGLossDataDict attribute) generation_logprobs_field (nemo_rl.algorithms.single_controller_utils.config.AdvantageConfig attribute) generation_router (nemo_rl.algorithms.single_controller_utils.config.AsyncRLConfig attribute) (nemo_rl.algorithms.single_controller_utils.setup.SingleControllerActorArgs attribute) generation_s (nemo_rl.experience.rollout_manager.RolloutTimeouts attribute) generation_timeout_s (nemo_rl.algorithms.single_controller_utils.config.NativeRolloutFTConfig attribute) GENERATION_WORKER_OVERRIDES (in module nemo_rl.models.generation.vllm.utils) GenerationConfig (class in nemo_rl.models.generation.interfaces) GenerationDatumSpec (class in nemo_rl.models.generation.interfaces) GenerationFleetExhausted GenerationFleetHealth (class in nemo_rl.models.generation.fleet_health) GenerationInterface (class in nemo_rl.models.generation.interfaces) GenerationOutputSpec (class in nemo_rl.models.generation.interfaces) GenerationRouterActor (class in nemo_rl.models.generation.generation_router) GenerationRouterConfig (class in nemo_rl.algorithms.single_controller_utils.config) GenerationRouterImpl (class in nemo_rl.models.generation.generation_router) GenerationSamplingParams (class in nemo_rl.models.generation.interfaces) GenerationUnavailable geo3k_reward() (in module nemo_rl.environments.rewards) Geometry3KDataset (class in nemo_rl.data.datasets.response_datasets.geometry3k) get() (nemo_rl.utils.weight_transfer_stream._S3ObjectStore method) (nemo_rl.weight_sync.nccl_reshard_utils.HFToLocalParamMap method) get_actor_python_env() (in module nemo_rl.distributed.ray_actor_environment_registry) get_agent_metadata() (nemo_rl.utils.checkpoint_engines.nixl.NixlAgent method) get_aggregated_metrics() (nemo_rl.data.packing.algorithms.SequencePacker method) get_aggregated_stats() (nemo_rl.data.packing.metrics.PackingMetrics method) get_all_worker_results() (nemo_rl.distributed.worker_groups.RayWorkerGroup method) get_and_validate_seqlen() (in module nemo_rl.models.megatron.data) get_attached_draft_model() (in module nemo_rl.models.megatron.draft.utils) get_aux_loss_track_names() (in module nemo_rl.models.megatron.common) get_available_address_and_port() (nemo_rl.distributed.virtual_cluster.RayVirtualCluster method) get_available_addresses_and_ports_batch() (nemo_rl.distributed.virtual_cluster.RayVirtualCluster method) get_axis_index() (nemo_rl.distributed.named_sharding.NamedSharding method) get_axis_size() (nemo_rl.distributed.named_sharding.NamedSharding method) get_batch() (nemo_rl.distributed.batched_data_dict.BatchedDataDict method) get_best_checkpoint_path() (nemo_rl.utils.checkpoint.CheckpointManager method) get_capture_context() (in module nemo_rl.models.megatron.draft.hidden_capture) get_captured_states() (nemo_rl.models.megatron.draft.hidden_capture.HiddenStateCapture method) get_checkpoint_dataloader_state() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector method) get_checkpoint_state() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector method) get_collective_sender_spec() (nemo_rl.models.generation.dynamo.dynamo_generation.DynamoGeneration method) (nemo_rl.models.generation.interfaces.GenerationInterface method) get_cp_sharded_next_token_logprobs() (in module nemo_rl.distributed.model_utils) get_cpu_state_dict() (in module nemo_rl.models.policy.workers.dtensor_policy_worker) get_data() (nemo_rl.data_plane.adapters.noop.NoOpDataPlaneClient method) (nemo_rl.data_plane.adapters.transfer_queue.TQDataPlaneClient method) (nemo_rl.data_plane.interfaces.DataPlaneClient method) (nemo_rl.data_plane.observability.MetricsDataPlaneClient method) get_data_records() (in module nemo_rl.data.datasets.response_datasets.oasst) get_dataloader_state() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector method) get_debug_info() (nemo_rl.algorithms.async_utils.replay_buffer.ReplayBufferImpl method) get_device_uuid() (in module nemo_rl.utils.nvml) get_dict() (nemo_rl.distributed.batched_data_dict.BatchedDataDict method) get_dim_to_pack_along() (in module nemo_rl.data.multimodal_utils) get_distillation_topk_logprobs_from_logits() (in module nemo_rl.distributed.model_utils) get_dp_leader_worker_idx() (nemo_rl.distributed.worker_groups.RayWorkerGroup method) get_dynamo_executable() (in module nemo_rl.models.generation.dynamo.venv) get_dynamo_python() (in module nemo_rl.models.generation.dynamo.venv) get_dynamo_venv_dir() (in module nemo_rl.models.generation.dynamo.venv) get_eagle3_aux_hidden_state_layers() (in module nemo_rl.models.megatron.draft.hidden_capture) get_efficiency_metrics() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector method) get_elapsed() (nemo_rl.utils.timer.ThreadSafeTimer method) (nemo_rl.utils.timer.Timer method) get_existing_target_weights() (nemo_rl.algorithms.async_utils.replay_buffer.ReplayBufferImpl method) get_first_index_that_differs() (in module nemo_rl.data.llm_message_utils) get_flash_attention_kwargs() (in module nemo_rl.models.huggingface.common) get_formatted_message_log() (in module nemo_rl.data.llm_message_utils) get_forward_loop_func() (in module nemo_rl.modelopt.models.policy.workers.utils) get_free_memory_bytes() (in module nemo_rl.utils.nvml) (nemo_rl.models.policy.lm_policy.Policy method) (nemo_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker method) get_full_logits_ipc() (nemo_rl.models.policy.lm_policy.Policy method) (nemo_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl method) get_gdpo_reward_component_keys() (in module nemo_rl.algorithms.utils) get_gpu_info() (in module nemo_rl.models.policy.utils) (nemo_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker method) get_grad_norm() (in module nemo_rl.models.dtensor.parallelize) get_group() (nemo_rl.experience.rollout_recovery.RolloutRecoveryLedger method) get_handle_from_tensor() (in module nemo_rl.models.policy.utils) get_held_task_indices() (nemo_rl.algorithms.async_utils.replay_buffer.ReplayBufferImpl method) get_hf_config() (in module nemo_rl.utils.flops_tracker) get_hf_tp_plan() (in module nemo_rl.models.dtensor.parallelize) get_huggingface_cache_path() (in module nemo_rl.data.datasets.utils) get_inference_cuda_graph_capture_count() (nemo_rl.models.generation.megatron.megatron_worker.MegatronGenerationMixin method) get_inference_world_size() (nemo_rl.models.generation.dynamo.dynamo_generation.DynamoGeneration method) (nemo_rl.models.generation.interfaces.GenerationInterface method) get_keys_from_message_log() (in module nemo_rl.data.llm_message_utils) get_last_target_weight_already_generated() (nemo_rl.algorithms.async_utils.replay_buffer.ReplayBufferImpl method) get_latest_checkpoint_path() (nemo_rl.utils.checkpoint.CheckpointManager method) get_latest_elapsed() (nemo_rl.utils.timer.ThreadSafeTimer method) (nemo_rl.utils.timer.Timer method) get_logger_metrics() (nemo_rl.models.generation.dynamo.dynamo_generation.DynamoGeneration method) (nemo_rl.models.generation.interfaces.GenerationInterface method) (nemo_rl.models.generation.trtllm.trtllm_generation.TrtllmGeneration method) (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) get_logprobs() (nemo_rl.models.policy.interfaces.PolicyInterface method) (nemo_rl.models.policy.lm_policy.Policy method) (nemo_rl.models.policy.teacher_worker_group.TeacherWorkerGroup method) (nemo_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) (nemo_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl method) (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) get_logprobs_from_meta() (nemo_rl.models.policy.teacher_worker_group.TeacherWorkerGroup method) (nemo_rl.models.policy.tq_policy.TQPolicy method) get_logprobs_from_vocab_parallel_logits() (in module nemo_rl.distributed.model_utils) get_logprobs_presharded() (nemo_rl.data_plane.worker_mixin.TQWorkerMixin method) get_ltor_masks_and_position_ids() (in module nemo_rl.models.megatron.data) get_markers() (nemo_rl.utils.timer.ThreadSafeTimer method) (nemo_rl.utils.timer.Timer method) get_master_address_and_port() (nemo_rl.distributed.virtual_cluster.RayVirtualCluster method) get_media_from_message() (in module nemo_rl.data.multimodal_utils) get_megatron_checkpoint_dir() (in module nemo_rl.models.policy.utils) get_microbatch_iterator() (in module nemo_rl.models.automodel.data) (in module nemo_rl.models.megatron.data) get_microbatch_iterator_dynamic_shapes_len() (nemo_rl.distributed.batched_data_dict.BatchedDataDict method) get_microbatch_iterator_for_packable_sequences_len() (nemo_rl.distributed.batched_data_dict.BatchedDataDict method) get_modelopt_checkpoint_dir() (in module nemo_rl.modelopt.models.policy.workers.utils) get_moe_metrics() (in module nemo_rl.models.megatron.common) get_mtp_metrics() (in module nemo_rl.models.megatron.common) get_multimodal_default_settings_from_processor() (in module nemo_rl.data.multimodal_utils) get_multimodal_dict() (nemo_rl.distributed.batched_data_dict.BatchedDataDict method) get_multimodal_keys_from_processor() (in module nemo_rl.data.multimodal_utils) get_nemo_gym_thinking_tags() (in module nemo_rl.experience.rollouts) get_nemo_gym_uv_cache_dir() (in module nemo_rl.environments.nemo_gym) get_nemo_gym_venv_dir() (in module nemo_rl.environments.nemo_gym) get_next_experiment_dir() (in module nemo_rl.utils.logger) get_next_token_logprobs_from_logits() (in module nemo_rl.distributed.model_utils) get_nsight_config_if_pattern_matches() (in module nemo_rl.distributed.worker_group_utils) get_num_buffers() (in module nemo_rl.utils.packed_tensor) get_num_routed_experts() (in module nemo_rl.models.generation.interfaces) get_packer() (in module nemo_rl.data.packing.algorithms) get_pad_dynamic_image_shapes() (in module nemo_rl.environments.nemo_gym) get_pad_to_max_shape() (in module nemo_rl.data.multimodal_utils) get_placement_groups() (nemo_rl.distributed.virtual_cluster.RayVirtualCluster method) get_placements() (in module nemo_rl.weight_sync.nccl_reshard_utils) get_policy_lm_head_weight() (in module nemo_rl.models.megatron.draft.utils) get_prompt() (nemo_rl.data.datasets.response_datasets.daily_omni.DailyOmniDataset class method) get_quantization_layer_spec() (in module nemo_rl.modelopt.models.policy.workers.utils) get_quantization_mamba_stack_spec() (in module nemo_rl.modelopt.models.policy.workers.utils) get_quantizer_stats() (nemo_rl.modelopt.models.generation.vllm_quant_backend.VllmQuantInternalWorkerExtension method) (nemo_rl.modelopt.models.generation.vllm_quant_worker.VllmQuantAsyncGenerationWorker method) (nemo_rl.modelopt.models.generation.vllm_quant_worker.VllmQuantGenerationWorker method) (nemo_rl.modelopt.models.policy.workers.megatron_quant_policy_worker.MegatronQuantPolicyWorker method) get_ranks() (nemo_rl.distributed.named_sharding.NamedSharding method) get_ranks_by_coord() (nemo_rl.distributed.named_sharding.NamedSharding method) get_ray_cluster_topology() (in module nemo_rl.distributed.virtual_cluster) get_reference_policy_logprobs() (nemo_rl.models.policy.interfaces.PolicyInterface method) (nemo_rl.models.policy.lm_policy.Policy method) (nemo_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker method) (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) get_reference_policy_logprobs_from_meta() (nemo_rl.models.policy.tq_policy.TQPolicy method) get_reference_policy_logprobs_presharded() (nemo_rl.data_plane.worker_mixin.TQWorkerMixin method) get_reordered_bundle() (in module nemo_rl.distributed.virtual_cluster) get_reserved_url() (nemo_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) get_results() (nemo_rl.distributed.worker_groups.MultiWorkerFuture method) get_resume_paths() (nemo_rl.utils.checkpoint.CheckpointManager static method) get_rollouts_state() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector method) get_runtime_env_for_policy_worker() (in module nemo_rl.models.policy.utils) get_samples() (nemo_rl.data_plane.adapters.noop.NoOpDataPlaneClient method) (nemo_rl.data_plane.adapters.transfer_queue.TQDataPlaneClient method) (nemo_rl.data_plane.interfaces.DataPlaneClient method) (nemo_rl.data_plane.observability.MetricsDataPlaneClient method) get_snapshot_str() (nemo_rl.utils.memory_tracker.MemoryTrackerDataPoint method) get_sparse_projection_matrix() (in module nemo_rl.algorithms.x_token.loss_utils) get_status() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector method) get_step_metrics() (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) get_target_packed_tensor_size() (in module nemo_rl.utils.packed_tensor) get_target_weight_layout() (nemo_rl.utils.checkpoint_engines.base.CheckpointEngine method) (nemo_rl.utils.checkpoint_engines.nixl.NIXLCheckpointEngine method) get_teacher_logprobs_presharded() (nemo_rl.data_plane.worker_mixin.TQWorkerMixin method) get_teacher_routing_metrics() (in module nemo_rl.algorithms.opd) get_telemetry_handle() (in module nemo_rl.telemetry.setup) get_theoretical_tflops() (in module nemo_rl.utils.flops_tracker) get_timing_metrics() (nemo_rl.utils.timer.ThreadSafeTimer method) (nemo_rl.utils.timer.Timer method) get_tokenizer() (in module nemo_rl.algorithms.utils) (in module nemo_rl.modelopt.models.policy.workers.utils) (in module nemo_rl.models.automodel.setup) get_topk_logits() (nemo_rl.models.policy.interfaces.PolicyInterface method) (nemo_rl.models.policy.lm_policy.Policy method) (nemo_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) (nemo_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl method) (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) get_topk_projection() (in module nemo_rl.algorithms.x_token.loss_utils) get_tp_shard_dim() (in module nemo_rl.weight_sync.nccl_reshard_utils) get_train_dataset_name() (in module nemo_rl.data.utils) get_trajectories_needed() (nemo_rl.algorithms.async_utils.interfaces.ReplayBufferProtocol method) (nemo_rl.algorithms.async_utils.replay_buffer.ReplayBufferImpl method) get_values() (nemo_rl.models.value.interfaces.ValueInterface method) (nemo_rl.models.value.lm_value.Value method) (nemo_rl.models.value.workers.dtensor_value_worker_v2.DTensorValueWorkerV2Impl method) (nemo_rl.models.value.workers.megatron_value_worker.MegatronValueWorkerImpl method) get_values_from_meta() (nemo_rl.models.value.tq_value.TQValue method) get_values_presharded() (nemo_rl.data_plane.worker_mixin.TQWorkerMixin method) get_vllm_logger_metrics() (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) (nemo_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) get_weight_snapshot() (nemo_rl.modelopt.models.generation.vllm_quant_backend.VllmQuantInternalWorkerExtension method) (nemo_rl.modelopt.models.generation.vllm_quant_worker.VllmQuantAsyncGenerationWorker method) (nemo_rl.modelopt.models.generation.vllm_quant_worker.VllmQuantGenerationWorker method) get_weight_version() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector method) get_worker_coords() (nemo_rl.distributed.named_sharding.NamedSharding method) get_zmq_address() (nemo_rl.models.generation.trtllm.trtllm_backend.NcclExtension method) (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) (nemo_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker method) git_root (in module nemo_rl.distributed.virtual_cluster) (in module nemo_rl.utils.venvs) glm_moe_dsa() (in module nemo_rl.utils.flops_formulas) GLOBAL_FORWARD_PAD_SEQLEN (in module nemo_rl.data_plane.schema) global_post_process_and_metrics() (nemo_rl.environments.code_environment.CodeEnvironment method) (nemo_rl.environments.code_jaccard_environment.CodeJaccardEnvironment method) (nemo_rl.environments.interfaces.EnvironmentInterface method) (nemo_rl.environments.math_environment.BaseMathEnvironment method) (nemo_rl.environments.nemo_gym.NemoGym method) (nemo_rl.environments.reward_model_environment.RewardModelEnvironment method) (nemo_rl.environments.vlm_environment.VLMEnvironment method) global_segment_size (nemo_rl.data_plane.interfaces.MooncakeCpuConfig attribute) global_valid_seqs (nemo_rl.algorithms.dpo.DPOValMetrics attribute) global_valid_toks (nemo_rl.algorithms.dpo.DPOValMetrics attribute) gold_loss (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute) (nemo_rl.algorithms.xtoken_off_policy_distillation.TeacherConfig attribute) goodput_span_attributes() (in module nemo_rl.telemetry.instrumentation) GPQADataset (class in nemo_rl.data.datasets.eval_datasets.gpqa) GPQAEvalDataConfig (class in nemo_rl.data) gpt3() (in module nemo_rl.utils.flops_formulas) GPU_CPU_AFFINITY_PATH (in module nemo_rl.distributed.numa_utils) gpu_memory_utilization (nemo_rl.models.generation.dynamo.config.DynamoVllmConfig attribute) (nemo_rl.models.generation.trtllm.config.TrtllmSpecificArgs attribute) (nemo_rl.models.generation.vllm.config.VllmSpecificArgs attribute) gpu_monitoring (nemo_rl.utils.logger.LoggerConfig attribute) GpuMetricSnapshot (class in nemo_rl.utils.logger) GPUMonitoringConfig (class in nemo_rl.utils.logger) gpus_per_node (nemo_rl.algorithms.opd.TeacherResourceConfig attribute) (nemo_rl.distributed.virtual_cluster.ClusterConfig attribute) (nemo_rl.models.generation.interfaces.OptionalResourcesConfig attribute) (nemo_rl.models.generation.interfaces.ResourcesConfig attribute) (nemo_rl.models.policy.teacher_worker_group.TeacherConfig attribute) grad_reduce_in_fp32 (nemo_rl.models.policy.MegatronDDPConfig attribute) gradient_accumulation_fusion (nemo_rl.models.policy.MegatronConfig attribute) ground_truth (nemo_rl.environments.code_jaccard_environment.CodeJaccardEnvironmentMetadata attribute) (nemo_rl.environments.math_environment.MathEnvironmentMetadata attribute) (nemo_rl.environments.vlm_environment.VLMEnvironmentMetadata attribute) group_all_reduce_sum() (in module nemo_rl.distributed.model_utils) group_all_reduce_sum_with_grad() (in module nemo_rl.distributed.model_utils) group_and_cat_tensors() (in module nemo_rl.models.huggingface.common) group_expert_params_in_metadata() (in module nemo_rl.weight_sync.nccl_reshard_utils) group_id (nemo_rl.algorithms.async_utils.replay_buffer.TQReplayGroupMetadata attribute) (nemo_rl.experience.rollout_recovery.PromptGroupRecoveryRecord attribute) (nemo_rl.experience.rollout_recovery.PromptGroupRecoveryState attribute) group_index (nemo_rl.experience.rollouts._CompletedNemoGymGroup attribute) (nemo_rl.experience.rollouts.RolloutGroupResult attribute) groups (nemo_rl.algorithms.async_utils.replay_buffer.TQReplayMetadataState attribute) (nemo_rl.experience.rollout_recovery.RolloutRecoveryLedgerState attribute) groups() (nemo_rl.experience.rollout_recovery.RolloutRecoveryLedger method) grpo (nemo_rl.algorithms.grpo.MasterConfig attribute) (nemo_rl.algorithms.single_controller_utils.config.MasterConfig attribute) grpo_group_size (nemo_rl.data_plane.adapters.noop._Partition attribute) grpo_train() (in module nemo_rl.algorithms.grpo) grpo_train_sync() (in module nemo_rl.algorithms.grpo_sync) GRPOAdvantageEstimator (class in nemo_rl.algorithms.advantage_estimator) GRPOConfig (class in nemo_rl.algorithms.grpo) GRPOLoggerConfig (class in nemo_rl.algorithms.grpo) GRPOSaveState (class in nemo_rl.algorithms.grpo) GSM8KDataset (class in nemo_rl.data.datasets.response_datasets.gsm8k) gym_row_redispatches (nemo_rl.experience.rollout_manager.RolloutStats attribute) gym_subprocess_check (nemo_rl.algorithms.single_controller_utils.config.WatchdogConfig attribute) GymTransportError H handle_model_import() (in module nemo_rl.models.megatron.setup) has_complete_batch() (nemo_rl.algorithms.async_utils.interfaces.ReplayBufferProtocol method) (nemo_rl.algorithms.async_utils.replay_buffer.ReplayBufferImpl method) has_flash_attention (nemo_rl.models.automodel.data.ProcessedInputs property) has_pending_finalization (nemo_rl.utils.checkpoint.CheckpointManager property) head_dim (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) health_check() (nemo_rl.environments.nemo_gym.NemoGym method) HEALTHY (nemo_rl.models.generation.fleet_health.ShardState attribute) healthy_threshold (nemo_rl.algorithms.single_controller_utils.config.FleetHealthConfig attribute) (nemo_rl.models.generation.fleet_health.FleetHealthPolicy attribute) HealthyShardSelector (class in nemo_rl.models.generation.fleet_health) HeldPortReservation (class in nemo_rl.distributed.held_port) helpsteer3_data_processor() (in module nemo_rl.data.processors) HelpSteer3Dataset (class in nemo_rl.data.datasets.preference_datasets.helpsteer3) (class in nemo_rl.data.datasets.response_datasets.helpsteer3) hf_config_overrides (nemo_rl.models.automodel.config.RuntimeConfig attribute) (nemo_rl.models.policy.PolicyConfig attribute) (nemo_rl.models.value.config.ValueConfig attribute) HfExpertWeight (class in nemo_rl.models.generation.vllm.refit_layout) HFMultiRewardVerifyWorker (class in nemo_rl.environments.math_environment) HFToLocalParamMap (class in nemo_rl.weight_sync.nccl_reshard_utils) HFVerifyWorker (class in nemo_rl.environments.math_environment) hidden_norm_key (nemo_rl.models.megatron.draft.utils._EagleLayerLayout attribute) hidden_states (nemo_rl.models.megatron.draft.hidden_capture.CapturedStates attribute) HiddenStateCapture (class in nemo_rl.models.megatron.draft.hidden_capture) hide_tensor_quantizers() (nemo_rl.modelopt.models.policy.workers.megatron_quant_policy_worker.MegatronQuantPolicyWorker method) high_lengths (nemo_rl.experience.rollouts._EffortShapingMetrics attribute) higher_is_better (nemo_rl.utils.checkpoint.CheckpointingConfig attribute) hold_refit_for_fault_injection() (in module nemo_rl.distributed.refit_watchdog) hs (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) http_post_json() (in module nemo_rl.models.generation.dynamo.http_client) http_refit_api_key_env_var (nemo_rl.models.generation.vllm.config.VllmSpecificArgs attribute) http_refit_server_port (nemo_rl.models.generation.vllm.config.VllmSpecificArgs attribute) http_server_serving_chat_kwargs (nemo_rl.models.generation.vllm.config.VllmSpecificArgs attribute) http_status_is_infra() (in module nemo_rl.experience.failures) hybrid_override_pattern (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) hybridep_num_ranks_per_nvlink_domain (nemo_rl.models.policy.MegatronConfig attribute) hybridep_use_mnnvl (nemo_rl.models.policy.MegatronConfig attribute) I IDLE (nemo_rl.telemetry.instrumentation.Bucket attribute) idx (nemo_rl.data.interfaces.DatumSpec attribute) (nemo_rl.data.interfaces.PreferenceDatumSpec attribute) ignore_router_for_ac (nemo_rl.models.policy.MoEParallelizerOptions attribute) IMAGE_CONTENT_TYPES (in module nemo_rl.data.multimodal_utils) image_counts_by_row() (in module nemo_rl.data.multimodal_utils) image_to_data_url() (in module nemo_rl.data.multimodal_utils) img_h (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) img_seq_len (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) img_w (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) impl (nemo_rl.data_plane.interfaces.DataPlaneConfig attribute) import_model_from_hf_name() (in module nemo_rl.models.megatron.community_import) in_channels (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) in_flight_weight_updates (nemo_rl.algorithms.grpo.AsyncGRPOConfig attribute) (nemo_rl.algorithms.ppo.AsyncPPOConfig attribute) (nemo_rl.models.generation.trtllm.config.TrtllmSpecificArgs attribute) in_memory (nemo_rl.models.generation.vllm.config.VllmRefitBaselineConfig attribute) inference_cluster (nemo_rl.algorithms.single_controller_utils.setup.SingleControllerActorArgs attribute) inference_cuda_graph_scope (nemo_rl.models.generation.megatron.config.MCoreGenerationSpecificArgs attribute) inference_grouped_gemm_backend (nemo_rl.models.policy.MegatronConfig attribute) inference_megatron_cfg (nemo_rl.models.megatron.config.ColocatedReshardPlan attribute) inference_model (nemo_rl.models.generation.megatron.megatron_worker.MegatronGenerationMixin attribute) inference_model_alloc_region() (in module nemo_rl.models.megatron.memory_saver) inference_moe_token_dispatcher_type (nemo_rl.models.policy.MegatronConfig attribute) inference_world_size (nemo_rl.models.generation.dynamo.refit.DynamoRefitChannel property) inflight() (nemo_rl.models.generation.fleet_health.HealthyShardSelector method) INFRA (nemo_rl.experience.failures.FailureClass attribute) infra_drops_by_reason (nemo_rl.experience.rollout_manager.RolloutStats attribute) init_checkpoint_engine() (nemo_rl.models.generation.vllm.checkpoint_engine.VllmCheckpointEngineMixin method) init_checkpoint_engine_process_group() (nemo_rl.models.generation.vllm.checkpoint_engine.VllmCheckpointEngineMixin method) init_checkpointer() (nemo_rl.models.automodel.checkpoint.AutomodelCheckpointManager method) init_cluster_placement_groups() (nemo_rl.models.generation.megatron.megatron_generation.MegatronGeneration class method) (nemo_rl.models.generation.trtllm.trtllm_generation.TrtllmGeneration static method) (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration static method) init_collective() (nemo_rl.models.generation.dynamo.dynamo_generation.DynamoGeneration method) (nemo_rl.models.generation.dynamo.refit.DynamoRefitChannel method) (nemo_rl.models.generation.interfaces.GenerationInterface method) (nemo_rl.models.generation.megatron.megatron_generation.MegatronGeneration method) (nemo_rl.models.generation.trtllm.trtllm_backend.NcclExtension method) (nemo_rl.models.generation.trtllm.trtllm_generation.TrtllmGeneration method) (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) (nemo_rl.models.generation.vllm.vllm_worker.VllmGenerationWorkerImpl method) (nemo_rl.models.policy.interfaces.ColocatablePolicyInterface method) (nemo_rl.models.policy.lm_policy.Policy method) (nemo_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker method) init_collective_async() (nemo_rl.models.generation.trtllm.trtllm_worker_async.TrtllmAsyncGenerationWorkerImpl method) (nemo_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) init_collective_mcore_generation() (nemo_rl.models.generation.megatron.megatron_worker.MegatronGenerationRefitMixin method) (nemo_rl.models.policy.lm_policy.Policy method) init_communicator() (nemo_rl.weight_sync.checkpoint_engine_weight_synchronizer.CheckpointEngineWeightSynchronizer method) (nemo_rl.weight_sync.collective_weight_synchronizer.CollectiveWeightSynchronizer method) (nemo_rl.weight_sync.interfaces.WeightSynchronizer method) (nemo_rl.weight_sync.ipc_weight_synchronizer.IPCWeightSynchronizer method) (nemo_rl.weight_sync.megatron_weight_synchronizer.MegatronWeightSynchronizer method) (nemo_rl.weight_sync.nccl_reshard_weight_synchronizer.NcclReshardWeightSynchronizer method) (nemo_rl.weight_sync.sglang_weight_synchronizer._SGLangWeightSynchronizer method) (nemo_rl.weight_sync.vllm_remote_sparse_weight_synchronizer.VllmRemoteSparseWeightSynchronizer method) init_nccl_communicator() (nemo_rl.distributed.stateless_process_group.StatelessProcessGroup method) init_nccl_reshard_comm_group() (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) (nemo_rl.models.generation.vllm.vllm_worker.VllmGenerationWorkerImpl method) (nemo_rl.models.policy.lm_policy.Policy method) (nemo_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker method) init_nccl_reshard_comm_group_async() (nemo_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) init_policy_process_group() (nemo_rl.utils.checkpoint_engines.base.CheckpointEngine method) (nemo_rl.utils.checkpoint_engines.nixl.NIXLCheckpointEngine method) init_process_group() (in module nemo_rl.models.policy.utils) init_ray() (in module nemo_rl.distributed.virtual_cluster) init_remote_sparse_delta_baseline() (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) init_rollout_process_group() (nemo_rl.utils.checkpoint_engines.base.CheckpointEngine method) (nemo_rl.utils.checkpoint_engines.nixl.NIXLCheckpointEngine method) init_sparse_delta_baseline_from_iterator() (in module nemo_rl.utils.weight_transfer_stream) init_telemetry_driver() (in module nemo_rl.telemetry.setup) init_telemetry_worker() (in module nemo_rl.telemetry.setup) init_tmp_checkpoint() (nemo_rl.utils.checkpoint.CheckpointManager method) initial_global_config_dict (nemo_rl.environments.nemo_gym.NemoGymConfig attribute) initial_model_provider (nemo_rl.models.megatron.config.ColocatedReshardPlan attribute) initialize_baseline() (nemo_rl.models.policy.workers.megatron_remote_sparse_refit.MegatronRemoteSparseRefit method) InOrderSampler (class in nemo_rl.algorithms.async_utils.staleness_sampler) InOrderSamplerConfig (class in nemo_rl.algorithms.async_utils.staleness_sampler) inp_s (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) INPUT_IDS (in module nemo_rl.data_plane.schema) input_ids (nemo_rl.algorithms.loss.loss_functions.ClippedPGLossDataDict attribute) (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossDataDict attribute) (nemo_rl.algorithms.loss.loss_functions.DistillationLossDataDict attribute) (nemo_rl.algorithms.loss.loss_functions.DPOLossDataDict attribute) (nemo_rl.algorithms.loss.loss_functions.PreferenceLossDataDict attribute) (nemo_rl.experience.rollouts.NemoGymRolloutResult attribute) (nemo_rl.models.automodel.data.ProcessedInputs attribute) (nemo_rl.models.generation.interfaces.GenerationDatumSpec attribute) (nemo_rl.models.megatron.data.ProcessedInputs attribute) (nemo_rl.models.megatron.data.ProcessedMicrobatch attribute) input_ids_cp_sharded (nemo_rl.models.megatron.data.ProcessedInputs attribute) (nemo_rl.models.megatron.data.ProcessedMicrobatch attribute) input_key (nemo_rl.data.ResponseDatasetConfig attribute) (nemo_rl.distributed.batched_data_dict.DynamicBatchingArgs attribute) (nemo_rl.distributed.batched_data_dict.SequencePackingArgs attribute) input_layernorm_key (nemo_rl.models.megatron.draft.utils._EagleLayerLayout attribute) INPUT_LENGTHS (in module nemo_rl.data_plane.schema) input_lengths (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossDataDict attribute) (nemo_rl.algorithms.loss.loss_functions.DistillationLossDataDict attribute) (nemo_rl.models.generation.interfaces.GenerationDatumSpec attribute) input_lengths_key (nemo_rl.distributed.batched_data_dict.DynamicBatchingArgs attribute) (nemo_rl.distributed.batched_data_dict.SequencePackingArgs attribute) input_type (nemo_rl.algorithms.loss.interfaces.LossFunction attribute) (nemo_rl.algorithms.loss.loss_functions.ClippedPGLossFn attribute) (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossFn attribute) (nemo_rl.algorithms.loss.loss_functions.DistillationLossFn attribute) (nemo_rl.algorithms.loss.loss_functions.DPOLossFn attribute) (nemo_rl.algorithms.loss.loss_functions.DraftCrossEntropyLossFn attribute) (nemo_rl.algorithms.loss.loss_functions.MseValueLossFn attribute) (nemo_rl.algorithms.loss.loss_functions.NLLLossFn attribute) (nemo_rl.algorithms.loss.loss_functions.PreferenceLossFn attribute) inputs_embeds (nemo_rl.models.megatron.draft.hidden_capture.CapturedStates attribute) instance_id (nemo_rl.models.generation.dynamo.refit.DynamoWorkerEndpoint attribute) integer_dtype_for_element_size() (in module nemo_rl.utils.weight_transfer_sparse_codec) integer_view() (in module nemo_rl.utils.weight_transfer_sparse_codec) IntentBenchDataset (class in nemo_rl.data.datasets.response_datasets.intent) IntentDataset (class in nemo_rl.data.datasets.response_datasets.intent) IntentTrainDataset (class in nemo_rl.data.datasets.response_datasets.intent) interval_s (nemo_rl.algorithms.single_controller_utils.config.WatchdogConfig attribute) invalid_tool_call_advantage (nemo_rl.algorithms.grpo.GRPOConfig attribute) invalid_tool_call_patterns (nemo_rl.environments.nemo_gym.NemoGymConfig attribute) invalidate_kv_cache() (nemo_rl.models.generation.dynamo.dynamo_generation.DynamoGeneration method) (nemo_rl.models.generation.interfaces.GenerationInterface method) (nemo_rl.models.generation.megatron.megatron_generation.MegatronGeneration method) (nemo_rl.models.generation.trtllm.trtllm_generation.TrtllmGeneration method) (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) (nemo_rl.models.policy.lm_policy.Policy method) IPCProtocol (class in nemo_rl.models.policy.utils) IPCWeightManifestError IPCWeightSynchronizer (class in nemo_rl.weight_sync.ipc_weight_synchronizer) is_alive() (nemo_rl.models.generation.dynamo.dynamo_worker.DynamoVllmWorker method) (nemo_rl.models.generation.dynamo.worker_pool.FixedDynamoWorkerPool method) (nemo_rl.models.generation.trtllm.trtllm_worker_async.TrtllmAsyncGenerationWorkerImpl method) (nemo_rl.models.generation.vllm.vllm_worker.BaseVllmGenerationWorker method) (nemo_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker method) is_async (nemo_rl.utils.checkpoint.CheckpointingConfig attribute) is_axis_zero() (nemo_rl.distributed.named_sharding.NamedSharding static method) is_complete (nemo_rl.experience.rollouts._NemoGymStreamAccumulator property) is_correct (nemo_rl.algorithms.x_token.token_aligner.AlignmentPair attribute) is_correct_minerva() (in module nemo_rl.environments.dapo_math_verifier) is_correct_strict_box() (in module nemo_rl.environments.dapo_math_verifier) is_data_exhausted() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector method) is_expert_param() (in module nemo_rl.weight_sync.nccl_reshard_utils) is_gemma_model() (in module nemo_rl.models.huggingface.common) is_generation_colocated (nemo_rl.models.automodel.config.RuntimeConfig attribute) (nemo_rl.models.megatron.config.RuntimeConfig attribute) is_hf_model (nemo_rl.models.automodel.config.ModelAndOptimizerState attribute) is_histogram_metric() (in module nemo_rl.experience.metric_utils) is_hybrid_model (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) is_moe_model (nemo_rl.models.automodel.config.ModelAndOptimizerState attribute) is_multimodal (nemo_rl.models.automodel.data.ProcessedInputs property) is_mx (nemo_rl.models.generation.vllm.config.VllmSpecificArgs attribute) is_nano_nemotron_vl_model() (in module nemo_rl.models.huggingface.common) is_nccl_reshard_param() (in module nemo_rl.weight_sync.nccl_reshard_utils) is_non_colocated_teachers_enabled() (in module nemo_rl.algorithms.opd) is_on_policy (nemo_rl.algorithms.async_utils.staleness_sampler.BaseSampler property) (nemo_rl.algorithms.async_utils.staleness_sampler.PromptGroupSampler property) is_opd_enabled() (in module nemo_rl.algorithms.opd) is_peft (nemo_rl.utils.checkpoint.CheckpointingConfig attribute) is_ppo_run() (in module nemo_rl.algorithms.single_controller_utils.config) is_refit_abort() (in module nemo_rl.distributed.refit_watchdog) is_refit_context_lost() (in module nemo_rl.distributed.refit_watchdog) is_reward_model (nemo_rl.models.automodel.config.ModelAndOptimizerState attribute) (nemo_rl.models.automodel.config.RuntimeConfig attribute) is_serving (nemo_rl.models.generation.fleet_health.ShardHealth property) is_serving() (nemo_rl.models.generation.generation_router.GenerationRouterImpl method) is_stale (nemo_rl.weight_sync.checkpoint_engine_weight_synchronizer.CheckpointEngineWeightSynchronizer property) (nemo_rl.weight_sync.collective_weight_synchronizer.CollectiveWeightSynchronizer property) (nemo_rl.weight_sync.interfaces.WeightSynchronizer property) (nemo_rl.weight_sync.ipc_weight_synchronizer.IPCWeightSynchronizer property) (nemo_rl.weight_sync.megatron_weight_synchronizer.MegatronWeightSynchronizer property) (nemo_rl.weight_sync.nccl_reshard_weight_synchronizer.NcclReshardWeightSynchronizer property) (nemo_rl.weight_sync.sglang_weight_synchronizer._SGLangWeightSynchronizer property) (nemo_rl.weight_sync.vllm_remote_sparse_weight_synchronizer.VllmRemoteSparseWeightSynchronizer property) is_using_tf32() (in module nemo_rl.utils.flops_tracker) is_vllm_v1_engine_enabled() (in module nemo_rl.models.policy.utils) is_vlm (nemo_rl.models.policy.PolicyConfig attribute) iter_logical_segments() (nemo_rl.data.multimodal_utils.PackedTensor method) iter_named_tensor_buckets() (in module nemo_rl.models.policy.utils) iter_quant_ignore_name_candidates() (in module nemo_rl.modelopt.utils) iter_sparse_weight_chunks() (in module nemo_rl.utils.weight_transfer_stream) iter_vlm_config_overrides() (in module nemo_rl.models.megatron.community_import) K k_value (nemo_rl.evals.eval.EvalConfig attribute) k_weight (nemo_rl.models.megatron.draft.utils._PendingLayerWeights attribute) kd_data_processor() (in module nemo_rl.data.processors) kd_loss_mode (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute) keep_top_k (nemo_rl.utils.checkpoint.CheckpointingConfig attribute) kl_input_clamp_value (nemo_rl.algorithms.loss.loss_functions.ClippedPGLossConfig attribute) kl_loss_weight (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute) kl_output_clamp_value (nemo_rl.algorithms.loss.loss_functions.ClippedPGLossConfig attribute) kl_type (nemo_rl.algorithms.loss.loss_functions.DistillationLossConfig attribute) kv_cache_dtype (nemo_rl.models.generation.dynamo.config.DynamoVllmConfig attribute) (nemo_rl.models.generation.vllm.config.VllmSpecificArgs attribute) kv_cache_management_mode (nemo_rl.models.generation.megatron.config.MCoreGenerationSpecificArgs attribute) kv_first_write() (in module nemo_rl.data_plane.column_io) kv_lora_rank (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) KVBatchMeta (class in nemo_rl.data_plane.interfaces) kwargs (nemo_rl.models.policy.PytorchOptimizerConfig attribute) (nemo_rl.models.policy.SinglePytorchSchedulerConfig attribute) L last_boxed_only_string() (in module nemo_rl.environments.dapo_math_verifier) last_checkpoint_path (nemo_rl.algorithms.single_controller_utils.setup.SingleControllerActorArgs attribute) last_error (nemo_rl.models.generation.fleet_health.ShardHealth attribute) last_ok_at (nemo_rl.models.generation.fleet_health.ShardHealth attribute) last_put_bytes_per_key (nemo_rl.data_plane.observability.DataPlaneStats attribute) layer_by_index (nemo_rl.models.megatron.draft.utils._EagleModelLayout property) layer_index (nemo_rl.models.megatron.draft.utils._EagleLayerLayout attribute) layers (nemo_rl.models.megatron.draft.utils._EagleModelLayout attribute) (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) Layout (in module nemo_rl.data_plane.schema) layout (nemo_rl.distributed.named_sharding.NamedSharding property) ledger_state (nemo_rl.experience.rollout_recovery.ParsedRolloutRecoveryState attribute) LEGACY_REPLAY_BUFFER_FILENAME (in module nemo_rl.algorithms.async_utils.replay_buffer) length (nemo_rl.data.interfaces.DatumSpec attribute) length_adaptive_alpha (nemo_rl.algorithms.advantage_estimator.GAEConfig attribute) length_chosen (nemo_rl.data.interfaces.PreferenceDatumSpec attribute) length_rejected (nemo_rl.data.interfaces.PreferenceDatumSpec attribute) length_rewards_low (nemo_rl.experience.rollouts._EffortShapingMetrics attribute) linear (nemo_rl.models.policy.AutomodelBackendConfig attribute) list_sample_ids() (nemo_rl.data_plane.adapters.noop.NoOpDataPlaneClient method) (nemo_rl.data_plane.adapters.transfer_queue.TQDataPlaneClient method) (nemo_rl.data_plane.interfaces.DataPlaneClient method) (nemo_rl.data_plane.observability.MetricsDataPlaneClient method) llama() (in module nemo_rl.utils.flops_formulas) llm() (nemo_rl.models.generation.vllm.vllm_worker.BaseVllmGenerationWorker method) LLMMessageLogType (in module nemo_rl.data.interfaces) lm_head_key (nemo_rl.models.megatron.draft.utils._EagleModelLayout attribute) lm_head_precision (nemo_rl.models.policy.MoEParallelizerOptions attribute) load_and_start() (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) load_audio_from_file() (in module nemo_rl.data.datasets.utils) load_checkpoint() (in module nemo_rl.utils.native_checkpoint) (nemo_rl.data_plane.adapters.noop.NoOpDataPlaneClient method) (nemo_rl.data_plane.adapters.transfer_queue.TQDataPlaneClient method) (nemo_rl.data_plane.interfaces.DataPlaneClient method) (nemo_rl.data_plane.observability.MetricsDataPlaneClient method) (nemo_rl.models.automodel.checkpoint.AutomodelCheckpointManager method) (nemo_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) (nemo_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl method) (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) (nemo_rl.models.value.workers.dtensor_value_worker_v2.DTensorValueWorkerV2Impl method) (nemo_rl.models.value.workers.megatron_value_worker.MegatronValueWorkerImpl method) load_config() (in module nemo_rl.utils.config) load_config_with_inheritance() (in module nemo_rl.utils.config) load_data_plane_checkpoint() (nemo_rl.models.policy.tq_policy.TQPolicy method) load_dataloader_state() (in module nemo_rl.data.utils) load_dataset_from_path() (in module nemo_rl.data.datasets.utils) load_eval_dataset() (in module nemo_rl.data.datasets.eval_datasets) load_format (nemo_rl.models.generation.dynamo.config.DynamoVllmConfig attribute) (nemo_rl.models.generation.vllm.config.VllmSpecificArgs attribute) load_from_path() (nemo_rl.algorithms.async_utils.interfaces.ReplayBufferProtocol method) (nemo_rl.algorithms.async_utils.replay_buffer.ReplayBufferImpl method) load_hf_weights_to_eagle() (in module nemo_rl.models.megatron.draft.utils) load_media_from_message() (in module nemo_rl.data.multimodal_utils) load_model() (nemo_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) load_mtp_weights_from_disk() (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) load_nemotron_video_model_config() (in module nemo_rl.environments.nemotron_utils) load_preference_dataset() (in module nemo_rl.data.datasets.preference_datasets) load_replay_buffer (nemo_rl.utils.checkpoint.CheckpointingConfig attribute) load_response_dataset() (in module nemo_rl.data.datasets.response_datasets) load_state_dict() (nemo_rl.algorithms.async_utils.interfaces.ReplayBufferProtocol method) (nemo_rl.algorithms.async_utils.replay_buffer.ReplayBufferImpl method) (nemo_rl.algorithms.async_utils.replay_buffer.TQReplayBuffer method) (nemo_rl.experience.rollout_recovery.RolloutRecoveryLedger method) (nemo_rl.utils.native_checkpoint.ModelState method) (nemo_rl.utils.native_checkpoint.OptimizerState method) load_training_info() (nemo_rl.utils.checkpoint.CheckpointManager method) load_video_frames_with_metadata() (in module nemo_rl.models.generation.vllm.video_utils) local_buffer_size (nemo_rl.data_plane.interfaces.MooncakeCpuConfig attribute) local_expert_ids (nemo_rl.models.generation.vllm.refit_layout.VllmExpertParamLayout attribute) localize_alignment() (in module nemo_rl.algorithms.x_token.loss_utils) LocalizedAlignment (class in nemo_rl.algorithms.x_token.loss_utils) LocalMathDataset (class in nemo_rl.data.datasets.eval_datasets.local_math_dataset) LocalMathEvalDataConfig (class in nemo_rl.data) LocalParamSpec (class in nemo_rl.weight_sync.nccl_reshard_utils) log (in module nemo_rl.algorithms.single_controller) (in module nemo_rl.models.policy.workers.megatron_policy_worker) log_batched_dict_as_jsonl() (nemo_rl.utils.logger.Logger method) log_container_init_timing() (in module nemo_rl.utils.logger) log_dir (nemo_rl.utils.logger.LoggerConfig attribute) (nemo_rl.utils.logger.TensorboardConfig attribute) log_event() (in module nemo_rl.data_plane.observability) log_generation_metrics() (in module nemo_rl.algorithms.utils) log_gpu_memory() (in module nemo_rl.models.generation.megatron.utils) log_gpu_memory_diagnostics() (in module nemo_rl.utils.nvml) log_histogram() (nemo_rl.utils.logger.Logger method) (nemo_rl.utils.logger.LoggerInterface method) (nemo_rl.utils.logger.MLflowLogger method) (nemo_rl.utils.logger.SwanlabLogger method) (nemo_rl.utils.logger.TensorboardLogger method) (nemo_rl.utils.logger.WandbLogger method) log_hyperparams() (nemo_rl.utils.logger.Logger method) (nemo_rl.utils.logger.LoggerInterface method) (nemo_rl.utils.logger.MLflowLogger method) (nemo_rl.utils.logger.SwanlabLogger method) (nemo_rl.utils.logger.TensorboardLogger method) (nemo_rl.utils.logger.WandbLogger method) log_metrics() (nemo_rl.utils.logger.Logger method) (nemo_rl.utils.logger.LoggerInterface method) (nemo_rl.utils.logger.MLflowLogger method) (nemo_rl.utils.logger.SwanlabLogger method) (nemo_rl.utils.logger.TensorboardLogger method) (nemo_rl.utils.logger.WandbLogger method) log_nemo_gym_full_result_tables (nemo_rl.utils.logger.WandbConfig attribute) log_plot() (nemo_rl.utils.logger.Logger method) (nemo_rl.utils.logger.LoggerInterface method) (nemo_rl.utils.logger.MLflowLogger method) (nemo_rl.utils.logger.SwanlabLogger method) (nemo_rl.utils.logger.TensorboardLogger method) (nemo_rl.utils.logger.WandbLogger method) log_plot_per_worker_timeline_metrics() (nemo_rl.utils.logger.Logger method) log_plot_token_mult_prob_error() (nemo_rl.utils.logger.Logger method) log_string_list_as_jsonl() (nemo_rl.utils.logger.Logger method) Logger (class in nemo_rl.utils.logger) logger (in module nemo_rl.data.datasets.response_datasets.intent) (in module nemo_rl.data.multimodal_utils) (in module nemo_rl.data_plane.observability) (in module nemo_rl.distributed.numa_utils) (in module nemo_rl.distributed.virtual_cluster) LOGGER (in module nemo_rl.models.generation.dynamo.dynamo_generation) (in module nemo_rl.models.generation.dynamo.managed_runtime) (in module nemo_rl.models.generation.dynamo.metrics) logger (in module nemo_rl.models.generation.trtllm.trtllm_http_server) (in module nemo_rl.models.generation.vllm.vllm_backend) (in module nemo_rl.models.generation.vllm.vllm_generation) (in module nemo_rl.models.generation.vllm.vllm_sparse_refit) (in module nemo_rl.models.generation.vllm.vllm_worker) LOGGER (in module nemo_rl.models.generation.vllm.vllm_worker_async) logger (in module nemo_rl.telemetry.metrics) (in module nemo_rl.telemetry.setup) (in module nemo_rl.utils.fastokens) (in module nemo_rl.utils.nvml) (in module nemo_rl.utils.timer) (in module nemo_rl.utils.venvs) (nemo_rl.algorithms.distillation.MasterConfig attribute) (nemo_rl.algorithms.dpo.MasterConfig attribute) (nemo_rl.algorithms.grpo.MasterConfig attribute) (nemo_rl.algorithms.ppo.MasterConfig attribute) (nemo_rl.algorithms.rm.MasterConfig attribute) (nemo_rl.algorithms.sft.MasterConfig attribute) (nemo_rl.algorithms.single_controller_utils.config.MasterConfig attribute) (nemo_rl.algorithms.xtoken_off_policy_distillation.MasterConfig attribute) LoggerConfig (class in nemo_rl.utils.logger) LoggerInterface (class in nemo_rl.utils.logger) logging_step_interval (nemo_rl.models.generation.megatron.config.MCoreGenerationSpecificArgs attribute) logical_segment_counts_by_row() (nemo_rl.data.multimodal_utils.PackedTensor method) LOGIT (nemo_rl.algorithms.loss.interfaces.LossInputType attribute) LOGPROB (nemo_rl.algorithms.loss.interfaces.LossInputType attribute) (nemo_rl.telemetry.span_groups.RLSpanGroup attribute) logprob_batch_size (nemo_rl.environments.reward_model_environment.RewardModelEnvironmentConfig attribute) (nemo_rl.models.policy.PolicyConfig attribute) (nemo_rl.models.value.config.ValueConfig attribute) logprob_chunk_size (nemo_rl.models.policy.PolicyConfig attribute) logprob_mb_tokens (nemo_rl.models.policy.DynamicBatchingConfig attribute) (nemo_rl.models.policy.SequencePackingConfig attribute) LogprobOutputSpec (class in nemo_rl.models.policy.interfaces) logprobs (nemo_rl.models.generation.interfaces.GenerationOutputSpec attribute) (nemo_rl.models.policy.interfaces.LogprobOutputSpec attribute) logprobs_mode (nemo_rl.models.generation.megatron.config.MCoreGenerationSpecificArgs attribute) (nemo_rl.models.generation.vllm.config.VllmSpecificArgs attribute) LogprobsPostProcessor (class in nemo_rl.models.automodel.train) (class in nemo_rl.models.megatron.train) logs_enabled (nemo_rl.telemetry.config.TelemetryConfig attribute) lora_A_init (nemo_rl.models.policy.LoRAConfig attribute) lora_A_init_method (nemo_rl.models.policy.MegatronPeftConfig attribute) lora_B_init_method (nemo_rl.models.policy.MegatronPeftConfig attribute) lora_cfg (nemo_rl.models.policy.DTensorConfig attribute) lora_dtype (nemo_rl.models.policy.MegatronPeftConfig attribute) LoRAConfig (class in nemo_rl.models.policy) LoRAConfigDisabled (class in nemo_rl.models.policy) loss (nemo_rl.algorithms.dpo.DPOValMetrics attribute) (nemo_rl.algorithms.rm.RMValMetrics attribute) loss_fn (nemo_rl.algorithms.distillation.MasterConfig attribute) (nemo_rl.algorithms.grpo.MasterConfig attribute) (nemo_rl.algorithms.ppo.MasterConfig attribute) (nemo_rl.algorithms.single_controller_utils.config.MasterConfig attribute) (nemo_rl.algorithms.single_controller_utils.setup.SingleControllerActorArgs attribute) (nemo_rl.algorithms.xtoken_off_policy_distillation.MasterConfig attribute) loss_multiplier (nemo_rl.data.interfaces.DatumSpec attribute) (nemo_rl.data.interfaces.PreferenceDatumSpec attribute) loss_type (nemo_rl.algorithms.loss.interfaces.LossFunction attribute) (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossFn attribute) (nemo_rl.algorithms.loss.loss_functions.DistillationLossFn attribute) (nemo_rl.algorithms.loss.loss_functions.DPOLossFn attribute) (nemo_rl.algorithms.loss.loss_functions.DraftCrossEntropyLossFn attribute) (nemo_rl.algorithms.loss.loss_functions.NLLLossFn attribute) (nemo_rl.algorithms.loss.loss_functions.PreferenceLossFn attribute) loss_weight (nemo_rl.models.policy.DraftConfig attribute) LossFunction (class in nemo_rl.algorithms.loss.interfaces) LossInputType (class in nemo_rl.algorithms.loss.interfaces) LossPostProcessor (class in nemo_rl.models.automodel.train) (class in nemo_rl.models.megatron.train) LossType (class in nemo_rl.algorithms.loss.interfaces) low_lengths (nemo_rl.experience.rollouts._EffortShapingMetrics attribute) low_penalty (nemo_rl.experience.rollouts.EffortLevelsConfig attribute) low_string (nemo_rl.experience.rollouts.EffortLevelsConfig attribute) low_ub (nemo_rl.experience.rollouts.EffortLevelsConfig attribute) low_weight (nemo_rl.experience.rollouts.EffortLevelsConfig attribute) LP_SEED_FIELDS (in module nemo_rl.data_plane.schema) lr (nemo_rl.models.policy.MegatronOptimizerConfig attribute) lr_decay_iters (nemo_rl.models.policy.MegatronSchedulerConfig attribute) lr_decay_style (nemo_rl.models.policy.MegatronSchedulerConfig attribute) lr_warmup_init (nemo_rl.models.policy.MegatronSchedulerConfig attribute) lr_warmup_iters (nemo_rl.models.policy.MegatronSchedulerConfig attribute) M main() (in module nemo_rl.models.generation.dynamo.validate_dynamo_vllm_args) MAJOR (in module nemo_rl.package_info) make_actor_runtime_env() (in module nemo_rl.utils.venvs) make_microbatch_iterator() (nemo_rl.distributed.batched_data_dict.BatchedDataDict method) make_microbatch_iterator_for_packable_sequences() (nemo_rl.distributed.batched_data_dict.BatchedDataDict method) make_microbatch_iterator_with_dynamic_shapes() (nemo_rl.distributed.batched_data_dict.BatchedDataDict method) make_nccl_reshard_refit_info_wire_safe() (in module nemo_rl.weight_sync.nccl_reshard_utils) make_policy_like_config() (in module nemo_rl.models.megatron.setup) make_processed_microbatch_iterator() (in module nemo_rl.models.automodel.data) (in module nemo_rl.models.megatron.data) make_sequence_length_divisible_by (nemo_rl.models.policy.PolicyConfig attribute) (nemo_rl.models.value.config.ValueConfig attribute) make_value_head_hook() (in module nemo_rl.models.value.workers.megatron_value_worker) malformed_thinking_advantage (nemo_rl.algorithms.grpo.GRPOConfig attribute) mamba_head_dim (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) mamba_inference_conv_states_dtype (nemo_rl.models.generation.megatron.config.MCoreGenerationSpecificArgs attribute) mamba_inference_ssm_states_dtype (nemo_rl.models.generation.megatron.config.MCoreGenerationSpecificArgs attribute) mamba_num_groups (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) mamba_num_heads (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) mamba_state_dim (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) managed_span() (in module nemo_rl.telemetry.instrumentation) ManagedDynamoRuntime (class in nemo_rl.models.generation.dynamo.managed_runtime) manifest_digest (nemo_rl.algorithms.async_utils.replay_buffer.TQReplayMetadataState attribute) map() (nemo_rl.data.datasets.response_datasets.oai_format_dataset.PreservingDataset method) map_efficiency_seconds() (in module nemo_rl.telemetry.metrics) mark() (nemo_rl.utils.timer.ThreadSafeTimer method) (nemo_rl.utils.timer.Timer method) mark_group_admitted() (nemo_rl.experience.rollout_recovery.RolloutRecoveryLedger method) mark_iteration() (nemo_rl.utils.timer.TimeoutChecker method) mark_loaded() (nemo_rl.models.generation.fleet_health.GenerationFleetHealth method) mark_prompt_group_admitted() (nemo_rl.experience.rollout_manager.RolloutManager method) mark_restarting() (nemo_rl.models.generation.fleet_health.GenerationFleetHealth method) mark_weights_partial() (nemo_rl.models.generation.fleet_health.GenerationFleetHealth method) mask_out_neg_inf_logprobs() (in module nemo_rl.algorithms.utils) masked_mean() (in module nemo_rl.algorithms.utils) masked_var() (in module nemo_rl.algorithms.utils) master_port_range_high (nemo_rl.distributed.virtual_cluster.ClusterConfig attribute) master_port_range_low (nemo_rl.distributed.virtual_cluster.ClusterConfig attribute) MasterConfig (class in nemo_rl.algorithms.distillation) (class in nemo_rl.algorithms.dpo) (class in nemo_rl.algorithms.grpo) (class in nemo_rl.algorithms.ppo) (class in nemo_rl.algorithms.rm) (class in nemo_rl.algorithms.sft) (class in nemo_rl.algorithms.single_controller_utils.config) (class in nemo_rl.algorithms.xtoken_off_policy_distillation) (class in nemo_rl.evals.eval) match_all_linear (nemo_rl.models.policy.LoRAConfig attribute) matches() (nemo_rl.models.huggingface.common.ModelFlag method) matches_quant_ignore_pattern() (in module nemo_rl.modelopt.utils) materialize() (in module nemo_rl.data_plane.codec) materialize_only_last_token_logits (nemo_rl.models.generation.megatron.config.MCoreGenerationSpecificArgs attribute) materialize_vllm_video_config() (in module nemo_rl.models.generation.vllm.config) math (nemo_rl.evals.eval._PassThroughEnvConfig attribute) math_data_processor() (in module nemo_rl.data.processors) math_expression_reward() (in module nemo_rl.environments.rewards) math_hf_data_processor() (in module nemo_rl.data.processors) math_verify_func (in module nemo_rl.environments.rewards) math_verify_impl (nemo_rl.environments.math_environment.MathEnvConfig attribute) MathDataset (class in nemo_rl.data.datasets.eval_datasets.math) MathEnvConfig (class in nemo_rl.environments.math_environment) MathEnvironment (class in nemo_rl.environments.math_environment) MathEnvironmentMetadata (class in nemo_rl.environments.math_environment) MathEvalDataConfig (class in nemo_rl.data) MathMultiRewardEnvironment (class in nemo_rl.environments.math_environment) max_backoff_s (nemo_rl.algorithms.single_controller_utils.config.RolloutFailureConfig attribute) (nemo_rl.experience.rollout_manager.RolloutRetryPolicy attribute) max_batch_size (nemo_rl.models.generation.trtllm.config.TrtllmSpecificArgs attribute) max_buffered_rollouts (nemo_rl.algorithms.single_controller_utils.config.AsyncRLConfig attribute) max_bytes_per_key_seen (nemo_rl.data_plane.observability.DataPlaneStats attribute) max_consecutive_dropped_prompts (nemo_rl.algorithms.single_controller_utils.config.RolloutFailureConfig attribute) (nemo_rl.experience.rollout_manager.RolloutRetryPolicy attribute) max_consecutive_infra_drops (nemo_rl.experience.rollout_manager.RolloutStats attribute) max_data_attempts (nemo_rl.experience.rollout_manager.RolloutRetryPolicy attribute) max_data_attempts_per_prompt (nemo_rl.algorithms.single_controller_utils.config.RolloutFailureConfig attribute) max_generation_failures (nemo_rl.algorithms.grpo.AsyncGRPOConfig attribute) max_grad_norm (nemo_rl.environments.reward_model_environment.RewardModelEnvironmentConfig attribute) (nemo_rl.models.automodel.config.RuntimeConfig attribute) (nemo_rl.models.policy.PolicyConfig attribute) (nemo_rl.models.value.config.ValueConfig attribute) max_gym_row_attempts (nemo_rl.experience.rollout_manager.RolloutRetryPolicy attribute) max_inflight_prompts (nemo_rl.algorithms.single_controller_utils.config.AsyncRLConfig attribute) max_infra_attempts (nemo_rl.experience.rollout_manager.RolloutRetryPolicy attribute) max_infra_attempts_per_prompt (nemo_rl.algorithms.single_controller_utils.config.RolloutFailureConfig attribute) max_input_seq_length (nemo_rl.data.AIMEEvalDataConfig attribute) (nemo_rl.data.DailyOmniEvalDataConfig attribute) (nemo_rl.data.DataConfig attribute) (nemo_rl.data.GPQAEvalDataConfig attribute) (nemo_rl.data.LocalMathEvalDataConfig attribute) (nemo_rl.data.MathEvalDataConfig attribute) (nemo_rl.data.MMAUEvalDataConfig attribute) (nemo_rl.data.MMLUEvalDataConfig attribute) (nemo_rl.data.MMLUProEvalDataConfig attribute) max_lookahead_versions (nemo_rl.algorithms.async_utils.staleness_sampler.InOrderSamplerConfig attribute) max_model_len (nemo_rl.models.generation.dynamo.config.DynamoVllmConfig attribute) (nemo_rl.models.generation.megatron.config.MCoreGenerationSpecificArgs attribute) (nemo_rl.models.generation.trtllm.config.TrtllmSpecificArgs attribute) (nemo_rl.models.generation.vllm.config.VllmSpecificArgs attribute) max_new_tokens (nemo_rl.models.generation.interfaces.GenerationConfig attribute) max_num_epochs (nemo_rl.algorithms.distillation.DistillationConfig attribute) (nemo_rl.algorithms.dpo.DPOConfig attribute) (nemo_rl.algorithms.grpo.GRPOConfig attribute) (nemo_rl.algorithms.ppo.PPOConfig attribute) (nemo_rl.algorithms.rm.RMConfig attribute) (nemo_rl.algorithms.sft.SFTConfig attribute) (nemo_rl.algorithms.xtoken_off_policy_distillation.OffPolicyDistillationConfig attribute) max_num_steps (nemo_rl.algorithms.distillation.DistillationConfig attribute) (nemo_rl.algorithms.dpo.DPOConfig attribute) (nemo_rl.algorithms.grpo.GRPOConfig attribute) (nemo_rl.algorithms.ppo.PPOConfig attribute) (nemo_rl.algorithms.rm.RMConfig attribute) (nemo_rl.algorithms.sft.SFTConfig attribute) (nemo_rl.algorithms.xtoken_off_policy_distillation.OffPolicyDistillationConfig attribute) max_num_tokens (nemo_rl.models.generation.trtllm.config.TrtllmSpecificArgs attribute) MAX_OUTPUT_LEN (in module nemo_rl.modelopt.models.policy.workers.utils) max_replacement_attempts (nemo_rl.algorithms.single_controller_utils.config.RolloutFailureConfig attribute) max_response_length (nemo_rl.algorithms.reward_functions.RewardShapingConfig attribute) max_restart_attempts_per_shard (nemo_rl.algorithms.single_controller_utils.config.FleetHealthConfig attribute) (nemo_rl.models.generation.fleet_health.FleetHealthPolicy attribute) max_rollout_turns (nemo_rl.algorithms.distillation.DistillationConfig attribute) (nemo_rl.algorithms.grpo.GRPOConfig attribute) (nemo_rl.algorithms.ppo.PPOConfig attribute) max_row_attempts (nemo_rl.algorithms.single_controller_utils.config.NemoGymRolloutFTConfig attribute) MAX_SEQ_LEN (in module nemo_rl.modelopt.models.policy.workers.utils) max_seqlen_k (nemo_rl.models.huggingface.common.FlashAttentionKwargs attribute) max_seqlen_q (nemo_rl.models.huggingface.common.FlashAttentionKwargs attribute) max_sequences_per_bin (nemo_rl.data.packing.algorithms.ConcatenativePacker attribute) max_skipped_prompts (nemo_rl.algorithms.single_controller_utils.config.RolloutFailureConfig attribute) (nemo_rl.experience.rollout_manager.RolloutRetryPolicy attribute) max_staleness_versions (nemo_rl.algorithms.async_utils.staleness_sampler.ReadyFirstSamplerConfig attribute) (nemo_rl.algorithms.async_utils.staleness_sampler.WeightFifoSamplerConfig attribute) (nemo_rl.algorithms.async_utils.staleness_sampler.WindowedSamplerConfig attribute) max_tokens (nemo_rl.models.generation.megatron.config.MCoreGenerationSpecificArgs attribute) max_tokens_per_microbatch (nemo_rl.distributed.batched_data_dict.DynamicBatchingArgs attribute) (nemo_rl.distributed.batched_data_dict.SequencePackingArgs attribute) max_total_sequence_length (nemo_rl.models.policy.PolicyConfig attribute) (nemo_rl.models.value.config.ValueConfig attribute) max_trajectory_age_steps (nemo_rl.algorithms.grpo.AsyncGRPOConfig attribute) (nemo_rl.algorithms.ppo.AsyncPPOConfig attribute) max_val_samples (nemo_rl.algorithms.distillation.DistillationConfig attribute) (nemo_rl.algorithms.grpo.GRPOConfig attribute) (nemo_rl.algorithms.ppo.PPOConfig attribute) maybe_configure_data_plane_env() (in module nemo_rl.data_plane.factory) maybe_gpu_profile_step() (in module nemo_rl.utils.nsys) maybe_init_zmq() (nemo_rl.modelopt.models.generation.vllm_quant_backend.VllmQuantInternalWorkerExtension method) (nemo_rl.modelopt.models.policy.workers.megatron_quant_policy_worker.MegatronQuantPolicyWorker method) (nemo_rl.models.generation.trtllm.trtllm_backend.NcclExtension method) (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) (nemo_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker method) maybe_pad_last_batch() (in module nemo_rl.algorithms.utils) maybe_patch_fastokens() (in module nemo_rl.utils.fastokens) maybe_preinit_nixl_checkpoint_engine() (in module nemo_rl.models.policy.workers.checkpoint_engine) maybe_r3_trace_stage() (in module nemo_rl.utils.r3_trace) MCORE (nemo_rl.distributed.virtual_cluster.PY_EXECUTABLES attribute) MCORE_EXECUTABLE (in module nemo_rl.distributed.ray_actor_environment_registry) mcore_generation_config (nemo_rl.models.generation.megatron.config.MCoreGenerationConfig attribute) MCoreGenerationConfig (class in nemo_rl.models.generation.megatron.config) MCoreGenerationSpecificArgs (class in nemo_rl.models.generation.megatron.config) media_placeholder_token_id_from_chunks() (in module nemo_rl.data.multimodal_utils) MEDIA_TAG_PATTERN (in module nemo_rl.data.multimodal_utils) MEDIA_TAGS (in module nemo_rl.data.multimodal_utils) MEDIA_TAGS_REVERSED (in module nemo_rl.data.multimodal_utils) MEDIA_TAGS_TO_ALLOWED (in module nemo_rl.data.multimodal_utils) media_token_validity_mask (nemo_rl.models.megatron.data.ProcessedInputs attribute) (nemo_rl.models.megatron.data.ProcessedMicrobatch attribute) MEGATRON_BACKEND (in module nemo_rl.models.generation.constants) megatron_cfg (nemo_rl.models.megatron.config.RuntimeConfig attribute) (nemo_rl.models.policy.PolicyConfig attribute) (nemo_rl.models.value.config.ValueConfig attribute) megatron_cfg_overrides (nemo_rl.algorithms.opd.TeacherResourceConfig attribute) (nemo_rl.models.policy.teacher_worker_group.TeacherConfig attribute) megatron_conversion_is_complete() (in module nemo_rl.models.megatron.community_import) megatron_forward_backward() (in module nemo_rl.models.megatron.train) megatron_path (in module nemo_rl) MegatronCheckpointConfig (class in nemo_rl.models.policy) MegatronCheckpointEngineSendMixin (class in nemo_rl.models.policy.workers.checkpoint_engine) MegatronConfig (class in nemo_rl.models.policy) MegatronConfigDisabled (class in nemo_rl.models.policy) MegatronDDPConfig (class in nemo_rl.models.policy) MegatronGeneration (class in nemo_rl.models.generation.megatron.megatron_generation) MegatronGenerationMixin (class in nemo_rl.models.generation.megatron.megatron_worker) MegatronGenerationRefitMixin (class in nemo_rl.models.generation.megatron.megatron_worker) MegatronOptimizerConfig (class in nemo_rl.models.policy) MegatronPeftConfig (class in nemo_rl.models.policy) MegatronPeftConfigDisabled (class in nemo_rl.models.policy) MegatronPolicyWorker (class in nemo_rl.models.policy.workers.megatron_policy_worker) MegatronPolicyWorkerImpl (class in nemo_rl.models.policy.workers.megatron_policy_worker) MegatronQuantPolicyWorker (class in nemo_rl.modelopt.models.policy.workers.megatron_quant_policy_worker) MegatronRemoteSparseRefit (class in nemo_rl.models.policy.workers.megatron_remote_sparse_refit) MegatronSchedulerConfig (class in nemo_rl.models.policy) MegatronValueWorker (class in nemo_rl.models.value.workers.megatron_value_worker) MegatronValueWorkerImpl (class in nemo_rl.models.value.workers.megatron_value_worker) MegatronWeightSynchronizer (class in nemo_rl.weight_sync.megatron_weight_synchronizer) mem_used_diff_gb (nemo_rl.utils.memory_tracker.MemoryTrackerDataPoint property) membership_epoch (nemo_rl.models.generation.fleet_health.GenerationFleetHealth property) memory_used_after_stage_gb (nemo_rl.utils.memory_tracker.MemoryTrackerDataPoint attribute) memory_used_before_stage_gb (nemo_rl.utils.memory_tracker.MemoryTrackerDataPoint attribute) MemoryTracker (class in nemo_rl.utils.memory_tracker) MemoryTrackerDataPoint (class in nemo_rl.utils.memory_tracker) merge_datasets() (in module nemo_rl.data.datasets.utils) merge_multimodal_payload_metrics() (in module nemo_rl.utils.multimodal_payload_metrics) merge_segments() (nemo_rl.data.multimodal_utils.PackedTensor class method) merge_sparse_payloads() (in module nemo_rl.utils.weight_transfer_sparse_codec) merge_vllm_refit_metrics() (in module nemo_rl.utils.weight_transfer_http) merge_weight_chunk_batches() (in module nemo_rl.utils.checkpoint_engines.base) merge_with_override() (in module nemo_rl.utils.config) merged_inference_megatron_cfg() (in module nemo_rl.models.generation.megatron.config) MeshInfo (class in nemo_rl.weight_sync.nccl_reshard_utils) message_log (nemo_rl.data.interfaces.DatumSpec attribute) (nemo_rl.experience.interfaces.Completion attribute) MESSAGE_LOG_BULK_FIELDS (in module nemo_rl.data.llm_message_utils) message_log_chosen (nemo_rl.data.interfaces.PreferenceDatumSpec attribute) message_log_rejected (nemo_rl.data.interfaces.PreferenceDatumSpec attribute) message_log_shape() (in module nemo_rl.data.llm_message_utils) message_log_to_flat_messages() (in module nemo_rl.data.llm_message_utils) meta (nemo_rl.algorithms.async_utils.replay_buffer.TQReplayGroupMetadata attribute) META_IDX (in module nemo_rl.data_plane.schema) metadata (nemo_rl.environments.interfaces.EnvironmentReturn attribute) (nemo_rl.experience.interfaces.PromptGroupRecord attribute) metadata() (nemo_rl.models.generation.dynamo.dynamo_worker.DynamoGpuReservation method) (nemo_rl.models.generation.dynamo.dynamo_worker.DynamoVllmWorker method) metadata_state_dict() (nemo_rl.algorithms.async_utils.replay_buffer.TQReplayBuffer method) MetadataT (in module nemo_rl.environments.interfaces) metric (nemo_rl.evals.eval.EvalConfig attribute) metric_name (nemo_rl.utils.checkpoint.CheckpointingConfig attribute) MetricNormalizer (class in nemo_rl.algorithms.loss.interfaces) metrics (nemo_rl.utils.logger.GpuMetricSnapshot attribute) metrics() (nemo_rl.models.generation.generation_router.GenerationRouterImpl method) metrics_enabled (nemo_rl.telemetry.config.TelemetryConfig attribute) metrics_exclude_prefixes (nemo_rl.models.generation.dynamo.config.DynamoCfg attribute) metrics_include_prefixes (nemo_rl.models.generation.dynamo.config.DynamoCfg attribute) MetricsDataPlaneClient (class in nemo_rl.data_plane.observability) MICRO_BATCH_INDICES (in module nemo_rl.data_plane.schema) MICRO_BATCH_LENGTHS (in module nemo_rl.data_plane.schema) micro_batch_size (nemo_rl.algorithms.opd.TeacherResourceConfig attribute) (nemo_rl.models.policy.teacher_worker_group.TeacherConfig attribute) microbatch_order (nemo_rl.distributed.batched_data_dict.SequencePackingArgs attribute) (nemo_rl.models.policy.SequencePackingConfig attribute) milestones (nemo_rl.models.policy.SinglePytorchMilestonesConfig attribute) min_generation_tokens (nemo_rl.data.interfaces.TaskDataSpec attribute) (nemo_rl.data.PreferenceDatasetConfig attribute) (nemo_rl.data.ResponseDatasetConfig attribute) min_groups_for_streaming_train (nemo_rl.algorithms.single_controller_utils.config.AsyncRLConfig attribute) min_healthy_shards (nemo_rl.algorithms.single_controller_utils.config.FleetHealthConfig attribute) (nemo_rl.models.generation.fleet_health.FleetHealthPolicy attribute) min_lr (nemo_rl.models.policy.MegatronOptimizerConfig attribute) min_step_batch_fraction (nemo_rl.algorithms.single_controller_utils.config.RolloutFailureConfig attribute) MINOR (in module nemo_rl.package_info) minus_baseline (nemo_rl.algorithms.advantage_estimator.AdvEstimatorConfig attribute) missing_weight_prefixes (nemo_rl.models.generation.vllm.refit_layout.VllmWeightLayout attribute) mixed_kl_weight (nemo_rl.algorithms.loss.loss_functions.DistillationLossConfig attribute) mixtral() (in module nemo_rl.utils.flops_formulas) mlflow (nemo_rl.utils.logger.LoggerConfig attribute) mlflow_enabled (nemo_rl.utils.logger.LoggerConfig attribute) MLflowConfig (class in nemo_rl.utils.logger) MLflowLogger (class in nemo_rl.utils.logger) mmap_dir (nemo_rl.models.generation.vllm.config.VllmRefitBaselineConfig attribute) mmau (nemo_rl.evals.eval._PassThroughEnvConfig attribute) MMAUDataset (class in nemo_rl.data.datasets.eval_datasets.mmau) MMAUEvalDataConfig (class in nemo_rl.data) MMLUDataset (class in nemo_rl.data.datasets.eval_datasets.mmlu) MMLUEvalDataConfig (class in nemo_rl.data) MMLUProDataset (class in nemo_rl.data.datasets.eval_datasets.mmlu_pro) MMLUProEvalDataConfig (class in nemo_rl.data) MMPRTinyDataset (class in nemo_rl.data.datasets.response_datasets.mmpr_tiny) mode (nemo_rl.algorithms.async_utils.replay_buffer.DataPlaneCheckpointMetadata attribute) model (nemo_rl.models.automodel.config.ModelAndOptimizerState attribute) (nemo_rl.models.megatron.config.ModelAndOptimizerState attribute) (nemo_rl.models.policy.workers.checkpoint_engine.DTensorCheckpointEngineSendMixin attribute) model_batch (nemo_rl.models.automodel.train.PreparedModelForward attribute) model_cache_dir (nemo_rl.utils.checkpoint.CheckpointingConfig attribute) model_cfg (nemo_rl.models.megatron.config.RuntimeConfig attribute) model_channels (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) model_class (nemo_rl.models.automodel.config.ModelAndOptimizerState attribute) (nemo_rl.models.automodel.config.RuntimeConfig attribute) model_config (nemo_rl.models.automodel.config.ModelAndOptimizerState attribute) (nemo_rl.models.automodel.config.RuntimeConfig attribute) (nemo_rl.models.generation.vllm.config.VllmVideoConfig attribute) model_context_factory (nemo_rl.models.automodel.train.PreparedModelForward attribute) model_dump_chat_response_with_dynamic_message_fields() (in module nemo_rl.models.generation.vllm.utils) model_forward() (in module nemo_rl.models.automodel.train) (in module nemo_rl.models.megatron.train) model_name (nemo_rl.environments.nemo_gym.NemoGymConfig attribute) (nemo_rl.environments.reward_model_environment.RewardModelEnvironmentConfig attribute) (nemo_rl.models.generation.interfaces.GenerationConfig attribute) (nemo_rl.models.generation.trtllm.config.TrtllmSpecificArgs attribute) (nemo_rl.models.policy.DraftConfig attribute) (nemo_rl.models.policy.PolicyConfig attribute) (nemo_rl.models.policy.teacher_worker_group.TeacherConfig attribute) (nemo_rl.models.value.config.ValueConfig attribute) model_overrides (nemo_rl.models.policy.MegatronConfig attribute) model_pattern (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) model_post_init() (nemo_rl.utils.memory_tracker.MemoryTracker method) model_prefix (nemo_rl.models.megatron.draft.utils._EagleLayerLayout attribute) model_repo_id (nemo_rl.utils.checkpoint.CheckpointingConfig attribute) model_save_format (nemo_rl.utils.checkpoint.CheckpointingConfig attribute) model_update_group (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension attribute) (nemo_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker attribute) ModelAndOptimizerState (class in nemo_rl.models.automodel.config) (class in nemo_rl.models.megatron.config) ModelFlag (class in nemo_rl.models.huggingface.common) MODELOPT_ACTOR_REGISTRY (in module nemo_rl.modelopt.registry) MODELOPT_AUTOMODEL_EXECUTABLE (in module nemo_rl.modelopt.registry) MODELOPT_MCORE_EXECUTABLE (in module nemo_rl.modelopt.registry) MODELOPT_REAL_QUANT_ZMQ_TIMEOUT_MS (in module nemo_rl.modelopt.utils) MODELOPT_VLLM_EXECUTABLE (in module nemo_rl.modelopt.registry) ModelState (class in nemo_rl.utils.native_checkpoint) MODIFIED_FIRST_FIT_DECREASING (nemo_rl.data.packing.algorithms.PackingAlgorithm attribute) ModifiedFirstFitDecreasingPacker (class in nemo_rl.data.packing.algorithms) module nemo_rl nemo_rl.algorithms nemo_rl.algorithms.advantage_estimator nemo_rl.algorithms.async_utils nemo_rl.algorithms.async_utils.interfaces nemo_rl.algorithms.async_utils.replay_buffer nemo_rl.algorithms.async_utils.staleness_sampler nemo_rl.algorithms.async_utils.trajectory_collector nemo_rl.algorithms.distillation nemo_rl.algorithms.dpo nemo_rl.algorithms.grpo nemo_rl.algorithms.grpo_sync nemo_rl.algorithms.logits_sampling_utils nemo_rl.algorithms.loss nemo_rl.algorithms.loss.interfaces nemo_rl.algorithms.loss.loss_functions nemo_rl.algorithms.loss.utils nemo_rl.algorithms.loss.wrapper nemo_rl.algorithms.metric_utils nemo_rl.algorithms.opd nemo_rl.algorithms.ppo nemo_rl.algorithms.reward_functions nemo_rl.algorithms.rm nemo_rl.algorithms.sft nemo_rl.algorithms.single_controller nemo_rl.algorithms.single_controller_utils nemo_rl.algorithms.single_controller_utils.config nemo_rl.algorithms.single_controller_utils.setup nemo_rl.algorithms.single_controller_utils.utils nemo_rl.algorithms.utils nemo_rl.algorithms.x_token nemo_rl.algorithms.x_token.loss_utils nemo_rl.algorithms.x_token.token_aligner nemo_rl.algorithms.x_token.utils nemo_rl.algorithms.xtoken_off_policy_distillation nemo_rl.data nemo_rl.data.chat_templates nemo_rl.data.collate_fn nemo_rl.data.cross_tokenizer_collate nemo_rl.data.dataloader nemo_rl.data.datasets nemo_rl.data.datasets.eval_datasets nemo_rl.data.datasets.eval_datasets.daily_omni nemo_rl.data.datasets.eval_datasets.gpqa nemo_rl.data.datasets.eval_datasets.local_math_dataset nemo_rl.data.datasets.eval_datasets.math nemo_rl.data.datasets.eval_datasets.mmau nemo_rl.data.datasets.eval_datasets.mmlu nemo_rl.data.datasets.eval_datasets.mmlu_pro nemo_rl.data.datasets.preference_datasets nemo_rl.data.datasets.preference_datasets.binary_preference_dataset nemo_rl.data.datasets.preference_datasets.helpsteer3 nemo_rl.data.datasets.preference_datasets.preference_dataset nemo_rl.data.datasets.preference_datasets.tulu3 nemo_rl.data.datasets.processed_dataset nemo_rl.data.datasets.raw_dataset nemo_rl.data.datasets.response_datasets nemo_rl.data.datasets.response_datasets.aime nemo_rl.data.datasets.response_datasets.arrow_text_dataset nemo_rl.data.datasets.response_datasets.audiomcq nemo_rl.data.datasets.response_datasets.avqa nemo_rl.data.datasets.response_datasets.clevr nemo_rl.data.datasets.response_datasets.daily_omni nemo_rl.data.datasets.response_datasets.dapo_math nemo_rl.data.datasets.response_datasets.deepscaler nemo_rl.data.datasets.response_datasets.general_conversations_dataset nemo_rl.data.datasets.response_datasets.geometry3k nemo_rl.data.datasets.response_datasets.gsm8k nemo_rl.data.datasets.response_datasets.helpsteer3 nemo_rl.data.datasets.response_datasets.intent nemo_rl.data.datasets.response_datasets.mmpr_tiny nemo_rl.data.datasets.response_datasets.nemogym_dataset nemo_rl.data.datasets.response_datasets.nemotron_cascade2_sft nemo_rl.data.datasets.response_datasets.numinamath nemo_rl.data.datasets.response_datasets.oai_format_dataset nemo_rl.data.datasets.response_datasets.oasst nemo_rl.data.datasets.response_datasets.openmathinstruct2 nemo_rl.data.datasets.response_datasets.openr1_math nemo_rl.data.datasets.response_datasets.refcoco nemo_rl.data.datasets.response_datasets.response_dataset nemo_rl.data.datasets.response_datasets.squad nemo_rl.data.datasets.response_datasets.tulu3 nemo_rl.data.datasets.utils nemo_rl.data.interfaces nemo_rl.data.llm_message_utils nemo_rl.data.multimodal_utils nemo_rl.data.packing nemo_rl.data.packing.algorithms nemo_rl.data.packing.metrics nemo_rl.data.processors nemo_rl.data.utils nemo_rl.data_plane nemo_rl.data_plane.adapters nemo_rl.data_plane.adapters.noop nemo_rl.data_plane.adapters.transfer_queue nemo_rl.data_plane.adapters.transfer_queue_env nemo_rl.data_plane.async_utils nemo_rl.data_plane.codec nemo_rl.data_plane.column_io nemo_rl.data_plane.driver_mixin nemo_rl.data_plane.factory nemo_rl.data_plane.interfaces nemo_rl.data_plane.observability nemo_rl.data_plane.preshard nemo_rl.data_plane.schema nemo_rl.data_plane.worker_mixin nemo_rl.distributed nemo_rl.distributed.batched_data_dict nemo_rl.distributed.collectives nemo_rl.distributed.held_port nemo_rl.distributed.model_utils nemo_rl.distributed.named_sharding nemo_rl.distributed.numa_utils nemo_rl.distributed.ray_actor_environment_registry nemo_rl.distributed.refit_watchdog nemo_rl.distributed.stateless_process_group nemo_rl.distributed.virtual_cluster nemo_rl.distributed.worker_group_utils nemo_rl.distributed.worker_groups nemo_rl.environments nemo_rl.environments.code_environment nemo_rl.environments.code_jaccard_environment nemo_rl.environments.dapo_math_verifier nemo_rl.environments.interfaces nemo_rl.environments.math_environment nemo_rl.environments.metrics nemo_rl.environments.nemo_gym nemo_rl.environments.nemo_gym_video nemo_rl.environments.nemotron_utils nemo_rl.environments.reward_model_environment nemo_rl.environments.rewards nemo_rl.environments.utils nemo_rl.environments.vlm_environment nemo_rl.evals nemo_rl.evals.answer_parsing nemo_rl.evals.eval nemo_rl.experience nemo_rl.experience.failures nemo_rl.experience.interfaces nemo_rl.experience.metric_utils nemo_rl.experience.payload nemo_rl.experience.rollout_manager nemo_rl.experience.rollout_recovery nemo_rl.experience.rollouts nemo_rl.experience.sync_rollout_actor nemo_rl.modelopt nemo_rl.modelopt.models nemo_rl.modelopt.models.generation nemo_rl.modelopt.models.generation.vllm_modelopt nemo_rl.modelopt.models.generation.vllm_quant_backend nemo_rl.modelopt.models.generation.vllm_quant_patch nemo_rl.modelopt.models.generation.vllm_quant_worker nemo_rl.modelopt.models.policy nemo_rl.modelopt.models.policy.workers nemo_rl.modelopt.models.policy.workers.dtensor_quant_policy_worker nemo_rl.modelopt.models.policy.workers.dtensor_quant_policy_worker_v2 nemo_rl.modelopt.models.policy.workers.megatron_quant_policy_worker nemo_rl.modelopt.models.policy.workers.utils nemo_rl.modelopt.registry nemo_rl.modelopt.utils nemo_rl.models nemo_rl.models.automodel nemo_rl.models.automodel.checkpoint nemo_rl.models.automodel.config nemo_rl.models.automodel.data nemo_rl.models.automodel.setup nemo_rl.models.automodel.train nemo_rl.models.dtensor nemo_rl.models.dtensor.parallelize nemo_rl.models.generation nemo_rl.models.generation.constants nemo_rl.models.generation.dynamo nemo_rl.models.generation.dynamo.arguments nemo_rl.models.generation.dynamo.config nemo_rl.models.generation.dynamo.dynamo_generation nemo_rl.models.generation.dynamo.dynamo_worker nemo_rl.models.generation.dynamo.http_client nemo_rl.models.generation.dynamo.managed_runtime nemo_rl.models.generation.dynamo.metrics nemo_rl.models.generation.dynamo.refit nemo_rl.models.generation.dynamo.token_wrapper nemo_rl.models.generation.dynamo.validate_dynamo_vllm_args nemo_rl.models.generation.dynamo.venv nemo_rl.models.generation.dynamo.worker_pool nemo_rl.models.generation.fleet_health nemo_rl.models.generation.generation_router nemo_rl.models.generation.interfaces nemo_rl.models.generation.megatron nemo_rl.models.generation.megatron.config nemo_rl.models.generation.megatron.megatron_generation nemo_rl.models.generation.megatron.megatron_worker nemo_rl.models.generation.megatron.utils nemo_rl.models.generation.openai_server_utils nemo_rl.models.generation.trtllm nemo_rl.models.generation.trtllm.config nemo_rl.models.generation.trtllm.trtllm_backend nemo_rl.models.generation.trtllm.trtllm_generation nemo_rl.models.generation.trtllm.trtllm_http_server nemo_rl.models.generation.trtllm.trtllm_worker_async nemo_rl.models.generation.vllm nemo_rl.models.generation.vllm.checkpoint_engine nemo_rl.models.generation.vllm.collective_rpc nemo_rl.models.generation.vllm.config nemo_rl.models.generation.vllm.patches nemo_rl.models.generation.vllm.refit_layout nemo_rl.models.generation.vllm.refit_loader nemo_rl.models.generation.vllm.utils nemo_rl.models.generation.vllm.video_utils nemo_rl.models.generation.vllm.vllm_backend nemo_rl.models.generation.vllm.vllm_generation nemo_rl.models.generation.vllm.vllm_sparse_delta nemo_rl.models.generation.vllm.vllm_sparse_refit nemo_rl.models.generation.vllm.vllm_worker nemo_rl.models.generation.vllm.vllm_worker_async nemo_rl.models.generation.vllm.worker_utils nemo_rl.models.huggingface nemo_rl.models.huggingface.common nemo_rl.models.megatron nemo_rl.models.megatron.common nemo_rl.models.megatron.community_import nemo_rl.models.megatron.config nemo_rl.models.megatron.data nemo_rl.models.megatron.draft nemo_rl.models.megatron.draft.eagle nemo_rl.models.megatron.draft.hidden_capture nemo_rl.models.megatron.draft.utils nemo_rl.models.megatron.memory_saver nemo_rl.models.megatron.pipeline_parallel nemo_rl.models.megatron.router_replay nemo_rl.models.megatron.setup nemo_rl.models.megatron.train nemo_rl.models.policy nemo_rl.models.policy.interfaces nemo_rl.models.policy.lm_policy nemo_rl.models.policy.teacher_worker_group nemo_rl.models.policy.tq_policy nemo_rl.models.policy.utils nemo_rl.models.policy.workers nemo_rl.models.policy.workers.base_policy_worker nemo_rl.models.policy.workers.checkpoint_engine nemo_rl.models.policy.workers.dtensor_policy_worker nemo_rl.models.policy.workers.dtensor_policy_worker_v2 nemo_rl.models.policy.workers.megatron_policy_worker nemo_rl.models.policy.workers.megatron_remote_sparse_refit nemo_rl.models.policy.workers.patches nemo_rl.models.value nemo_rl.models.value.config nemo_rl.models.value.interfaces nemo_rl.models.value.lm_value nemo_rl.models.value.tq_value nemo_rl.models.value.workers nemo_rl.models.value.workers.dtensor_value_worker_v2 nemo_rl.models.value.workers.megatron_value_worker nemo_rl.package_info nemo_rl.telemetry nemo_rl.telemetry.config nemo_rl.telemetry.instrumentation nemo_rl.telemetry.metrics nemo_rl.telemetry.setup nemo_rl.telemetry.span_groups nemo_rl.utils nemo_rl.utils.checkpoint nemo_rl.utils.checkpoint_engines nemo_rl.utils.checkpoint_engines.base nemo_rl.utils.checkpoint_engines.nixl nemo_rl.utils.config nemo_rl.utils.fastokens nemo_rl.utils.flops_formulas nemo_rl.utils.flops_tracker nemo_rl.utils.grad_norm nemo_rl.utils.logger nemo_rl.utils.memory_tracker nemo_rl.utils.multimodal_payload_metrics nemo_rl.utils.native_checkpoint nemo_rl.utils.nsys nemo_rl.utils.nvml nemo_rl.utils.packed_tensor nemo_rl.utils.prefetch_venvs nemo_rl.utils.r3_trace nemo_rl.utils.routed_experts_codec nemo_rl.utils.timer nemo_rl.utils.venvs nemo_rl.utils.weight_transfer_http nemo_rl.utils.weight_transfer_sparse_codec nemo_rl.utils.weight_transfer_stream nemo_rl.utils.weight_transfer_zmq nemo_rl.weight_sync nemo_rl.weight_sync.checkpoint_engine_config nemo_rl.weight_sync.checkpoint_engine_weight_synchronizer nemo_rl.weight_sync.collective_weight_synchronizer nemo_rl.weight_sync.factory nemo_rl.weight_sync.interfaces nemo_rl.weight_sync.ipc_weight_synchronizer nemo_rl.weight_sync.megatron_weight_synchronizer nemo_rl.weight_sync.membership nemo_rl.weight_sync.nccl_reshard_utils nemo_rl.weight_sync.nccl_reshard_weight_synchronizer nemo_rl.weight_sync.sglang_weight_synchronizer nemo_rl.weight_sync.vllm_remote_sparse_weight_synchronizer nemo_rl.weight_sync.xferdtensor nemo_rl.weight_sync.xferdtensor_python moe_config (nemo_rl.models.automodel.config.DistributedContext attribute) moe_enable_deepep (nemo_rl.models.policy.MegatronConfig attribute) moe_expert_parallel_size (nemo_rl.models.generation.trtllm.config.TrtllmSpecificArgs attribute) moe_ffn_hidden_size (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) moe_flex_dispatcher_backend (nemo_rl.models.policy.MegatronConfig attribute) moe_grouped_gemm (nemo_rl.models.policy.MegatronConfig attribute) moe_hybridep_num_sms (nemo_rl.models.policy.MegatronConfig attribute) moe_layer_freq (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) moe_mesh (nemo_rl.models.automodel.config.DistributedContext attribute) moe_pad_experts_for_cuda_graph_inference (nemo_rl.models.policy.MegatronConfig attribute) moe_parallelizer (nemo_rl.models.policy.DTensorConfig attribute) moe_per_layer_logging (nemo_rl.models.policy.MegatronConfig attribute) moe_router_group_topk (nemo_rl.models.policy.MegatronConfig attribute) moe_router_num_groups (nemo_rl.models.policy.MegatronConfig attribute) moe_router_topk (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) moe_shared_expert_intermediate_size (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) moe_shared_expert_overlap (nemo_rl.models.policy.MegatronConfig attribute) moe_tensor_parallel_size (nemo_rl.models.generation.trtllm.config.TrtllmSpecificArgs attribute) moe_token_dispatcher_type (nemo_rl.models.policy.MegatronConfig attribute) MoEFloat16Module (class in nemo_rl.models.megatron.setup) MoEParallelizerOptions (class in nemo_rl.models.policy) monitor (nemo_rl.models.generation.fleet_health.HealthyShardSelector attribute) monitor_gpus (nemo_rl.utils.logger.LoggerConfig attribute) mooncake_cpu (nemo_rl.data_plane.interfaces.DataPlaneConfig attribute) MooncakeCpuConfig (class in nemo_rl.data_plane.interfaces) move_buffer_to_device() (nemo_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) (nemo_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl method) (nemo_rl.models.value.workers.dtensor_value_worker_v2.DTensorValueWorkerV2Impl method) move_model() (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) (nemo_rl.models.value.workers.megatron_value_worker.MegatronValueWorkerImpl method) move_optimizer() (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) (nemo_rl.models.value.workers.megatron_value_worker.MegatronValueWorkerImpl method) move_optimizer_to_device() (nemo_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) (nemo_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl method) (nemo_rl.models.value.workers.dtensor_value_worker_v2.DTensorValueWorkerV2Impl method) move_to_cpu() (nemo_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) (nemo_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl method) (nemo_rl.models.value.workers.dtensor_value_worker_v2.DTensorValueWorkerV2Impl method) move_to_cuda() (nemo_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) (nemo_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl method) (nemo_rl.models.value.workers.dtensor_value_worker_v2.DTensorValueWorkerV2Impl method) move_to_device() (nemo_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) (nemo_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl method) (nemo_rl.models.value.workers.dtensor_value_worker_v2.DTensorValueWorkerV2Impl method) MseValueLossConfig (class in nemo_rl.algorithms.loss.loss_functions) MseValueLossFn (class in nemo_rl.algorithms.loss.loss_functions) mtp_detach_heads (nemo_rl.models.policy.MegatronConfig attribute) mtp_loss_mask (nemo_rl.models.megatron.data.ProcessedInputs attribute) (nemo_rl.models.megatron.data.ProcessedMicrobatch attribute) mtp_loss_scaling_factor (nemo_rl.models.policy.MegatronConfig attribute) mtp_num_layers (nemo_rl.models.policy.MegatronConfig attribute) (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) mtp_use_repeated_layer (nemo_rl.models.policy.MegatronConfig attribute) multichoice_qa_processor() (in module nemo_rl.data.processors) MULTILINGUAL_ANSWER_PATTERN_TEMPLATE (in module nemo_rl.evals.answer_parsing) MULTILINGUAL_ANSWER_REGEXES (in module nemo_rl.evals.answer_parsing) MultilingualMultichoiceVerifyWorker (class in nemo_rl.environments.math_environment) MULTIMODAL_CONTENT_TYPES (in module nemo_rl.data.multimodal_utils) MULTIMODAL_DATASETS (in module nemo_rl.data.datasets.eval_datasets) MultipleDataloaderWrapper (class in nemo_rl.data.dataloader) MultiWorkerFuture (class in nemo_rl.distributed.worker_groups) mutation() (nemo_rl.algorithms.async_utils.replay_buffer.DataPlaneCheckpointBarrier method) N n_bytes (nemo_rl.data_plane.observability.DataPlaneEvent attribute) n_keys (nemo_rl.data_plane.observability.DataPlaneEvent attribute) name (nemo_rl.algorithms.advantage_estimator.AdvEstimatorConfig attribute) (nemo_rl.algorithms.advantage_estimator.GAEConfig attribute) (nemo_rl.algorithms.async_utils.staleness_sampler.CustomSamplerConfig attribute) (nemo_rl.algorithms.async_utils.staleness_sampler.InOrderSamplerConfig attribute) (nemo_rl.algorithms.async_utils.staleness_sampler.ReadyFirstSamplerConfig attribute) (nemo_rl.algorithms.async_utils.staleness_sampler.WeightFifoSamplerConfig attribute) (nemo_rl.algorithms.async_utils.staleness_sampler.WindowedSamplerConfig attribute) (nemo_rl.models.policy.PytorchOptimizerConfig attribute) (nemo_rl.models.policy.SinglePytorchSchedulerConfig attribute) (nemo_rl.models.policy.TokenizerConfig attribute) (nemo_rl.utils.checkpoint_engines.base.TensorMeta attribute) (nemo_rl.utils.logger.SwanlabConfig attribute) (nemo_rl.utils.logger.WandbConfig attribute) (nemo_rl.utils.weight_transfer_stream.SparseRefitTransport attribute) NamedSharding (class in nemo_rl.distributed.named_sharding) NamedTensor (in module nemo_rl.utils.weight_transfer_sparse_codec) names (nemo_rl.distributed.named_sharding.NamedSharding property) native (nemo_rl.algorithms.single_controller_utils.config.RolloutFailureConfig attribute) NATIVE_MULTIMODAL_KEYS (in module nemo_rl.data.multimodal_utils) NativeRolloutFTConfig (class in nemo_rl.algorithms.single_controller_utils.config) nbytes (nemo_rl.utils.checkpoint_engines.base.TensorMeta property) nccl_peer (nemo_rl.models.generation.interfaces.CollectiveSenderSpec attribute) nccl_reshard_refit() (nemo_rl.models.generation.interfaces.GenerationInterface method) (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) (nemo_rl.models.generation.vllm.vllm_worker.VllmGenerationWorkerImpl method) (nemo_rl.models.policy.interfaces.ColocatablePolicyInterface method) (nemo_rl.models.policy.lm_policy.Policy method) (nemo_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker method) (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) nccl_reshard_refit_async() (nemo_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) NcclExtension (class in nemo_rl.models.generation.trtllm.trtllm_backend) NcclReshardWeightSynchronizer (class in nemo_rl.weight_sync.nccl_reshard_weight_synchronizer) ndim (nemo_rl.distributed.named_sharding.NamedSharding property) (nemo_rl.weight_sync.nccl_reshard_utils.MeshInfo property) need_top_k_or_top_p_filtering() (in module nemo_rl.algorithms.logits_sampling_utils) nemo_gym (nemo_rl.algorithms.single_controller_utils.config.RolloutFailureConfig attribute) NEMO_GYM (nemo_rl.distributed.virtual_cluster.PY_EXECUTABLES attribute) nemo_gym_data_processor() (in module nemo_rl.data.processors) nemo_gym_example_to_video_datum_spec() (in module nemo_rl.environments.nemo_gym_video) NEMO_GYM_IMAGE_ENCODE_MAX_WORKERS (in module nemo_rl.data.multimodal_utils) nemo_gym_init_time_s (nemo_rl.algorithms.metric_utils.SetupTimingMetrics attribute) NEMO_GYM_TASK_INDEX_KEY (in module nemo_rl.experience.interfaces) NEMO_MODELOPT_W4A16 (in module nemo_rl.modelopt.models.generation.vllm_modelopt) NEMO_MODELOPT_W4A4 (in module nemo_rl.modelopt.models.generation.vllm_modelopt) nemo_rl module nemo_rl.algorithms module nemo_rl.algorithms.advantage_estimator module nemo_rl.algorithms.async_utils module nemo_rl.algorithms.async_utils.interfaces module nemo_rl.algorithms.async_utils.replay_buffer module nemo_rl.algorithms.async_utils.staleness_sampler module nemo_rl.algorithms.async_utils.trajectory_collector module nemo_rl.algorithms.distillation module nemo_rl.algorithms.dpo module nemo_rl.algorithms.grpo module nemo_rl.algorithms.grpo_sync module nemo_rl.algorithms.logits_sampling_utils module nemo_rl.algorithms.loss module nemo_rl.algorithms.loss.interfaces module nemo_rl.algorithms.loss.loss_functions module nemo_rl.algorithms.loss.utils module nemo_rl.algorithms.loss.wrapper module nemo_rl.algorithms.metric_utils module nemo_rl.algorithms.opd module nemo_rl.algorithms.ppo module nemo_rl.algorithms.reward_functions module nemo_rl.algorithms.rm module nemo_rl.algorithms.sft module nemo_rl.algorithms.single_controller module nemo_rl.algorithms.single_controller_utils module nemo_rl.algorithms.single_controller_utils.config module nemo_rl.algorithms.single_controller_utils.setup module nemo_rl.algorithms.single_controller_utils.utils module nemo_rl.algorithms.utils module nemo_rl.algorithms.x_token module nemo_rl.algorithms.x_token.loss_utils module nemo_rl.algorithms.x_token.token_aligner module nemo_rl.algorithms.x_token.utils module nemo_rl.algorithms.xtoken_off_policy_distillation module nemo_rl.data module nemo_rl.data.chat_templates module nemo_rl.data.collate_fn module nemo_rl.data.cross_tokenizer_collate module nemo_rl.data.dataloader module nemo_rl.data.datasets module nemo_rl.data.datasets.eval_datasets module nemo_rl.data.datasets.eval_datasets.daily_omni module nemo_rl.data.datasets.eval_datasets.gpqa module nemo_rl.data.datasets.eval_datasets.local_math_dataset module nemo_rl.data.datasets.eval_datasets.math module nemo_rl.data.datasets.eval_datasets.mmau module nemo_rl.data.datasets.eval_datasets.mmlu module nemo_rl.data.datasets.eval_datasets.mmlu_pro module nemo_rl.data.datasets.preference_datasets module nemo_rl.data.datasets.preference_datasets.binary_preference_dataset module nemo_rl.data.datasets.preference_datasets.helpsteer3 module nemo_rl.data.datasets.preference_datasets.preference_dataset module nemo_rl.data.datasets.preference_datasets.tulu3 module nemo_rl.data.datasets.processed_dataset module nemo_rl.data.datasets.raw_dataset module nemo_rl.data.datasets.response_datasets module nemo_rl.data.datasets.response_datasets.aime module nemo_rl.data.datasets.response_datasets.arrow_text_dataset module nemo_rl.data.datasets.response_datasets.audiomcq module nemo_rl.data.datasets.response_datasets.avqa module nemo_rl.data.datasets.response_datasets.clevr module nemo_rl.data.datasets.response_datasets.daily_omni module nemo_rl.data.datasets.response_datasets.dapo_math module nemo_rl.data.datasets.response_datasets.deepscaler module nemo_rl.data.datasets.response_datasets.general_conversations_dataset module nemo_rl.data.datasets.response_datasets.geometry3k module nemo_rl.data.datasets.response_datasets.gsm8k module nemo_rl.data.datasets.response_datasets.helpsteer3 module nemo_rl.data.datasets.response_datasets.intent module nemo_rl.data.datasets.response_datasets.mmpr_tiny module nemo_rl.data.datasets.response_datasets.nemogym_dataset module nemo_rl.data.datasets.response_datasets.nemotron_cascade2_sft module nemo_rl.data.datasets.response_datasets.numinamath module nemo_rl.data.datasets.response_datasets.oai_format_dataset module nemo_rl.data.datasets.response_datasets.oasst module nemo_rl.data.datasets.response_datasets.openmathinstruct2 module nemo_rl.data.datasets.response_datasets.openr1_math module nemo_rl.data.datasets.response_datasets.refcoco module nemo_rl.data.datasets.response_datasets.response_dataset module nemo_rl.data.datasets.response_datasets.squad module nemo_rl.data.datasets.response_datasets.tulu3 module nemo_rl.data.datasets.utils module nemo_rl.data.interfaces module nemo_rl.data.llm_message_utils module nemo_rl.data.multimodal_utils module nemo_rl.data.packing module nemo_rl.data.packing.algorithms module nemo_rl.data.packing.metrics module nemo_rl.data.processors module nemo_rl.data.utils module nemo_rl.data_plane module nemo_rl.data_plane.adapters module nemo_rl.data_plane.adapters.noop module nemo_rl.data_plane.adapters.transfer_queue module nemo_rl.data_plane.adapters.transfer_queue_env module nemo_rl.data_plane.async_utils module nemo_rl.data_plane.codec module nemo_rl.data_plane.column_io module nemo_rl.data_plane.driver_mixin module nemo_rl.data_plane.factory module nemo_rl.data_plane.interfaces module nemo_rl.data_plane.observability module nemo_rl.data_plane.preshard module nemo_rl.data_plane.schema module nemo_rl.data_plane.worker_mixin module nemo_rl.distributed module nemo_rl.distributed.batched_data_dict module nemo_rl.distributed.collectives module nemo_rl.distributed.held_port module nemo_rl.distributed.model_utils module nemo_rl.distributed.named_sharding module nemo_rl.distributed.numa_utils module nemo_rl.distributed.ray_actor_environment_registry module nemo_rl.distributed.refit_watchdog module nemo_rl.distributed.stateless_process_group module nemo_rl.distributed.virtual_cluster module nemo_rl.distributed.worker_group_utils module nemo_rl.distributed.worker_groups module nemo_rl.environments module nemo_rl.environments.code_environment module nemo_rl.environments.code_jaccard_environment module nemo_rl.environments.dapo_math_verifier module nemo_rl.environments.interfaces module nemo_rl.environments.math_environment module nemo_rl.environments.metrics module nemo_rl.environments.nemo_gym module nemo_rl.environments.nemo_gym_video module nemo_rl.environments.nemotron_utils module nemo_rl.environments.reward_model_environment module nemo_rl.environments.rewards module nemo_rl.environments.utils module nemo_rl.environments.vlm_environment module nemo_rl.evals module nemo_rl.evals.answer_parsing module nemo_rl.evals.eval module nemo_rl.experience module nemo_rl.experience.failures module nemo_rl.experience.interfaces module nemo_rl.experience.metric_utils module nemo_rl.experience.payload module nemo_rl.experience.rollout_manager module nemo_rl.experience.rollout_recovery module nemo_rl.experience.rollouts module nemo_rl.experience.sync_rollout_actor module nemo_rl.modelopt module nemo_rl.modelopt.models module nemo_rl.modelopt.models.generation module nemo_rl.modelopt.models.generation.vllm_modelopt module nemo_rl.modelopt.models.generation.vllm_quant_backend module nemo_rl.modelopt.models.generation.vllm_quant_patch module nemo_rl.modelopt.models.generation.vllm_quant_worker module nemo_rl.modelopt.models.policy module nemo_rl.modelopt.models.policy.workers module nemo_rl.modelopt.models.policy.workers.dtensor_quant_policy_worker module nemo_rl.modelopt.models.policy.workers.dtensor_quant_policy_worker_v2 module nemo_rl.modelopt.models.policy.workers.megatron_quant_policy_worker module nemo_rl.modelopt.models.policy.workers.utils module nemo_rl.modelopt.registry module nemo_rl.modelopt.utils module nemo_rl.models module nemo_rl.models.automodel module nemo_rl.models.automodel.checkpoint module nemo_rl.models.automodel.config module nemo_rl.models.automodel.data module nemo_rl.models.automodel.setup module nemo_rl.models.automodel.train module nemo_rl.models.dtensor module nemo_rl.models.dtensor.parallelize module nemo_rl.models.generation module nemo_rl.models.generation.constants module nemo_rl.models.generation.dynamo module nemo_rl.models.generation.dynamo.arguments module nemo_rl.models.generation.dynamo.config module nemo_rl.models.generation.dynamo.dynamo_generation module nemo_rl.models.generation.dynamo.dynamo_worker module nemo_rl.models.generation.dynamo.http_client module nemo_rl.models.generation.dynamo.managed_runtime module nemo_rl.models.generation.dynamo.metrics module nemo_rl.models.generation.dynamo.refit module nemo_rl.models.generation.dynamo.token_wrapper module nemo_rl.models.generation.dynamo.validate_dynamo_vllm_args module nemo_rl.models.generation.dynamo.venv module nemo_rl.models.generation.dynamo.worker_pool module nemo_rl.models.generation.fleet_health module nemo_rl.models.generation.generation_router module nemo_rl.models.generation.interfaces module nemo_rl.models.generation.megatron module nemo_rl.models.generation.megatron.config module nemo_rl.models.generation.megatron.megatron_generation module nemo_rl.models.generation.megatron.megatron_worker module nemo_rl.models.generation.megatron.utils module nemo_rl.models.generation.openai_server_utils module nemo_rl.models.generation.trtllm module nemo_rl.models.generation.trtllm.config module nemo_rl.models.generation.trtllm.trtllm_backend module nemo_rl.models.generation.trtllm.trtllm_generation module nemo_rl.models.generation.trtllm.trtllm_http_server module nemo_rl.models.generation.trtllm.trtllm_worker_async module nemo_rl.models.generation.vllm module nemo_rl.models.generation.vllm.checkpoint_engine module nemo_rl.models.generation.vllm.collective_rpc module nemo_rl.models.generation.vllm.config module nemo_rl.models.generation.vllm.patches module nemo_rl.models.generation.vllm.refit_layout module nemo_rl.models.generation.vllm.refit_loader module nemo_rl.models.generation.vllm.utils module nemo_rl.models.generation.vllm.video_utils module nemo_rl.models.generation.vllm.vllm_backend module nemo_rl.models.generation.vllm.vllm_generation module nemo_rl.models.generation.vllm.vllm_sparse_delta module nemo_rl.models.generation.vllm.vllm_sparse_refit module nemo_rl.models.generation.vllm.vllm_worker module nemo_rl.models.generation.vllm.vllm_worker_async module nemo_rl.models.generation.vllm.worker_utils module nemo_rl.models.huggingface module nemo_rl.models.huggingface.common module nemo_rl.models.megatron module nemo_rl.models.megatron.common module nemo_rl.models.megatron.community_import module nemo_rl.models.megatron.config module nemo_rl.models.megatron.data module nemo_rl.models.megatron.draft module nemo_rl.models.megatron.draft.eagle module nemo_rl.models.megatron.draft.hidden_capture module nemo_rl.models.megatron.draft.utils module nemo_rl.models.megatron.memory_saver module nemo_rl.models.megatron.pipeline_parallel module nemo_rl.models.megatron.router_replay module nemo_rl.models.megatron.setup module nemo_rl.models.megatron.train module nemo_rl.models.policy module nemo_rl.models.policy.interfaces module nemo_rl.models.policy.lm_policy module nemo_rl.models.policy.teacher_worker_group module nemo_rl.models.policy.tq_policy module nemo_rl.models.policy.utils module nemo_rl.models.policy.workers module nemo_rl.models.policy.workers.base_policy_worker module nemo_rl.models.policy.workers.checkpoint_engine module nemo_rl.models.policy.workers.dtensor_policy_worker module nemo_rl.models.policy.workers.dtensor_policy_worker_v2 module nemo_rl.models.policy.workers.megatron_policy_worker module nemo_rl.models.policy.workers.megatron_remote_sparse_refit module nemo_rl.models.policy.workers.patches module nemo_rl.models.value module nemo_rl.models.value.config module nemo_rl.models.value.interfaces module nemo_rl.models.value.lm_value module nemo_rl.models.value.tq_value module nemo_rl.models.value.workers module nemo_rl.models.value.workers.dtensor_value_worker_v2 module nemo_rl.models.value.workers.megatron_value_worker module nemo_rl.package_info module nemo_rl.telemetry module nemo_rl.telemetry.config module nemo_rl.telemetry.instrumentation module nemo_rl.telemetry.metrics module nemo_rl.telemetry.setup module nemo_rl.telemetry.span_groups module nemo_rl.utils module nemo_rl.utils.checkpoint module nemo_rl.utils.checkpoint_engines module nemo_rl.utils.checkpoint_engines.base module nemo_rl.utils.checkpoint_engines.nixl module nemo_rl.utils.config module nemo_rl.utils.fastokens module nemo_rl.utils.flops_formulas module nemo_rl.utils.flops_tracker module nemo_rl.utils.grad_norm module nemo_rl.utils.logger module nemo_rl.utils.memory_tracker module nemo_rl.utils.multimodal_payload_metrics module nemo_rl.utils.native_checkpoint module nemo_rl.utils.nsys module nemo_rl.utils.nvml module nemo_rl.utils.packed_tensor module nemo_rl.utils.prefetch_venvs module nemo_rl.utils.r3_trace module nemo_rl.utils.routed_experts_codec module nemo_rl.utils.timer module nemo_rl.utils.venvs module nemo_rl.utils.weight_transfer_http module nemo_rl.utils.weight_transfer_sparse_codec module nemo_rl.utils.weight_transfer_stream module nemo_rl.utils.weight_transfer_zmq module nemo_rl.weight_sync module nemo_rl.weight_sync.checkpoint_engine_config module nemo_rl.weight_sync.checkpoint_engine_weight_synchronizer module nemo_rl.weight_sync.collective_weight_synchronizer module nemo_rl.weight_sync.factory module nemo_rl.weight_sync.interfaces module nemo_rl.weight_sync.ipc_weight_synchronizer module nemo_rl.weight_sync.megatron_weight_synchronizer module nemo_rl.weight_sync.membership module nemo_rl.weight_sync.nccl_reshard_utils module nemo_rl.weight_sync.nccl_reshard_weight_synchronizer module nemo_rl.weight_sync.sglang_weight_synchronizer module nemo_rl.weight_sync.vllm_remote_sparse_weight_synchronizer module nemo_rl.weight_sync.xferdtensor module nemo_rl.weight_sync.xferdtensor_python module NemoGym (class in nemo_rl.environments.nemo_gym) NemoGymCompatibleConfig (class in nemo_rl.environments.nemo_gym) NemoGymConfig (class in nemo_rl.environments.nemo_gym) NemoGymDataset (class in nemo_rl.data.datasets.response_datasets.nemogym_dataset) NemoGymRolloutFTConfig (class in nemo_rl.algorithms.single_controller_utils.config) NemoGymRolloutResult (class in nemo_rl.experience.rollouts) nemotron() (in module nemo_rl.utils.flops_formulas) NEMOTRON_VIDEO_PROCESSOR_NAMES (in module nemo_rl.environments.nemotron_utils) NemotronCascade2SFTMathDataset (class in nemo_rl.data.datasets.response_datasets.nemotron_cascade2_sft) nemotronh() (in module nemo_rl.utils.flops_formulas) new_variables (nemo_rl.utils.memory_tracker.MemoryTrackerDataPoint property) next_index (nemo_rl.utils.weight_transfer_stream._SparsePayloadBucket attribute) NEXT_NEMO_GYM_TASK_INDEX_KEY (in module nemo_rl.experience.interfaces) next_shard() (nemo_rl.models.generation.fleet_health.HealthyShardSelector method) next_stop_strings (nemo_rl.environments.interfaces.EnvironmentReturn attribute) next_token_accuracy() (in module nemo_rl.algorithms.x_token.loss_utils) nixl (nemo_rl.models.generation.vllm.config.VllmRefitConfig attribute) NIXL_DEFAULT_BACKEND_NAME (in module nemo_rl.utils.checkpoint_engines.nixl) NIXL_TRANSFER_BUFFER_COUNT (in module nemo_rl.utils.checkpoint_engines.nixl) NIXL_VLLM_WORKER (in module nemo_rl.models.generation.vllm.checkpoint_engine) NixlAgent (class in nemo_rl.utils.checkpoint_engines.nixl) NixlAgentMetadata (in module nemo_rl.utils.checkpoint_engines.nixl) NIXLCheckpointEngine (class in nemo_rl.utils.checkpoint_engines.nixl) NixlVllmWorker (class in nemo_rl.models.generation.vllm.vllm_backend) NLLLossFn (class in nemo_rl.algorithms.loss.loss_functions) no_healthy_backend_status (nemo_rl.algorithms.single_controller_utils.config.GenerationRouterConfig attribute) node_count() (nemo_rl.distributed.virtual_cluster.RayVirtualCluster method) NoHealthyShards non_colocated_teachers (nemo_rl.algorithms.opd.OnPolicyDistillationConfig attribute) NON_VERIFIABLE_ANSWERS (in module nemo_rl.data.datasets.response_datasets.numinamath) NonColocatedTeachersConfig (class in nemo_rl.algorithms.opd) NONE (nemo_rl.algorithms.loss.interfaces.MetricNormalizer attribute) NoOpDataPlaneClient (class in nemo_rl.data_plane.adapters.noop) normalize_advantages (nemo_rl.algorithms.advantage_estimator.GAEConfig attribute) normalize_extracted_answer() (in module nemo_rl.evals.answer_parsing) normalize_final_answer() (in module nemo_rl.environments.dapo_math_verifier) normalize_response() (in module nemo_rl.evals.answer_parsing) normalize_rewards (nemo_rl.algorithms.advantage_estimator.AdvEstimatorConfig attribute) (nemo_rl.algorithms.grpo.GRPOConfig attribute) normalize_teacher_by_vocab (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute) normalize_video_urls_in_examples() (in module nemo_rl.environments.nemo_gym_video) normalize_vllm_refit_config() (in module nemo_rl.models.generation.vllm.config) NoSurvivingShards NRL_NSYS_EXTRA_OPTIONS (in module nemo_rl.utils.nsys) NRL_NSYS_PROFILE_STEP_RANGE (in module nemo_rl.utils.nsys) NRL_NSYS_WORKER_PATTERNS (in module nemo_rl.utils.nsys) num_buffers (nemo_rl.models.generation.interfaces.CollectiveSenderSpec attribute) num_chunks (nemo_rl.algorithms.x_token.token_aligner.AlignmentBatch attribute) num_cuda_graphs (nemo_rl.models.generation.megatron.config.MCoreGenerationSpecificArgs attribute) num_frames (nemo_rl.data.interfaces.TaskDataSpec attribute) (nemo_rl.data.PreferenceDatasetConfig attribute) (nemo_rl.data.ResponseDatasetConfig attribute) (nemo_rl.models.generation.vllm.config.VllmVideoConfig attribute) num_generations_per_prompt (nemo_rl.algorithms.distillation.DistillationConfig attribute) (nemo_rl.algorithms.grpo.GRPOConfig attribute) (nemo_rl.algorithms.ppo.PPOConfig attribute) num_layers (nemo_rl.models.policy.DraftConfig attribute) num_layers_in_first_pipeline_stage (nemo_rl.models.policy.MegatronConfig attribute) num_layers_in_last_pipeline_stage (nemo_rl.models.policy.MegatronConfig attribute) num_nodes (nemo_rl.algorithms.opd.TeacherResourceConfig attribute) (nemo_rl.distributed.virtual_cluster.ClusterConfig attribute) (nemo_rl.models.generation.interfaces.OptionalResourcesConfig attribute) (nemo_rl.models.generation.interfaces.ResourcesConfig attribute) (nemo_rl.models.policy.teacher_worker_group.TeacherConfig attribute) num_prompts_per_dataloader (nemo_rl.data.DataConfig attribute) num_prompts_per_step (nemo_rl.algorithms.distillation.DistillationConfig attribute) (nemo_rl.algorithms.grpo.GRPOConfig attribute) (nemo_rl.algorithms.ppo.PPOConfig attribute) (nemo_rl.algorithms.xtoken_off_policy_distillation.OffPolicyDistillationConfig attribute) num_samples (nemo_rl.data_plane.adapters.noop._Partition attribute) num_speculative_tokens (nemo_rl.models.generation.megatron.config.MCoreGenerationSpecificArgs attribute) num_storage_units (nemo_rl.data_plane.interfaces.SimpleStorageConfig attribute) num_tests_per_prompt (nemo_rl.evals.eval.EvalConfig attribute) num_val_samples_to_print (nemo_rl.algorithms.grpo.GRPOLoggerConfig attribute) (nemo_rl.algorithms.ppo.PPOLoggerConfig attribute) (nemo_rl.utils.logger.LoggerConfig attribute) num_valid_samples (nemo_rl.algorithms.dpo.DPOValMetrics attribute) (nemo_rl.algorithms.rm.RMValMetrics attribute) num_workers (nemo_rl.data.DataConfig attribute) (nemo_rl.environments.code_environment.CodeEnvConfig attribute) (nemo_rl.environments.code_jaccard_environment.CodeJaccardEnvConfig attribute) (nemo_rl.environments.math_environment.MathEnvConfig attribute) (nemo_rl.environments.vlm_environment.VLMEnvConfig attribute) NuminaMath15Dataset (class in nemo_rl.data.datasets.response_datasets.numinamath) NVFP4RealQuantMode (in module nemo_rl.modelopt.utils) NVLINK_DOMAIN_PREFIX (in module nemo_rl.distributed.virtual_cluster) nvlink_domain_span() (nemo_rl.models.generation.megatron.megatron_generation.MegatronGeneration class method) NVLINK_DOMAIN_UNKNOWN (in module nemo_rl.distributed.virtual_cluster) nvml_context() (in module nemo_rl.utils.nvml) O OasstDataset (class in nemo_rl.data.datasets.response_datasets.oasst) observability (nemo_rl.data_plane.interfaces.DataPlaneConfig attribute) ObservabilityConfig (class in nemo_rl.data_plane.interfaces) observations (nemo_rl.environments.interfaces.EnvironmentReturn attribute) offload_after_refit() (nemo_rl.models.policy.interfaces.ColocatablePolicyInterface method) (nemo_rl.models.policy.lm_policy.Policy method) (nemo_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) (nemo_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl method) (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) offload_before_refit() (nemo_rl.models.policy.interfaces.ColocatablePolicyInterface method) (nemo_rl.models.policy.lm_policy.Policy method) (nemo_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) (nemo_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl method) (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) offload_modules (nemo_rl.models.policy.MegatronConfig attribute) offload_optimizer_for_logprob (nemo_rl.models.automodel.config.RuntimeConfig attribute) (nemo_rl.models.megatron.config.RuntimeConfig attribute) offload_optimizer_for_refit (nemo_rl.models.megatron.config.RuntimeConfig attribute) offload_to_cpu() (nemo_rl.models.policy.interfaces.ColocatablePolicyInterface method) (nemo_rl.models.policy.lm_policy.Policy method) OffPolicyDistillationConfig (class in nemo_rl.algorithms.xtoken_off_policy_distillation) OffPolicyDistillationSaveState (class in nemo_rl.algorithms.xtoken_off_policy_distillation) offset (nemo_rl.utils.checkpoint_engines.base.TensorMeta attribute) on_dead_shard (nemo_rl.algorithms.single_controller_utils.config.FleetHealthConfig attribute) on_dropped_prompt (nemo_rl.algorithms.single_controller_utils.config.RolloutFailureConfig attribute) on_policy_distillation (nemo_rl.algorithms.grpo.MasterConfig attribute) (nemo_rl.algorithms.single_controller_utils.config.MasterConfig attribute) on_sync_failed() (nemo_rl.utils.weight_transfer_sparse_codec.DeltaCompressionTracker method) on_sync_succeeded() (nemo_rl.utils.weight_transfer_sparse_codec.DeltaCompressionTracker method) only_unmask_final (nemo_rl.algorithms.sft.SFTConfig attribute) OnPolicyDistillationConfig (class in nemo_rl.algorithms.opd) op (nemo_rl.data_plane.observability.DataPlaneEvent attribute) OPDAdvantageEstimator (class in nemo_rl.algorithms.advantage_estimator) OpenAIFormatDataset (class in nemo_rl.data.datasets.response_datasets.oai_format_dataset) OpenMathInstruct2Dataset (class in nemo_rl.data.datasets.response_datasets.openmathinstruct2) OpenR1Math220KDataset (class in nemo_rl.data.datasets.response_datasets.openr1_math) OPT_IN_CARRY_KEYS (in module nemo_rl.experience.sync_rollout_actor) optimizer (nemo_rl.models.automodel.config.ModelAndOptimizerState attribute) (nemo_rl.models.megatron.config.ModelAndOptimizerState attribute) (nemo_rl.models.policy.MegatronConfig attribute) (nemo_rl.models.policy.MegatronOptimizerConfig attribute) (nemo_rl.models.policy.PolicyConfig attribute) (nemo_rl.models.value.config.ValueConfig attribute) optimizer_cpu_offload (nemo_rl.models.megatron.config.RuntimeConfig attribute) (nemo_rl.models.policy.MegatronOptimizerConfig attribute) optimizer_offload_fraction (nemo_rl.models.policy.MegatronOptimizerConfig attribute) OptimizerState (class in nemo_rl.utils.native_checkpoint) OptionalResourcesConfig (class in nemo_rl.models.generation.interfaces) original_batch_size (nemo_rl.models.automodel.data.ProcessedMicrobatch attribute) original_seq_len (nemo_rl.models.automodel.data.ProcessedMicrobatch attribute) original_seq_length (nemo_rl.models.megatron.data.ProcessedInputs attribute) (nemo_rl.models.megatron.data.ProcessedMicrobatch attribute) other_setup_time_s (nemo_rl.algorithms.metric_utils.SetupTimingMetrics attribute) output_field (nemo_rl.algorithms.single_controller_utils.config.AdvantageConfig attribute) output_ids (nemo_rl.models.generation.interfaces.GenerationOutputSpec attribute) output_key (nemo_rl.data.ResponseDatasetConfig attribute) OVERHEAD (nemo_rl.telemetry.instrumentation.Bucket attribute) overlap_cpu_optimizer_d2h_h2d (nemo_rl.models.policy.MegatronOptimizerConfig attribute) overlap_grad_reduce (nemo_rl.models.policy.MegatronDDPConfig attribute) overlap_param_gather (nemo_rl.models.policy.MegatronDDPConfig attribute) overlong_buffer_length (nemo_rl.algorithms.reward_functions.RewardShapingConfig attribute) overlong_buffer_penalty (nemo_rl.algorithms.reward_functions.RewardShapingConfig attribute) overlong_filtering (nemo_rl.algorithms.grpo.GRPOConfig attribute) (nemo_rl.algorithms.ppo.PPOConfig attribute) OverridesError P pack() (nemo_rl.data.packing.algorithms.SequencePacker method) pack_jagged_fields() (in module nemo_rl.data_plane.codec) pack_payload() (in module nemo_rl.experience.payload) pack_per_token_field() (in module nemo_rl.data_plane.codec) pack_sequences() (in module nemo_rl.models.huggingface.common) packed_broadcast_consumer() (in module nemo_rl.utils.packed_tensor) packed_broadcast_producer() (in module nemo_rl.utils.packed_tensor) packed_seq_params (nemo_rl.models.megatron.data.ProcessedInputs attribute) (nemo_rl.models.megatron.data.ProcessedMicrobatch attribute) PackedTensor (class in nemo_rl.data.multimodal_utils) PackingAlgorithm (class in nemo_rl.data.packing.algorithms) PackingMetrics (class in nemo_rl.data.packing.metrics) pad_and_align_routed_expert_indices() (in module nemo_rl.models.generation.vllm.utils) pad_distillation_val_batch() (in module nemo_rl.algorithms.x_token.utils) pad_dynamic_image_shapes (nemo_rl.environments.nemo_gym.NemoGymConfig attribute) pair_is_correct (nemo_rl.algorithms.x_token.loss_utils.LocalizedAlignment attribute) (nemo_rl.algorithms.x_token.token_aligner.AlignmentBatch attribute) pair_valid (nemo_rl.algorithms.x_token.loss_utils.LocalizedAlignment attribute) (nemo_rl.algorithms.x_token.token_aligner.AlignmentBatch attribute) parallel_init_enabled (nemo_rl.algorithms.metric_utils.SetupTimingMetrics attribute) parallel_wall_time_s (nemo_rl.algorithms.metric_utils.SetupTimingMetrics attribute) PARALLIZE_FUNCTIONS (in module nemo_rl.models.dtensor.parallelize) param_sync_func (nemo_rl.models.megatron.config.ModelAndOptimizerState attribute) parameter_name (nemo_rl.models.generation.vllm.refit_layout.HfExpertWeight attribute) params_dtype (nemo_rl.models.policy.MegatronOptimizerConfig attribute) parse_conversations() (in module nemo_rl.data.datasets.response_datasets.oasst) parse_hf_expert_weight() (in module nemo_rl.models.generation.vllm.refit_layout) parse_hydra_overrides() (in module nemo_rl.utils.config) parse_projection_file() (in module nemo_rl.algorithms.x_token.loss_utils) parse_prometheus_metrics() (in module nemo_rl.models.generation.dynamo.metrics) parse_rollout_recovery_state() (in module nemo_rl.experience.rollout_recovery) ParsedRolloutRecoveryState (class in nemo_rl.experience.rollout_recovery) parsers (nemo_rl.models.generation.megatron.config.MCoreGenerationSpecificArgs attribute) partition_id (nemo_rl.algorithms.async_utils.replay_buffer.DataPlaneCheckpointMetadata attribute) (nemo_rl.algorithms.async_utils.replay_buffer.TQReplayMetadataState attribute) (nemo_rl.algorithms.single_controller_utils.setup.SingleControllerActorArgs attribute) (nemo_rl.data_plane.interfaces.KVBatchMeta attribute) (nemo_rl.data_plane.observability.DataPlaneEvent attribute) partition_workers (nemo_rl.models.generation.vllm.config.VllmRefitTuningConfig attribute) passthrough_prompt_response (nemo_rl.data.chat_templates.COMMON_CHAT_TEMPLATES attribute) PATCH (in module nemo_rl.package_info) patch_dim (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) patch_gpt_model_forward_for_linear_ce_fusion() (in module nemo_rl.distributed.model_utils) patch_transformers_module_dir() (in module nemo_rl) path (nemo_rl.models.generation.vllm.vllm_sparse_refit._StagedSparsePayload attribute) (nemo_rl.utils.checkpoint.PretrainedCheckpointConfig attribute) PathLike (in module nemo_rl.data.interfaces) (in module nemo_rl.models.policy.lm_policy) (in module nemo_rl.models.value.lm_value) (in module nemo_rl.utils.checkpoint) pause() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector method) pause_generation() (nemo_rl.models.generation.interfaces.GenerationInterface method) pause_generation_async() (nemo_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) pause_generation_for_refit() (nemo_rl.models.generation.interfaces.GenerationInterface method) (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) pause_inference_weights() (in module nemo_rl.models.megatron.memory_saver) payloads (nemo_rl.utils.weight_transfer_stream._SparsePayloadBucket attribute) pct() (in module nemo_rl.experience.metric_utils) peak_bytes_outstanding (nemo_rl.data_plane.observability.DataPlaneStats attribute) peak_lookahead_versions (nemo_rl.algorithms.async_utils.staleness_sampler.InOrderSamplerConfig property) peft (nemo_rl.models.policy.MegatronConfig attribute) peft_config (nemo_rl.models.automodel.config.ModelAndOptimizerState attribute) (nemo_rl.utils.checkpoint.CheckpointingConfig attribute) penalize_duplicated_reasoning (nemo_rl.algorithms.grpo.RewardPenaltyConfig attribute) penalize_empty_final_answer (nemo_rl.algorithms.grpo.RewardPenaltyConfig attribute) penalize_malformed_think_tag (nemo_rl.algorithms.grpo.RewardPenaltyConfig attribute) penalize_unwanted_tokens (nemo_rl.algorithms.grpo.RewardPenaltyConfig attribute) PENDING_PROMPTS_KEY (in module nemo_rl.experience.interfaces) phase (nemo_rl.experience.rollout_recovery.PromptGroupRecoveryRecord attribute) (nemo_rl.experience.rollout_recovery.PromptGroupRecoveryState attribute) pil_to_base64() (in module nemo_rl.data.datasets.utils) ping() (nemo_rl.algorithms.single_controller.SingleControllerActor method) pipeline_dtype (nemo_rl.models.policy.MegatronConfig attribute) pipeline_model_parallel_size (nemo_rl.algorithms.opd.TeacherResourceConfig attribute) (nemo_rl.models.policy.MegatronConfig attribute) (nemo_rl.models.policy.teacher_worker_group.TeacherConfig attribute) pipeline_parallel_size (nemo_rl.models.generation.dynamo.config.DynamoVllmConfig attribute) (nemo_rl.models.generation.vllm.config.VllmSpecificArgs attribute) plan_refit_membership() (in module nemo_rl.weight_sync.membership) Policy (class in nemo_rl.models.policy.lm_policy) policy (nemo_rl.algorithms.distillation.MasterConfig attribute) (nemo_rl.algorithms.dpo.MasterConfig attribute) (nemo_rl.algorithms.grpo.MasterConfig attribute) (nemo_rl.algorithms.ppo.MasterConfig attribute) (nemo_rl.algorithms.rm.MasterConfig attribute) (nemo_rl.algorithms.sft.MasterConfig attribute) (nemo_rl.algorithms.single_controller_utils.config.MasterConfig attribute) (nemo_rl.algorithms.xtoken_off_policy_distillation.MasterConfig attribute) (nemo_rl.environments.nemo_gym.NemoGymCompatibleConfig property) policy_config() (nemo_rl.algorithms.xtoken_off_policy_distillation.TeacherConfig method) policy_init_time_s (nemo_rl.algorithms.metric_utils.SetupTimingMetrics attribute) policy_logprobs_field (nemo_rl.algorithms.single_controller_utils.config.AdvantageConfig attribute) policy_training_start_step (nemo_rl.algorithms.ppo.PPOConfig attribute) POLICY_UPDATE (nemo_rl.telemetry.span_groups.RLSpanGroup attribute) POLICY_WORKER_OVERRIDES (in module nemo_rl.models.policy.utils) PolicyCheckpointEngineMixin (class in nemo_rl.models.policy.workers.checkpoint_engine) PolicyConfig (class in nemo_rl.models.policy) PolicyInterface (class in nemo_rl.models.policy.interfaces) pool_for() (nemo_rl.data_plane.adapters.transfer_queue._StagingPoolRegistry method) port_range_high (nemo_rl.algorithms.single_controller_utils.config.GenerationRouterConfig attribute) (nemo_rl.environments.nemo_gym.NemoGymConfig attribute) (nemo_rl.models.generation.interfaces.GenerationConfig attribute) port_range_low (nemo_rl.algorithms.single_controller_utils.config.GenerationRouterConfig attribute) (nemo_rl.environments.nemo_gym.NemoGymConfig attribute) (nemo_rl.models.generation.interfaces.GenerationConfig attribute) position_ids (nemo_rl.models.automodel.data.ProcessedInputs attribute) (nemo_rl.models.megatron.data.ProcessedInputs attribute) (nemo_rl.models.megatron.data.ProcessedMicrobatch attribute) positive_example_nll_weight (nemo_rl.algorithms.loss.loss_functions.ClippedPGLossConfig attribute) post (nemo_rl.weight_sync.nccl_reshard_utils.LocalParamSpec attribute) post_attention_layernorm_key (nemo_rl.models.megatron.draft.utils._EagleLayerLayout attribute) post_init() (nemo_rl.models.generation.vllm.vllm_worker.VllmGenerationWorkerImpl method) post_init_async() (nemo_rl.models.generation.trtllm.trtllm_worker_async.TrtllmAsyncGenerationWorkerImpl method) (nemo_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) post_vllm_refit_endpoints() (in module nemo_rl.utils.weight_transfer_http) PostProcessingFunction (in module nemo_rl.models.automodel.train) (in module nemo_rl.models.megatron.train) PostWriteEnrichmentError pp_comm_group (nemo_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker attribute) pp_comm_groups (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension attribute) ppo (nemo_rl.algorithms.ppo.MasterConfig attribute) (nemo_rl.algorithms.single_controller_utils.config.MasterConfig attribute) ppo_epochs (nemo_rl.algorithms.ppo.PPOConfig attribute) ppo_train() (in module nemo_rl.algorithms.ppo) PPO_VALUE_FIELDS (in module nemo_rl.data_plane.schema) PPOConfig (class in nemo_rl.algorithms.ppo) PPOLoggerConfig (class in nemo_rl.algorithms.ppo) PPOSaveState (class in nemo_rl.algorithms.ppo) pre (nemo_rl.weight_sync.nccl_reshard_utils.LocalParamSpec attribute) PRE_RELEASE (in module nemo_rl.package_info) precision (nemo_rl.algorithms.opd.TeacherResourceConfig attribute) (nemo_rl.environments.reward_model_environment.RewardModelEnvironmentConfig attribute) (nemo_rl.models.generation.dynamo.config.DynamoVllmConfig attribute) (nemo_rl.models.generation.trtllm.config.TrtllmSpecificArgs attribute) (nemo_rl.models.generation.vllm.config.VllmSpecificArgs attribute) (nemo_rl.models.policy.PolicyConfig attribute) (nemo_rl.models.policy.teacher_worker_group.TeacherConfig attribute) (nemo_rl.models.value.config.ValueConfig attribute) preference_average_log_probs (nemo_rl.algorithms.dpo.DPOConfig attribute) (nemo_rl.algorithms.loss.loss_functions.DPOLossConfig attribute) preference_collate_fn() (in module nemo_rl.data.collate_fn) preference_loss (nemo_rl.algorithms.dpo.DPOValMetrics attribute) preference_loss_weight (nemo_rl.algorithms.dpo.DPOConfig attribute) (nemo_rl.algorithms.loss.loss_functions.DPOLossConfig attribute) preference_preprocessor() (in module nemo_rl.data.processors) PreferenceDataset (class in nemo_rl.data.datasets.preference_datasets.preference_dataset) PreferenceDatasetConfig (class in nemo_rl.data) PreferenceDatumSpec (class in nemo_rl.data.interfaces) PreferenceLossDataDict (class in nemo_rl.algorithms.loss.loss_functions) PreferenceLossFn (class in nemo_rl.algorithms.loss.loss_functions) prefetch_venvs() (in module nemo_rl.utils.prefetch_venvs) preinit_nixl_agent() (in module nemo_rl.utils.checkpoint_engines.nixl) preinit_nixl_from_vllm_config() (in module nemo_rl.models.generation.vllm.checkpoint_engine) preinit_nvshmem() (nemo_rl.models.policy.lm_policy.Policy method) preinit_nvshmem_collective() (nemo_rl.models.generation.megatron.megatron_generation.MegatronGeneration method) (nemo_rl.models.generation.megatron.megatron_worker.MegatronGenerationRefitMixin method) prepare() (nemo_rl.models.generation.dynamo.refit.DynamoRefitChannel method) (nemo_rl.utils.checkpoint_engines.base.CheckpointEngine method) (nemo_rl.utils.checkpoint_engines.nixl.NIXLCheckpointEngine method) prepare_checkpoint_engine() (nemo_rl.models.generation.vllm.checkpoint_engine.VllmCheckpointEngineMixin method) prepare_dynamo_chat_completion_request() (in module nemo_rl.models.generation.dynamo.token_wrapper) prepare_for_generation() (nemo_rl.models.generation.dynamo.dynamo_generation.DynamoGeneration method) (nemo_rl.models.generation.interfaces.GenerationInterface method) (nemo_rl.models.generation.megatron.megatron_generation.MegatronGeneration method) (nemo_rl.models.generation.megatron.megatron_worker.MegatronGenerationMixin method) (nemo_rl.models.generation.trtllm.trtllm_generation.TrtllmGeneration method) (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) (nemo_rl.models.policy.lm_policy.Policy method) prepare_for_inference() (nemo_rl.models.value.interfaces.ValueInterface method) (nemo_rl.models.value.lm_value.Value method) (nemo_rl.models.value.workers.dtensor_value_worker_v2.DTensorValueWorkerV2Impl method) (nemo_rl.models.value.workers.megatron_value_worker.MegatronValueWorkerImpl method) prepare_for_lp_inference() (nemo_rl.models.policy.interfaces.ColocatablePolicyInterface method) (nemo_rl.models.policy.lm_policy.Policy method) (nemo_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) (nemo_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl method) (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) prepare_for_refit() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector method) prepare_for_training() (nemo_rl.models.policy.interfaces.PolicyInterface method) (nemo_rl.models.policy.lm_policy.Policy method) (nemo_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) (nemo_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl method) (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) (nemo_rl.models.value.interfaces.ValueInterface method) (nemo_rl.models.value.lm_value.Value method) (nemo_rl.models.value.workers.dtensor_value_worker_v2.DTensorValueWorkerV2Impl method) (nemo_rl.models.value.workers.megatron_value_worker.MegatronValueWorkerImpl method) prepare_loss_input() (in module nemo_rl.algorithms.loss.utils) prepare_model_forward() (in module nemo_rl.models.automodel.train) prepare_nccl_reshard_refit_info() (nemo_rl.models.generation.interfaces.GenerationInterface method) (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) (nemo_rl.models.generation.vllm.vllm_worker.VllmGenerationWorkerImpl method) (nemo_rl.models.policy.interfaces.ColocatablePolicyInterface method) (nemo_rl.models.policy.lm_policy.Policy method) (nemo_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker method) (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) prepare_nccl_reshard_refit_info_async() (nemo_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) prepare_packed_loss_input() (in module nemo_rl.algorithms.loss.utils) prepare_refit_info() (nemo_rl.modelopt.models.generation.vllm_quant_backend.VllmQuantInternalWorkerExtension method) (nemo_rl.models.generation.dynamo.dynamo_generation.DynamoGeneration method) (nemo_rl.models.generation.interfaces.GenerationInterface method) (nemo_rl.models.generation.megatron.megatron_generation.MegatronGeneration method) (nemo_rl.models.generation.trtllm.trtllm_backend.NcclExtension method) (nemo_rl.models.generation.trtllm.trtllm_generation.TrtllmGeneration method) (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) (nemo_rl.models.generation.vllm.vllm_worker.VllmGenerationWorkerImpl method) (nemo_rl.models.policy.interfaces.ColocatablePolicyInterface method) (nemo_rl.models.policy.lm_policy.Policy method) (nemo_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) (nemo_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl method) (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) prepare_refit_info_async() (nemo_rl.models.generation.trtllm.trtllm_worker_async.TrtllmAsyncGenerationWorkerImpl method) (nemo_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) prepare_segment_topology() (in module nemo_rl.distributed.virtual_cluster) prepare_sparse_delta_payload() (nemo_rl.utils.weight_transfer_sparse_codec.DeltaCompressionTracker method) prepare_sparse_delta_refit_info() (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) prepare_step() (nemo_rl.models.policy.tq_policy.TQPolicy method) prepare_val_partition() (nemo_rl.models.policy.tq_policy.TQPolicy method) prepare_xtoken_cross_tokenizer_loss_input() (in module nemo_rl.algorithms.x_token.loss_utils) PreparedModelForward (class in nemo_rl.models.automodel.train) PreparedTensorPayload (in module nemo_rl.utils.weight_transfer_sparse_codec) preprocess_data() (nemo_rl.environments.reward_model_environment.RewardModelEnvironment method) preprocessor (nemo_rl.data.datasets.raw_dataset.RawDataset attribute) PreservingDataset (class in nemo_rl.data.datasets.response_datasets.oai_format_dataset) pretrained_checkpoint (nemo_rl.models.policy.PolicyConfig attribute) (nemo_rl.utils.checkpoint.CheckpointingConfig attribute) PretrainedCheckpointConfig (class in nemo_rl.utils.checkpoint) prev_logprobs (nemo_rl.algorithms.loss.loss_functions.ClippedPGLossDataDict attribute) print_aggregated_stats() (nemo_rl.data.packing.metrics.PackingMetrics method) print_efficiency_summary() (in module nemo_rl.algorithms.utils) print_message_log_samples() (in module nemo_rl.utils.logger) print_metrics() (nemo_rl.data.packing.algorithms.SequencePacker method) print_multimodal_payload_metrics() (in module nemo_rl.utils.multimodal_payload_metrics) print_node_ip_and_gpu_id() (nemo_rl.models.policy.lm_policy.Policy method) print_performance_metrics() (in module nemo_rl.algorithms.utils) print_setup_timing_summary() (in module nemo_rl.algorithms.metric_utils) probe_interval_s (nemo_rl.algorithms.single_controller_utils.config.FleetHealthConfig attribute) probe_timeout_s (nemo_rl.algorithms.single_controller_utils.config.FleetHealthConfig attribute) problem_key (nemo_rl.data.LocalMathEvalDataConfig attribute) process_global_batch() (in module nemo_rl.models.automodel.data) (in module nemo_rl.models.megatron.data) process_message_fragment() (nemo_rl.data.datasets.response_datasets.general_conversations_dataset.GeneralConversationsJsonlDataset class method) process_microbatch() (in module nemo_rl.models.automodel.data) (in module nemo_rl.models.megatron.data) process_nemotron_video_frames() (in module nemo_rl.environments.nemotron_utils) processed_inputs (nemo_rl.models.automodel.data.ProcessedMicrobatch attribute) ProcessedInputs (class in nemo_rl.models.automodel.data) (class in nemo_rl.models.megatron.data) ProcessedMicrobatch (class in nemo_rl.models.automodel.data) (class in nemo_rl.models.megatron.data) processor (nemo_rl.data.AIMEEvalDataConfig attribute) (nemo_rl.data.datasets.raw_dataset.RawDataset attribute) (nemo_rl.data.ResponseDatasetConfig attribute) PROCESSOR_REGISTRY (in module nemo_rl.data.processors) PRODUCTIVE (nemo_rl.telemetry.instrumentation.Bucket attribute) ProfilablePolicy (class in nemo_rl.utils.nsys) proj_weight_key (nemo_rl.models.megatron.draft.utils._EagleLayerLayout property) project (nemo_rl.utils.logger.SwanlabConfig attribute) (nemo_rl.utils.logger.WandbConfig attribute) project_student_to_teacher_vocab() (in module nemo_rl.algorithms.x_token.loss_utils) projection_matrix_path (nemo_rl.algorithms.xtoken_off_policy_distillation.TeacherConfig attribute) projection_matrix_paths (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute) projector_type (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) PROMOTE_1D_FIELDS (in module nemo_rl.data_plane.schema) promote_ready_group() (nemo_rl.algorithms.async_utils.replay_buffer.TQReplayBuffer method) prompt (nemo_rl.experience.interfaces.PromptGroupRecord attribute) prompt_file (nemo_rl.data.AIMEEvalDataConfig attribute) (nemo_rl.data.DailyOmniEvalDataConfig attribute) (nemo_rl.data.GPQAEvalDataConfig attribute) (nemo_rl.data.interfaces.TaskDataSpec attribute) (nemo_rl.data.LocalMathEvalDataConfig attribute) (nemo_rl.data.MathEvalDataConfig attribute) (nemo_rl.data.MMAUEvalDataConfig attribute) (nemo_rl.data.MMLUEvalDataConfig attribute) (nemo_rl.data.MMLUProEvalDataConfig attribute) (nemo_rl.data.PreferenceDatasetConfig attribute) (nemo_rl.data.ResponseDatasetConfig attribute) prompt_id (nemo_rl.experience.rollout_recovery.PromptGroupRecoveryRecord attribute) (nemo_rl.experience.rollout_recovery.PromptGroupRecoveryState attribute) prompt_ids_field (nemo_rl.algorithms.single_controller_utils.config.AdvantageConfig attribute) prompt_idx (nemo_rl.experience.interfaces.PromptGroupRecord attribute) prompt_key (nemo_rl.data.PreferenceDatasetConfig attribute) prompt_payload (nemo_rl.experience.rollout_recovery.PromptGroupRecoveryRecord property) prompt_ref (nemo_rl.experience.rollout_recovery.PromptGroupRecoveryRecord attribute) (nemo_rl.experience.rollout_recovery.PromptGroupRecoveryState attribute) PromptGroupPhase (class in nemo_rl.experience.rollout_recovery) PromptGroupRecord (class in nemo_rl.experience.interfaces) PromptGroupRecoveryRecord (class in nemo_rl.experience.rollout_recovery) PromptGroupRecoveryState (class in nemo_rl.experience.rollout_recovery) PromptGroupSampler (class in nemo_rl.algorithms.async_utils.staleness_sampler) PromptRef (class in nemo_rl.experience.rollout_recovery) PromptRefState (class in nemo_rl.experience.rollout_recovery) protocol5_serialized_nbytes() (in module nemo_rl.utils.multimodal_payload_metrics) publish_megatron_conversion() (in module nemo_rl.models.megatron.community_import) put() (nemo_rl.utils.weight_transfer_stream._S3ObjectStore method) put_samples() (nemo_rl.data_plane.adapters.noop.NoOpDataPlaneClient method) (nemo_rl.data_plane.adapters.transfer_queue.TQDataPlaneClient method) (nemo_rl.data_plane.interfaces.DataPlaneClient method) (nemo_rl.data_plane.observability.MetricsDataPlaneClient method) PY_EXECUTABLES (class in nemo_rl.distributed.virtual_cluster) PytorchOptimizerConfig (class in nemo_rl.models.policy) Q q_lora_rank (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) q_weight (nemo_rl.models.megatron.draft.utils._PendingLayerWeights attribute) qk_head_dim (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) qk_pos_emb_head_dim (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) qkv_weight (nemo_rl.models.megatron.draft.utils._PendingLayerWeights attribute) qkv_weight_key (nemo_rl.models.megatron.draft.utils._EagleLayerLayout property) quant_batch_size (nemo_rl.models.policy.PolicyConfig attribute) quant_calib_data (nemo_rl.models.policy.PolicyConfig attribute) quant_calib_size (nemo_rl.models.policy.PolicyConfig attribute) quant_cfg (nemo_rl.models.generation.vllm.config.VllmConfig attribute) (nemo_rl.models.policy.PolicyConfig attribute) quant_sequence_length (nemo_rl.models.policy.PolicyConfig attribute) quantization_ignore_patterns (nemo_rl.models.generation.vllm.config.VllmSpecificArgs attribute) quantization_ignored_layer_kws (nemo_rl.models.generation.vllm.config.VllmSpecificArgs attribute) quantization_method_for_mode() (in module nemo_rl.modelopt.models.generation.vllm_modelopt) quantize_model() (in module nemo_rl.modelopt.models.policy.workers.utils) query_groups (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) qwen2() (in module nemo_rl.utils.flops_formulas) qwen3() (in module nemo_rl.utils.flops_formulas) R R3_MISSING_ROUTE_SENTINEL (in module nemo_rl.models.generation.vllm.utils) r3_routed_experts_actual_routes (nemo_rl.models.generation.interfaces.GenerationOutputSpec attribute) r3_routed_experts_expected_routes (nemo_rl.models.generation.interfaces.GenerationOutputSpec attribute) r3_routed_experts_missing_routes (nemo_rl.models.generation.interfaces.GenerationOutputSpec attribute) r3_trace_enabled() (in module nemo_rl.utils.r3_trace) r3_trace_stage() (in module nemo_rl.utils.r3_trace) r3_trace_verify_forward_enabled() (in module nemo_rl.utils.r3_trace) radio_force_cpe_eval_mode (nemo_rl.models.policy.MegatronConfig attribute) rail_link_layers() (in module nemo_rl.data_plane.adapters.transfer_queue_env) raise_if_exhausted() (nemo_rl.models.generation.fleet_health.GenerationFleetHealth method) rank (nemo_rl.models.policy.workers.checkpoint_engine.PolicyCheckpointEngineMixin attribute) ratio_clip_c (nemo_rl.algorithms.loss.loss_functions.ClippedPGLossConfig attribute) ratio_clip_max (nemo_rl.algorithms.loss.loss_functions.ClippedPGLossConfig attribute) ratio_clip_min (nemo_rl.algorithms.loss.loss_functions.ClippedPGLossConfig attribute) RawDataset (class in nemo_rl.data.datasets.raw_dataset) RawRewardAdvantageEstimator (class in nemo_rl.algorithms.advantage_estimator) RayGpuMonitorLogger (class in nemo_rl.utils.logger) RayVirtualCluster (class in nemo_rl.distributed.virtual_cluster) RayWorkerBuilder (class in nemo_rl.distributed.worker_groups) RayWorkerBuilder.IsolatedWorkerInitializer (class in nemo_rl.distributed.worker_groups) RayWorkerGroup (class in nemo_rl.distributed.worker_groups) rdma_devices() (in module nemo_rl.data_plane.adapters.transfer_queue) re_enable_float32_expert_bias() (nemo_rl.models.megatron.setup.MoEFloat16Module method) read_columns() (in module nemo_rl.data_plane.column_io) read_from_dataplane() (nemo_rl.data_plane.driver_mixin.TQDriverMixin method) read_message() (nemo_rl.utils.checkpoint_engines.nixl.NixlAgent method) ReadyFirstSampler (class in nemo_rl.algorithms.async_utils.staleness_sampler) ReadyFirstSamplerConfig (class in nemo_rl.algorithms.async_utils.staleness_sampler) real_quant (nemo_rl.models.generation.vllm.config.VllmConfig attribute) real_quant_export_cpu_offload (nemo_rl.models.generation.vllm.config.VllmConfig attribute) real_quant_ignore (nemo_rl.models.generation.vllm.config.VllmConfig attribute) reasoning_parser (nemo_rl.models.generation.dynamo.config.DynamoWorkerArgs attribute) (nemo_rl.models.generation.trtllm.config.TrtllmSpecificArgs attribute) reasoning_parser_plugin (nemo_rl.models.generation.vllm.config.VllmSpecificArgs attribute) rebalance_nd_tensor() (in module nemo_rl.distributed.collectives) rebuild_collective() (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) rebuild_cuda_tensor_from_ipc() (in module nemo_rl.models.policy.utils) rebuild_nccl_reshard_comm_group() (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) rebuild_teacher_full_logits_from_ipc() (in module nemo_rl.algorithms.x_token.loss_utils) receive_held_socket() (in module nemo_rl.distributed.held_port) receive_weight_batches() (nemo_rl.utils.checkpoint_engines.base.CheckpointEngine method) (nemo_rl.utils.checkpoint_engines.nixl.NIXLCheckpointEngine method) recompute_granularity (nemo_rl.models.policy.MegatronConfig attribute) recompute_kv_cache_after_weight_updates (nemo_rl.algorithms.grpo.AsyncGRPOConfig attribute) (nemo_rl.algorithms.ppo.AsyncPPOConfig attribute) (nemo_rl.algorithms.single_controller_utils.config.AsyncRLConfig attribute) (nemo_rl.models.generation.trtllm.config.TrtllmSpecificArgs attribute) recompute_modules (nemo_rl.models.policy.MegatronConfig attribute) reconcile_communicator() (nemo_rl.weight_sync.collective_weight_synchronizer.CollectiveWeightSynchronizer method) (nemo_rl.weight_sync.interfaces.WeightSynchronizer method) (nemo_rl.weight_sync.nccl_reshard_weight_synchronizer.NcclReshardWeightSynchronizer method) reconstruct_message_log() (in module nemo_rl.data.llm_message_utils) record() (nemo_rl.utils.timer.ThreadSafeTimer method) (nemo_rl.utils.timer.Timer method) record_actor_death() (nemo_rl.models.generation.fleet_health.GenerationFleetHealth method) record_data_failure() (nemo_rl.experience.rollout_manager.RolloutStats method) record_data_retry() (nemo_rl.experience.rollout_manager.RolloutStats method) record_gym_row_redispatch() (nemo_rl.experience.rollout_manager.RolloutStats method) record_infra_drop() (nemo_rl.experience.rollout_manager.RolloutStats method) record_load_failure() (nemo_rl.models.generation.vllm.vllm_backend._IPCWeightManifest method) record_loaded() (nemo_rl.models.generation.vllm.vllm_backend._IPCWeightManifest method) record_probe() (nemo_rl.models.generation.fleet_health.GenerationFleetHealth method) record_redispatch() (nemo_rl.experience.rollout_manager.RolloutStats method) record_to_train_batch() (in module nemo_rl.experience.payload) recovery_ledger (nemo_rl.experience.rollout_manager.RolloutManager property) recursive_merge_options() (in module nemo_rl.distributed.worker_group_utils) redact_argv() (in module nemo_rl.models.generation.dynamo.arguments) redact_environment() (in module nemo_rl.models.generation.dynamo.arguments) redispatches_by_reason (nemo_rl.experience.rollout_manager.RolloutStats attribute) reduce() (nemo_rl.utils.timer.ThreadSafeTimer method) (nemo_rl.utils.timer.Timer method) reduce_advantage_pump_metrics() (in module nemo_rl.algorithms.single_controller_utils.utils) RefCOCODataset (class in nemo_rl.data.datasets.response_datasets.refcoco) reference_logprobs (nemo_rl.models.policy.interfaces.ReferenceLogprobOutputSpec attribute) reference_logprobs_field (nemo_rl.algorithms.single_controller_utils.config.AdvantageConfig attribute) REFERENCE_POLICY (nemo_rl.telemetry.span_groups.RLSpanGroup attribute) reference_policy_kl_penalty (nemo_rl.algorithms.dpo.DPOConfig attribute) (nemo_rl.algorithms.loss.loss_functions.ClippedPGLossConfig attribute) (nemo_rl.algorithms.loss.loss_functions.DPOLossConfig attribute) reference_policy_kl_type (nemo_rl.algorithms.loss.loss_functions.ClippedPGLossConfig attribute) reference_policy_logprobs (nemo_rl.algorithms.loss.loss_functions.ClippedPGLossDataDict attribute) (nemo_rl.algorithms.loss.loss_functions.DPOLossDataDict attribute) ReferenceLogprobOutputSpec (class in nemo_rl.models.policy.interfaces) REFIT_ABORTED_TOKEN (in module nemo_rl.distributed.refit_watchdog) refit_backend (nemo_rl.models.generation.megatron.config.MCoreGenerationSpecificArgs attribute) refit_buffer_size_gb (nemo_rl.models.policy.PolicyConfig attribute) refit_cfg (nemo_rl.models.generation.vllm.config.VllmConfig attribute) REFIT_CONTEXT_LOST_TOKEN (in module nemo_rl.distributed.refit_watchdog) refit_http_session() (in module nemo_rl.utils.weight_transfer_http) refit_policy_generation() (in module nemo_rl.algorithms.grpo) refit_timeout_s (nemo_rl.algorithms.single_controller_utils.config.FleetHealthConfig attribute) refit_transport (nemo_rl.models.generation.vllm.config.VllmConfig attribute) refit_workers() (nemo_rl.models.generation.dynamo.managed_runtime.ManagedDynamoRuntime method) (nemo_rl.models.generation.dynamo.worker_pool.FixedDynamoWorkerPool method) RefitAborted RefitAbortWatchdog (class in nemo_rl.distributed.refit_watchdog) RefitBuilderInterface (class in nemo_rl.weight_sync.nccl_reshard_utils) RefitCtx (class in nemo_rl.weight_sync.nccl_reshard_utils) RefitMembership (class in nemo_rl.weight_sync.membership) register_draft_grad_norm_group() (in module nemo_rl.models.megatron.draft.utils) register_env() (in module nemo_rl.environments.utils) register_hooks() (nemo_rl.models.megatron.draft.hidden_capture.HiddenStateCapture method) register_nemo_modelopt_nvfp4() (in module nemo_rl.modelopt.models.generation.vllm_modelopt) register_omegaconf_resolvers() (in module nemo_rl.utils.config) register_partition() (nemo_rl.data_plane.adapters.noop.NoOpDataPlaneClient method) (nemo_rl.data_plane.adapters.transfer_queue.TQDataPlaneClient method) (nemo_rl.data_plane.interfaces.DataPlaneClient method) (nemo_rl.data_plane.observability.MetricsDataPlaneClient method) register_process_group() (nemo_rl.models.generation.dynamo.dynamo_worker.DynamoGpuReservation method) register_processor() (in module nemo_rl.data.processors) register_torchcodec_vllm_video_loader() (in module nemo_rl.models.generation.vllm.video_utils) ReinforcePlusPlusAdvantageEstimator (class in nemo_rl.algorithms.advantage_estimator) reject_unenforceable_refit_deadline() (in module nemo_rl.models.generation.interfaces) rejected_key (nemo_rl.data.PreferenceDatasetConfig attribute) release() (nemo_rl.models.generation.fleet_health.HealthyShardSelector method) release_after_refit (nemo_rl.models.generation.vllm.config.VllmCheckpointEnginePluginConfig attribute) (nemo_rl.models.generation.vllm.config.VllmNixlRefitConfig attribute) RELEASE_GRACE_S (in module nemo_rl.distributed.refit_watchdog) release_ipc_buffer() (nemo_rl.models.policy.lm_policy.Policy method) (nemo_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl method) release_within() (in module nemo_rl.distributed.refit_watchdog) remap_dataset_keys() (in module nemo_rl.data.llm_message_utils) remote_trace_context() (in module nemo_rl.telemetry.instrumentation) RemoteHeldPortReservation (class in nemo_rl.distributed.held_port) remove() (nemo_rl.algorithms.async_utils.replay_buffer.TQReplayBuffer method) remove_boxed() (in module nemo_rl.environments.dapo_math_verifier) remove_group() (nemo_rl.algorithms.async_utils.replay_buffer.TQReplayBuffer method) remove_old_checkpoints() (nemo_rl.utils.checkpoint.CheckpointManager method) remove_remote_agent() (nemo_rl.utils.checkpoint_engines.nixl.NixlAgent method) REMOVED_EXPRESSIONS (in module nemo_rl.environments.dapo_math_verifier) reorder_data() (nemo_rl.distributed.batched_data_dict.BatchedDataDict method) repeat (nemo_rl.data.AIMEEvalDataConfig attribute) repeat_interleave() (nemo_rl.distributed.batched_data_dict.BatchedDataDict method) repeated_batch_fields (nemo_rl.algorithms.single_controller_utils.config.AdvantageConfig attribute) replace_prefix_tokens() (in module nemo_rl.models.generation.openai_server_utils) REPLACEMENT_RESERVE_FILENAME (in module nemo_rl.algorithms.async_utils.replay_buffer) replacement_reserve_prompts (nemo_rl.algorithms.single_controller_utils.config.RolloutFailureConfig attribute) REPLAY_BUFFER_METADATA_FILENAME (in module nemo_rl.algorithms.async_utils.replay_buffer) REPLAY_BUFFER_METADATA_SCHEMA_VERSION (in module nemo_rl.algorithms.async_utils.replay_buffer) REPLAY_BUFFER_METADATA_STORAGE (in module nemo_rl.algorithms.async_utils.replay_buffer) replay_group_count (nemo_rl.algorithms.async_utils.replay_buffer.DataPlaneCheckpointMetadata attribute) replay_manifest_digest (nemo_rl.algorithms.async_utils.replay_buffer.DataPlaneCheckpointMetadata attribute) replay_manifest_digest() (in module nemo_rl.algorithms.async_utils.replay_buffer) replay_metadata_schema_version (nemo_rl.algorithms.async_utils.replay_buffer.DataPlaneCheckpointMetadata attribute) ReplayBuffer (class in nemo_rl.algorithms.async_utils.replay_buffer) ReplayBufferImpl (class in nemo_rl.algorithms.async_utils.replay_buffer) ReplayBufferProtocol (class in nemo_rl.algorithms.async_utils.interfaces) REPLICATED_AXES (in module nemo_rl.distributed.named_sharding) report_device_id() (nemo_rl.models.generation.trtllm.trtllm_backend.NcclExtension method) (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) (nemo_rl.models.generation.vllm.vllm_worker.VllmGenerationWorkerImpl method) (nemo_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker method) report_device_id_async() (nemo_rl.models.generation.trtllm.trtllm_worker_async.TrtllmAsyncGenerationWorkerImpl method) (nemo_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) report_dp_openai_server_base_url() (nemo_rl.models.generation.megatron.megatron_worker.MegatronGenerationMixin method) (nemo_rl.models.generation.trtllm.trtllm_worker_async.TrtllmAsyncGenerationWorkerImpl method) (nemo_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) report_failure() (nemo_rl.models.generation.fleet_health.GenerationFleetHealth method) report_node_hostname() (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) report_node_ip_and_gpu_id() (nemo_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker method) report_refit() (nemo_rl.models.generation.fleet_health.GenerationFleetHealth method) report_refit_server_base_url() (nemo_rl.models.generation.vllm.vllm_sparse_refit.VllmSparseRefitReceiver method) (nemo_rl.models.generation.vllm.vllm_worker.BaseVllmGenerationWorker method) report_success() (nemo_rl.models.generation.fleet_health.GenerationFleetHealth method) request_timeout_s (nemo_rl.models.generation.dynamo.config.DynamoCfg attribute) (nemo_rl.models.generation.vllm.config.VllmSparseRefitConfig attribute) require_complete() (nemo_rl.models.generation.vllm.vllm_backend._IPCWeightManifest method) require_live() (nemo_rl.algorithms.async_utils.replay_buffer.DataPlaneMutationCut method) require_routed_experts (nemo_rl.environments.nemo_gym.NemoGymConfig attribute) required_buffer_capacity() (nemo_rl.algorithms.async_utils.staleness_sampler._GatedSampler method) (nemo_rl.algorithms.async_utils.staleness_sampler.BaseSampler method) (nemo_rl.algorithms.async_utils.staleness_sampler.InOrderSampler method) (nemo_rl.algorithms.async_utils.staleness_sampler.PromptGroupSampler method) required_buffer_capacity_for_config() (in module nemo_rl.algorithms.async_utils.staleness_sampler) requires_kv_scale_sync (nemo_rl.models.generation.interfaces.GenerationInterface property) (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration property) reserve() (nemo_rl.algorithms.async_utils.replay_buffer.TQReplayBuffer method) reserve_group() (nemo_rl.experience.rollout_recovery.RolloutRecoveryLedger method) reserve_http_server_address() (nemo_rl.models.generation.megatron.megatron_generation.MegatronGeneration class method) reserve_prompt_group() (nemo_rl.experience.rollout_manager.RolloutManager method) reserve_teacher_clusters() (in module nemo_rl.algorithms.opd) RESERVED (nemo_rl.experience.rollout_recovery.PromptGroupPhase attribute) reset() (nemo_rl.data.packing.metrics.PackingMetrics method) (nemo_rl.utils.flops_tracker.FLOPTracker method) (nemo_rl.utils.timer.ThreadSafeTimer method) (nemo_rl.utils.timer.Timer method) reset_encoder_cache_after_weight_update (nemo_rl.models.generation.vllm.config.VllmSpecificArgs attribute) reset_metrics() (nemo_rl.data.packing.algorithms.SequencePacker method) reset_peak_memory_stats() (nemo_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker method) reset_prefix_cache() (nemo_rl.models.generation.vllm.vllm_worker.VllmGenerationWorkerImpl method) reset_prefix_cache_async() (nemo_rl.models.generation.trtllm.trtllm_worker_async.TrtllmAsyncGenerationWorkerImpl method) (nemo_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) reshard_after_forward (nemo_rl.models.policy.MoEParallelizerOptions attribute) resolve_collective_rpc_result() (in module nemo_rl.models.generation.vllm.collective_rpc) resolve_data_parallel_local_rank() (in module nemo_rl.models.generation.vllm.worker_utils) resolve_distributed_executor_backend() (in module nemo_rl.models.generation.vllm.worker_utils) resolve_external_dataset_class() (in module nemo_rl.data.datasets.utils) resolve_generation_class() (in module nemo_rl.models.generation) resolve_generation_worker_cls() (in module nemo_rl.models.generation.vllm.utils) resolve_model_class() (in module nemo_rl.models.policy.utils) resolve_nixl_backend_kwargs() (in module nemo_rl.utils.checkpoint_engines.nixl) resolve_nvfp4_real_quant_mode() (in module nemo_rl.modelopt.utils) resolve_path() (in module nemo_rl.utils.config) resolve_policy_worker_cls() (in module nemo_rl.models.policy.utils) resolve_quant_cfg() (in module nemo_rl.modelopt.utils) resolve_reference_aliases() (in module nemo_rl.algorithms.opd) resolve_reward_penalty_config() (in module nemo_rl.experience.rollouts) resolve_rollout_rank() (in module nemo_rl.models.generation.vllm.checkpoint_engine) resolve_routed_experts_dtype() (in module nemo_rl.models.generation.interfaces) resolve_routed_experts_dtype_name_for_model() (in module nemo_rl.models.generation.interfaces) resolve_to_image() (in module nemo_rl.data.multimodal_utils) resolve_torch_dtype() (in module nemo_rl.models.generation.megatron.utils) resolve_visible_gpu_id() (in module nemo_rl.distributed.numa_utils) resolve_vllm_video_config() (in module nemo_rl.models.generation.vllm.config) resolved_warmup_generation_lead_steps (nemo_rl.algorithms.ppo.AsyncPPOConfig property) ResourceInsufficientError resources (nemo_rl.environments.reward_model_environment.RewardModelEnvironmentConfig attribute) (nemo_rl.models.generation.interfaces.ColocationConfig attribute) ResourcesConfig (class in nemo_rl.models.generation.interfaces) response_from_nested() (in module nemo_rl.data_plane.codec) ResponseDataset (class in nemo_rl.data.datasets.response_datasets.response_dataset) ResponseDatasetConfig (class in nemo_rl.data) restart_attempts (nemo_rl.models.generation.fleet_health.ShardHealth attribute) RESTARTING (nemo_rl.models.generation.fleet_health.ShardState attribute) restore_dispatch_index() (nemo_rl.algorithms.async_utils.staleness_sampler.BaseSampler method) (nemo_rl.algorithms.async_utils.staleness_sampler.PromptGroupSampler method) restore_refit_info_placements() (in module nemo_rl.weight_sync.nccl_reshard_utils) results (nemo_rl.experience.rollouts._CompletedNemoGymGroup attribute) resume() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector method) resume_after_refit() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector method) (nemo_rl.models.generation.megatron.megatron_generation.MegatronGeneration method) (nemo_rl.models.generation.megatron.megatron_worker.MegatronGenerationRefitMixin method) RESUME_BASE_ORDINAL_KEY (in module nemo_rl.experience.interfaces) resume_generation_after_refit() (nemo_rl.models.generation.interfaces.GenerationInterface method) (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) resume_generation_async() (nemo_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) resume_inference_weights() (in module nemo_rl.models.megatron.memory_saver) RETAINED_TASK_INDICES_KEY (in module nemo_rl.experience.interfaces) retire() (nemo_rl.models.generation.fleet_health.GenerationFleetHealth method) RETIRED (nemo_rl.models.generation.fleet_health.ShardState attribute) return_from_workers (nemo_rl.distributed.worker_groups.MultiWorkerFuture attribute) return_model_config() (nemo_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) (nemo_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl method) return_state_dict() (nemo_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) (nemo_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl method) returns_field (nemo_rl.algorithms.single_controller_utils.config.AdvantageConfig attribute) reuse_registered_buffers (nemo_rl.data_plane.interfaces.MooncakeCpuConfig attribute) reverse_kl (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute) reward (nemo_rl.experience.interfaces.Completion attribute) REWARD (nemo_rl.telemetry.span_groups.RLSpanGroup attribute) reward_field (nemo_rl.algorithms.single_controller_utils.config.AdvantageConfig attribute) reward_functions (nemo_rl.environments.vlm_environment.VLMEnvConfig attribute) reward_model_cfg (nemo_rl.models.policy.PolicyConfig attribute) (nemo_rl.models.value.config.ValueConfig attribute) reward_model_type (nemo_rl.models.policy.RewardModelConfig attribute) REWARD_NAMES (nemo_rl.environments.math_environment.HFMultiRewardVerifyWorker attribute) reward_penalties (nemo_rl.algorithms.grpo.MasterConfig attribute) reward_scaling (nemo_rl.algorithms.grpo.GRPOConfig attribute) (nemo_rl.algorithms.ppo.PPOConfig attribute) reward_shaping (nemo_rl.algorithms.grpo.GRPOConfig attribute) (nemo_rl.algorithms.ppo.PPOConfig attribute) reward_weights (nemo_rl.algorithms.advantage_estimator.AdvEstimatorConfig attribute) RewardModelConfig (class in nemo_rl.models.policy) RewardModelEnvironment (class in nemo_rl.environments.reward_model_environment) RewardModelEnvironmentConfig (class in nemo_rl.environments.reward_model_environment) RewardPenaltyConfig (class in nemo_rl.algorithms.grpo) RewardPenaltyTokenIdsConfig (class in nemo_rl.algorithms.grpo) rewards (nemo_rl.environments.interfaces.EnvironmentReturn attribute) rewards_chosen_mean (nemo_rl.algorithms.dpo.DPOValMetrics attribute) (nemo_rl.algorithms.rm.RMValMetrics attribute) rewards_low (nemo_rl.experience.rollouts._EffortShapingMetrics attribute) rewards_rejected_mean (nemo_rl.algorithms.dpo.DPOValMetrics attribute) (nemo_rl.algorithms.rm.RMValMetrics attribute) RewardScalingConfig (class in nemo_rl.algorithms.grpo) RewardShapingConfig (class in nemo_rl.algorithms.reward_functions) right_shift_values() (in module nemo_rl.models.value.workers.dtensor_value_worker_v2) RightShiftLossWrapper (class in nemo_rl.models.value.workers.dtensor_value_worker_v2) RL_BUCKET_ATTR (in module nemo_rl.telemetry.instrumentation) rl_collate_fn() (in module nemo_rl.data.collate_fn) RL_EFFICIENCY_CATEGORY_ATTR (in module nemo_rl.telemetry.instrumentation) RL_EFFICIENCY_MEASUREMENT_ATTR (in module nemo_rl.telemetry.metrics) RL_EFFICIENCY_PCT_METRIC (in module nemo_rl.telemetry.metrics) RL_EFFICIENCY_SECONDS_METRIC (in module nemo_rl.telemetry.metrics) RL_EFFICIENCY_WINDOW_ATTR (in module nemo_rl.telemetry.metrics) RLSpanGroup (class in nemo_rl.telemetry.span_groups) rm (nemo_rl.algorithms.rm.MasterConfig attribute) rm_train() (in module nemo_rl.algorithms.rm) RMConfig (class in nemo_rl.algorithms.rm) rms_norm (nemo_rl.models.policy.AutomodelBackendConfig attribute) RMSaveState (class in nemo_rl.algorithms.rm) RMValMetrics (class in nemo_rl.algorithms.rm) ROLLOUT (nemo_rl.telemetry.span_groups.RLSpanGroup attribute) rollout_failure (nemo_rl.algorithms.single_controller_utils.config.AsyncRLConfig attribute) rollout_manager (nemo_rl.algorithms.single_controller_utils.setup.SingleControllerActorArgs attribute) rollout_metrics (nemo_rl.experience.interfaces.PromptGroupRecord attribute) (nemo_rl.experience.rollouts.NemoGymRolloutResult attribute) (nemo_rl.experience.rollouts.RolloutGroupResult attribute) rollout_recovery_group_count (nemo_rl.algorithms.async_utils.replay_buffer.DataPlaneCheckpointMetadata attribute) rollout_recovery_payload_sha256 (nemo_rl.algorithms.async_utils.replay_buffer.DataPlaneCheckpointMetadata attribute) ROLLOUT_RECOVERY_SCHEMA_VERSION (in module nemo_rl.experience.rollout_recovery) rollout_recovery_schema_version (nemo_rl.algorithms.async_utils.replay_buffer.DataPlaneCheckpointMetadata attribute) ROLLOUT_RECOVERY_STATE_FILENAME (in module nemo_rl.experience.rollout_recovery) rollout_s (nemo_rl.experience.rollout_manager.RolloutTimeouts attribute) rollout_timeout_s (nemo_rl.algorithms.single_controller_utils.config.NemoGymRolloutFTConfig attribute) rollout_to_tq() (nemo_rl.experience.sync_rollout_actor.SyncRolloutActor method) RolloutDataFailure RolloutFailure RolloutFailureConfig (class in nemo_rl.algorithms.single_controller_utils.config) RolloutGroupResult (class in nemo_rl.experience.rollouts) RolloutInfraFailure RolloutManager (class in nemo_rl.experience.rollout_manager) RolloutOutcome (class in nemo_rl.experience.rollout_manager) RolloutRecoveryLedger (class in nemo_rl.experience.rollout_recovery) RolloutRecoveryLedgerState (class in nemo_rl.experience.rollout_recovery) RolloutRecoveryState (class in nemo_rl.experience.rollout_recovery) RolloutRedispatchExhausted RolloutRetryPolicy (class in nemo_rl.experience.rollout_manager) RolloutStall RolloutStats (class in nemo_rl.experience.rollout_manager) RolloutTimeout RolloutTimeouts (class in nemo_rl.experience.rollout_manager) RotaryEmbedParallel (class in nemo_rl.models.dtensor.parallelize) round_up() (in module nemo_rl.data_plane.column_io) routed_experts (nemo_rl.models.generation.interfaces.GenerationOutputSpec attribute) (nemo_rl.models.megatron.data.ProcessedInputs attribute) (nemo_rl.models.megatron.data.ProcessedMicrobatch attribute) routed_experts_cp_sharded (nemo_rl.models.megatron.data.ProcessedInputs attribute) (nemo_rl.models.megatron.data.ProcessedMicrobatch attribute) routed_experts_dtype (nemo_rl.environments.nemo_gym.NemoGymConfig attribute) ROUTED_EXPERTS_FALLBACK_DTYPE (in module nemo_rl.models.generation.interfaces) ROUTED_EXPERTS_FIELD (in module nemo_rl.data_plane.schema) ROUTED_EXPERTS_MISSING_ROUTE_SENTINEL (in module nemo_rl.models.generation.interfaces) router_mode (nemo_rl.models.generation.dynamo.config.DynamoFrontendArgs attribute) router_replay (nemo_rl.models.policy.PolicyConfig attribute) router_replay_enabled() (in module nemo_rl.models.megatron.router_replay) router_reset_states (nemo_rl.models.generation.dynamo.config.DynamoFrontendArgs attribute) RouterReplayConfig (class in nemo_rl.models.policy) RouterReplayConfigDisabled (class in nemo_rl.models.policy) ROW_PARALLEL_SUFFIXES (in module nemo_rl.weight_sync.nccl_reshard_utils) rows (nemo_rl.data_plane.adapters.noop._Partition attribute) (nemo_rl.experience.rollouts._CompletedNemoGymGroup attribute) run() (nemo_rl.algorithms.single_controller.SingleControllerActor method) run_all_workers_multiple_data() (nemo_rl.distributed.worker_groups.RayWorkerGroup method) (nemo_rl.models.policy.lm_policy.Policy method) run_all_workers_sharded_data() (nemo_rl.distributed.worker_groups.RayWorkerGroup method) run_all_workers_single_data() (nemo_rl.distributed.worker_groups.RayWorkerGroup method) (nemo_rl.models.policy.lm_policy.Policy method) run_async_multi_turn_rollout() (in module nemo_rl.experience.rollouts) run_async_multi_turn_rollout_groups() (in module nemo_rl.experience.rollouts) run_async_nemo_gym_rollout() (in module nemo_rl.experience.rollouts) run_env_eval() (in module nemo_rl.evals.eval) run_id (nemo_rl.utils.logger.MLflowConfig attribute) run_multi_turn_rollout() (in module nemo_rl.experience.rollouts) run_name (nemo_rl.utils.logger.MLflowConfig attribute) run_nemo_gym_rollout_sync() (in module nemo_rl.experience.rollouts) run_rollout() (nemo_rl.experience.rollout_manager.AsyncNemoGymRolloutImpl method) (nemo_rl.experience.rollout_manager.AsyncRolloutImpl method) (nemo_rl.experience.rollout_manager.RolloutManager method) run_rollouts() (nemo_rl.environments.nemo_gym.NemoGym method) run_sample_multi_turn_rollout() (in module nemo_rl.experience.rollouts) run_single_worker_single_data() (nemo_rl.distributed.worker_groups.RayWorkerGroup method) RUN_WINDOW (in module nemo_rl.telemetry.metrics) RUN_WINDOW_WALL_CLOCK_CATEGORIES (in module nemo_rl.algorithms.utils) runtime_prompt_payload (nemo_rl.experience.rollout_recovery.PromptGroupRecoveryRecord attribute) RuntimeConfig (class in nemo_rl.models.automodel.config) (class in nemo_rl.models.megatron.config) S s3_bucket (nemo_rl.models.generation.vllm.config.VllmRefitStorageConfig attribute) s3_prefix (nemo_rl.models.generation.vllm.config.VllmRefitStorageConfig attribute) s3_region (nemo_rl.models.generation.vllm.config.VllmRefitStorageConfig attribute) s_end (nemo_rl.algorithms.x_token.token_aligner.AlignmentPair attribute) s_start (nemo_rl.algorithms.x_token.token_aligner.AlignmentPair attribute) s_tokens (nemo_rl.algorithms.x_token.token_aligner.AlignmentPair attribute) safe_import() (nemo_rl.environments.code_environment.CodeExecutionWorker method) safe_open() (nemo_rl.environments.code_environment.CodeExecutionWorker method) sample() (nemo_rl.algorithms.async_utils.interfaces.ReplayBufferProtocol method) (nemo_rl.algorithms.async_utils.replay_buffer.ReplayBufferImpl method) sample_freshest_first (nemo_rl.algorithms.async_utils.staleness_sampler.WindowedSamplerConfig attribute) sample_id (nemo_rl.experience.rollout_recovery.PromptRef attribute) (nemo_rl.experience.rollout_recovery.PromptRefState attribute) sample_ids (nemo_rl.data_plane.interfaces.KVBatchMeta attribute) SAMPLE_MASK (in module nemo_rl.data_plane.schema) sample_mask (nemo_rl.algorithms.loss.loss_functions.ClippedPGLossDataDict attribute) (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossDataDict attribute) (nemo_rl.algorithms.loss.loss_functions.DistillationLossDataDict attribute) (nemo_rl.algorithms.loss.loss_functions.DPOLossDataDict attribute) (nemo_rl.algorithms.loss.loss_functions.DraftCrossEntropyLossDataDict attribute) (nemo_rl.algorithms.loss.loss_functions.PreferenceLossDataDict attribute) (nemo_rl.algorithms.x_token.loss_utils.LocalizedAlignment attribute) sample_mask_field (nemo_rl.algorithms.single_controller_utils.config.AdvantageConfig attribute) sampler (nemo_rl.algorithms.single_controller_utils.config.AsyncRLConfig attribute) sampler_dispatch_index (nemo_rl.algorithms.grpo.GRPOSaveState attribute) sampler_enabled (nemo_rl.telemetry.config.TelemetryConfig attribute) sampler_name (nemo_rl.algorithms.async_utils.replay_buffer.DataPlaneCheckpointMetadata attribute) (nemo_rl.algorithms.grpo.GRPOSaveState attribute) sampler_stamps_target_steps (nemo_rl.experience.rollout_recovery.ParsedRolloutRecoveryState attribute) (nemo_rl.experience.rollout_recovery.RolloutRecoveryState attribute) sampler_supports_buffer_checkpoint() (in module nemo_rl.algorithms.async_utils.staleness_sampler) SamplerConfig (in module nemo_rl.algorithms.async_utils.staleness_sampler) sampling_params (nemo_rl.models.automodel.config.RuntimeConfig attribute) (nemo_rl.models.megatron.config.RuntimeConfig attribute) sampling_style (nemo_rl.models.generation.vllm.config.VllmVideoConfig attribute) sanitize() (nemo_rl.environments.code_environment.CodeExecutionWorker method) save_checkpoint() (in module nemo_rl.utils.native_checkpoint) (nemo_rl.data_plane.adapters.noop.NoOpDataPlaneClient method) (nemo_rl.data_plane.adapters.transfer_queue.TQDataPlaneClient method) (nemo_rl.data_plane.interfaces.DataPlaneClient method) (nemo_rl.data_plane.observability.MetricsDataPlaneClient method) (nemo_rl.modelopt.models.policy.workers.megatron_quant_policy_worker.MegatronQuantPolicyWorker method) (nemo_rl.models.automodel.checkpoint.AutomodelCheckpointManager method) (nemo_rl.models.policy.interfaces.PolicyInterface method) (nemo_rl.models.policy.lm_policy.Policy method) (nemo_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) (nemo_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl method) (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) (nemo_rl.models.value.interfaces.ValueInterface method) (nemo_rl.models.value.lm_value.Value method) (nemo_rl.models.value.workers.dtensor_value_worker_v2.DTensorValueWorkerV2Impl method) (nemo_rl.models.value.workers.megatron_value_worker.MegatronValueWorkerImpl method) save_consolidated (nemo_rl.utils.checkpoint.CheckpointingConfig attribute) save_data_plane (nemo_rl.utils.checkpoint.CheckpointingConfig attribute) save_optimizer (nemo_rl.utils.checkpoint.CheckpointingConfig attribute) save_path (nemo_rl.evals.eval.EvalConfig attribute) save_period (nemo_rl.utils.checkpoint.CheckpointingConfig attribute) save_s (nemo_rl.models.generation.vllm.vllm_sparse_refit._StagedSparsePayload attribute) save_state (nemo_rl.algorithms.single_controller_utils.setup.SingleControllerActorArgs attribute) save_to_path() (nemo_rl.algorithms.async_utils.interfaces.ReplayBufferProtocol method) (nemo_rl.algorithms.async_utils.replay_buffer.ReplayBufferImpl method) save_tokenizer_on_rank0() (in module nemo_rl.utils.native_checkpoint) saved_capacity (nemo_rl.algorithms.async_utils.replay_buffer.TQReplayMetadataState attribute) SC_ROLLOUT_SCHEMA_FIELDS (in module nemo_rl.data_plane.schema) scale (nemo_rl.algorithms.loss.loss_functions.MseValueLossConfig attribute) scale_rewards() (in module nemo_rl.algorithms.grpo) scheduler (nemo_rl.models.automodel.config.ModelAndOptimizerState attribute) (nemo_rl.models.megatron.config.ModelAndOptimizerState attribute) (nemo_rl.models.policy.MegatronConfig attribute) (nemo_rl.models.policy.PolicyConfig attribute) (nemo_rl.models.value.config.ValueConfig attribute) SchedulerMilestones (in module nemo_rl.models.policy) schema_version (nemo_rl.algorithms.async_utils.replay_buffer.TQReplayMetadataState attribute) (nemo_rl.experience.rollout_recovery.RolloutRecoveryLedgerState attribute) score() (nemo_rl.models.policy.lm_policy.Policy method) (nemo_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) (nemo_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl method) ScoreOutputSpec (class in nemo_rl.models.policy.interfaces) ScorePostProcessor (class in nemo_rl.models.automodel.train) scores (nemo_rl.models.policy.interfaces.ScoreOutputSpec attribute) seed (nemo_rl.algorithms.distillation.DistillationConfig attribute) (nemo_rl.algorithms.dpo.DPOConfig attribute) (nemo_rl.algorithms.grpo.GRPOConfig attribute) (nemo_rl.algorithms.ppo.PPOConfig attribute) (nemo_rl.algorithms.rm.RMConfig attribute) (nemo_rl.algorithms.sft.SFTConfig attribute) (nemo_rl.algorithms.xtoken_off_policy_distillation.OffPolicyDistillationConfig attribute) (nemo_rl.data.ResponseDatasetConfig attribute) (nemo_rl.evals.eval.EvalConfig attribute) segment_size (nemo_rl.distributed.virtual_cluster.ClusterConfig attribute) select() (nemo_rl.algorithms.async_utils.staleness_sampler.BaseSampler method) (nemo_rl.algorithms.async_utils.staleness_sampler.InOrderSampler method) (nemo_rl.algorithms.async_utils.staleness_sampler.PromptGroupSampler method) (nemo_rl.algorithms.async_utils.staleness_sampler.ReadyFirstSampler method) (nemo_rl.algorithms.async_utils.staleness_sampler.WeightFifoSampler method) (nemo_rl.algorithms.async_utils.staleness_sampler.WindowedSampler method) select_free_port() (nemo_rl.models.generation.dynamo.dynamo_worker.DynamoGpuReservation method) select_hf_weight_for_vllm_target() (in module nemo_rl.models.generation.vllm.refit_layout) select_indices() (nemo_rl.distributed.batched_data_dict.BatchedDataDict method) select_segment_nodes() (in module nemo_rl.distributed.virtual_cluster) select_teacher_topk_indices() (in module nemo_rl.algorithms.x_token.loss_utils) selection (nemo_rl.algorithms.single_controller_utils.config.FleetHealthConfig attribute) send (nemo_rl.utils.weight_transfer_stream.SparseRefitTransport attribute) send_hf_buckets_via_ipc_actor_impl() (in module nemo_rl.models.policy.utils) send_message() (nemo_rl.utils.checkpoint_engines.nixl.NixlAgent method) send_payload() (nemo_rl.utils.weight_transfer_zmq.ZmqSparseRefitClient method) send_weights() (nemo_rl.utils.checkpoint_engines.base.CheckpointEngine method) (nemo_rl.utils.checkpoint_engines.nixl.NIXLCheckpointEngine method) send_weights_via_checkpoint_engine() (nemo_rl.models.policy.workers.checkpoint_engine.PolicyCheckpointEngineMixin method) sender_spec (nemo_rl.models.generation.dynamo.refit.DynamoRefitChannel property) seq_len (nemo_rl.models.automodel.data.ProcessedInputs attribute) seq_logprob_error_threshold (nemo_rl.algorithms.grpo.GRPOConfig attribute) (nemo_rl.algorithms.ppo.PPOConfig attribute) sequence_length_pad_multiple (nemo_rl.distributed.batched_data_dict.SequencePackingArgs attribute) sequence_length_round (nemo_rl.distributed.batched_data_dict.DynamicBatchingArgs attribute) (nemo_rl.models.policy.DynamicBatchingConfig attribute) sequence_lengths (nemo_rl.data_plane.interfaces.KVBatchMeta attribute) SEQUENCE_LEVEL (nemo_rl.algorithms.loss.interfaces.LossType attribute) sequence_level_importance_ratios (nemo_rl.algorithms.loss.loss_functions.ClippedPGLossConfig attribute) sequence_packing (nemo_rl.environments.reward_model_environment.RewardModelEnvironmentConfig attribute) (nemo_rl.models.policy.PolicyConfig attribute) (nemo_rl.models.value.config.ValueConfig attribute) sequence_parallel (nemo_rl.models.policy.DTensorConfig attribute) (nemo_rl.models.policy.MegatronConfig attribute) SequencePacker (class in nemo_rl.data.packing.algorithms) SequencePackingArgs (class in nemo_rl.distributed.batched_data_dict) SequencePackingConfig (class in nemo_rl.models.policy) SequencePackingConfigDisabled (class in nemo_rl.models.policy) SequencePackingFusionLossWrapper (class in nemo_rl.algorithms.loss.wrapper) SequencePackingLossWrapper (class in nemo_rl.algorithms.loss.wrapper) SEQUENCES (nemo_rl.algorithms.loss.interfaces.MetricNormalizer attribute) serve_in_background() (nemo_rl.models.generation.generation_router.GenerationRouterImpl method) service_name (nemo_rl.telemetry.config.TelemetryConfig attribute) serving_base_urls() (nemo_rl.models.generation.fleet_health.GenerationFleetHealth method) serving_shards() (nemo_rl.models.generation.fleet_health.GenerationFleetHealth method) set_data_plane_checkpoint_barrier() (nemo_rl.algorithms.async_utils.replay_buffer.TQReplayBuffer method) set_dispatch_index() (nemo_rl.algorithms.async_utils.staleness_sampler.BaseSampler method) (nemo_rl.algorithms.async_utils.staleness_sampler.PromptGroupSampler method) set_gate_window() (nemo_rl.algorithms.async_utils.staleness_sampler._GatedSampler method) set_generation_window() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector method) set_post_write_enricher() (nemo_rl.algorithms.async_utils.replay_buffer.TQReplayBuffer method) set_processor() (nemo_rl.data.datasets.raw_dataset.RawDataset method) set_records() (nemo_rl.data.dataloader.MultipleDataloaderWrapper method) set_refit_membership() (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) set_rollout_num_gpus_per_engine() (nemo_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl method) set_router_replay_backward() (in module nemo_rl.models.megatron.router_replay) set_router_replay_forward() (in module nemo_rl.models.megatron.router_replay) set_seed() (in module nemo_rl.algorithms.utils) set_serving_backends() (nemo_rl.models.generation.generation_router.GenerationRouterImpl method) set_task_spec() (nemo_rl.data.datasets.raw_dataset.RawDataset method) set_tokenizer() (nemo_rl.environments.nemo_gym.NemoGym method) set_weight_version() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector method) (nemo_rl.experience.rollout_manager.RolloutManager method) set_worker_hostnames() (nemo_rl.models.generation.vllm.vllm_sparse_refit.VllmSparseRefitReceiver method) setup() (in module nemo_rl.algorithms.distillation) (in module nemo_rl.algorithms.dpo) (in module nemo_rl.algorithms.grpo) (in module nemo_rl.algorithms.ppo) (in module nemo_rl.algorithms.rm) (in module nemo_rl.algorithms.sft) (in module nemo_rl.algorithms.xtoken_off_policy_distillation) (in module nemo_rl.evals.eval) setup_api_server() (nemo_rl.models.generation.vllm.vllm_sparse_refit.VllmSparseRefitReceiver method) setup_data_plane() (nemo_rl.data_plane.worker_mixin.TQWorkerMixin method) (nemo_rl.models.policy.teacher_worker_group.TeacherWorkerGroup method) setup_distributed() (in module nemo_rl.models.automodel.setup) (in module nemo_rl.models.megatron.setup) setup_model_and_optimizer() (in module nemo_rl.models.automodel.setup) (in module nemo_rl.models.megatron.setup) setup_model_config() (in module nemo_rl.models.megatron.setup) setup_nemo_gym_config() (in module nemo_rl.environments.nemo_gym) setup_preference_data() (in module nemo_rl.data.utils) setup_reference_model_state() (in module nemo_rl.models.automodel.setup) (in module nemo_rl.models.megatron.setup) setup_response_data() (in module nemo_rl.data.utils) setup_single_controller() (in module nemo_rl.algorithms.single_controller_utils.setup) SetupTimingMetrics (class in nemo_rl.algorithms.metric_utils) sft (nemo_rl.algorithms.sft.MasterConfig attribute) sft_average_log_probs (nemo_rl.algorithms.dpo.DPOConfig attribute) (nemo_rl.algorithms.loss.loss_functions.DPOLossConfig attribute) sft_loss (nemo_rl.algorithms.dpo.DPOValMetrics attribute) sft_loss_weight (nemo_rl.algorithms.dpo.DPOConfig attribute) (nemo_rl.algorithms.loss.loss_functions.DPOLossConfig attribute) sft_processor() (in module nemo_rl.data.processors) sft_train() (in module nemo_rl.algorithms.sft) SFTConfig (class in nemo_rl.algorithms.sft) SFTSaveState (class in nemo_rl.algorithms.sft) sgd_momentum (nemo_rl.models.policy.MegatronOptimizerConfig attribute) SGLANG (nemo_rl.distributed.virtual_cluster.PY_EXECUTABLES attribute) SGLANG_BACKEND (in module nemo_rl.models.generation.constants) SGLANG_EXECUTABLE (in module nemo_rl.distributed.ray_actor_environment_registry) SGLangColocatedWeightSynchronizer (class in nemo_rl.weight_sync.sglang_weight_synchronizer) SGLangDisaggregatedWeightSynchronizer (class in nemo_rl.weight_sync.sglang_weight_synchronizer) shape (nemo_rl.distributed.named_sharding.NamedSharding property) (nemo_rl.utils.checkpoint_engines.base.TensorMeta attribute) shard_by_batch_size() (nemo_rl.distributed.batched_data_dict.BatchedDataDict method) shard_count (nemo_rl.models.generation.fleet_health.GenerationFleetHealth property) shard_expert_weights (nemo_rl.models.generation.vllm.config.VllmNixlRefitConfig attribute) (nemo_rl.utils.checkpoint_engines.base.CheckpointEngine attribute) shard_for_base_url() (nemo_rl.models.generation.fleet_health.GenerationFleetHealth method) shard_id (nemo_rl.models.generation.vllm.refit_layout.HfExpertWeight attribute) shard_meta_for_dp() (in module nemo_rl.data_plane.preshard) shard_prefixes (nemo_rl.weight_sync.membership.RefitMembership attribute) sharded_state_dict() (nemo_rl.models.megatron.draft.eagle.EagleModel method) ShardHealth (class in nemo_rl.models.generation.fleet_health) ShardState (class in nemo_rl.models.generation.fleet_health) should_abort_inflight() (nemo_rl.algorithms.async_utils.staleness_sampler.BaseSampler method) (nemo_rl.algorithms.async_utils.staleness_sampler.PromptGroupSampler method) (nemo_rl.algorithms.async_utils.staleness_sampler.WindowedSampler method) should_log_nemo_gym_full_result_tables() (in module nemo_rl.utils.logger) should_mask_flagged_samples() (in module nemo_rl.experience.rollouts) should_use_async_rollouts() (in module nemo_rl.models.generation.interfaces) should_use_nemo_gym() (in module nemo_rl.environments.nemo_gym) shuffle (nemo_rl.data.DataConfig attribute) shutdown() (nemo_rl.distributed.virtual_cluster.RayVirtualCluster method) (nemo_rl.distributed.worker_groups.RayWorkerGroup method) (nemo_rl.environments.code_environment.CodeEnvironment method) (nemo_rl.environments.code_jaccard_environment.CodeJaccardEnvironment method) (nemo_rl.environments.math_environment.BaseMathEnvironment method) (nemo_rl.environments.nemo_gym.NemoGym method) (nemo_rl.environments.reward_model_environment.RewardModelEnvironment method) (nemo_rl.environments.vlm_environment.VLMEnvironment method) (nemo_rl.experience.sync_rollout_actor.SyncRolloutActor method) (nemo_rl.models.generation.dynamo.dynamo_generation.DynamoGeneration method) (nemo_rl.models.generation.dynamo.dynamo_worker.DynamoVllmWorker method) (nemo_rl.models.generation.dynamo.managed_runtime.ManagedDynamoRuntime method) (nemo_rl.models.generation.dynamo.metrics.DynamoMetricsSampler method) (nemo_rl.models.generation.dynamo.token_wrapper.DynamoTokenWrapperServer method) (nemo_rl.models.generation.dynamo.worker_pool.FixedDynamoWorkerPool method) (nemo_rl.models.generation.interfaces.GenerationInterface method) (nemo_rl.models.generation.megatron.megatron_generation.MegatronGeneration method) (nemo_rl.models.generation.trtllm.trtllm_generation.TrtllmGeneration method) (nemo_rl.models.generation.trtllm.trtllm_worker_async.TrtllmAsyncGenerationWorkerImpl method) (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) (nemo_rl.models.generation.vllm.vllm_sparse_refit.VllmSparseRefitReceiver method) (nemo_rl.models.generation.vllm.vllm_worker.VllmGenerationWorkerImpl method) (nemo_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) (nemo_rl.models.policy.interfaces.PolicyInterface method) (nemo_rl.models.policy.lm_policy.Policy method) (nemo_rl.models.policy.teacher_worker_group.TeacherWorkerGroup method) (nemo_rl.models.policy.tq_policy.TQPolicy method) (nemo_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker method) (nemo_rl.models.value.interfaces.ValueInterface method) (nemo_rl.models.value.lm_value.Value method) (nemo_rl.models.value.tq_value.TQValue method) (nemo_rl.utils.checkpoint.CheckpointManager method) (nemo_rl.weight_sync.checkpoint_engine_weight_synchronizer.CheckpointEngineWeightSynchronizer method) (nemo_rl.weight_sync.collective_weight_synchronizer.CollectiveWeightSynchronizer method) (nemo_rl.weight_sync.interfaces.WeightSynchronizer method) (nemo_rl.weight_sync.ipc_weight_synchronizer.IPCWeightSynchronizer method) (nemo_rl.weight_sync.megatron_weight_synchronizer.MegatronWeightSynchronizer method) (nemo_rl.weight_sync.nccl_reshard_weight_synchronizer.NcclReshardWeightSynchronizer method) (nemo_rl.weight_sync.sglang_weight_synchronizer._SGLangWeightSynchronizer method) (nemo_rl.weight_sync.vllm_remote_sparse_weight_synchronizer.VllmRemoteSparseWeightSynchronizer method) shutdown_environments() (in module nemo_rl.algorithms.grpo) shutdown_telemetry() (in module nemo_rl.telemetry.setup) simple (nemo_rl.data_plane.interfaces.DataPlaneConfig attribute) simple_role_header (nemo_rl.data.chat_templates.COMMON_CHAT_TEMPLATES attribute) SimpleStorageConfig (class in nemo_rl.data_plane.interfaces) single_attempt() (nemo_rl.experience.rollout_manager.RolloutRetryPolicy class method) single_controller_epoch (nemo_rl.algorithms.async_utils.replay_buffer.DataPlaneCheckpointMetadata attribute) single_controller_train_steps (nemo_rl.algorithms.async_utils.replay_buffer.DataPlaneCheckpointMetadata attribute) single_controller_trainer_version (nemo_rl.algorithms.async_utils.replay_buffer.DataPlaneCheckpointMetadata attribute) SingleControllerActor (class in nemo_rl.algorithms.single_controller) SingleControllerActorArgs (class in nemo_rl.algorithms.single_controller_utils.setup) SinglePytorchMilestonesConfig (class in nemo_rl.models.policy) SinglePytorchSchedulerConfig (class in nemo_rl.models.policy) size (nemo_rl.data_plane.interfaces.KVBatchMeta property) (nemo_rl.distributed.batched_data_dict.BatchedDataDict property) (nemo_rl.distributed.named_sharding.NamedSharding property) (nemo_rl.models.generation.dynamo.worker_pool.FixedDynamoWorkerPool property) size() (nemo_rl.algorithms.async_utils.interfaces.ReplayBufferProtocol method) (nemo_rl.algorithms.async_utils.replay_buffer.ReplayBufferImpl method) (nemo_rl.algorithms.async_utils.replay_buffer.TQReplayBuffer method) skip_reference_policy_logprobs_calculation (nemo_rl.algorithms.grpo.GRPOConfig attribute) (nemo_rl.algorithms.ppo.PPOConfig attribute) skip_tokenizer_init (nemo_rl.models.generation.vllm.config.VllmSpecificArgs attribute) SKIPPED (nemo_rl.experience.rollout_manager.RolloutOutcome attribute) skipped (nemo_rl.experience.rollout_manager.RolloutStats attribute) sleep() (nemo_rl.models.generation.vllm.vllm_worker.VllmGenerationWorkerImpl method) sleep_async() (nemo_rl.models.generation.trtllm.trtllm_worker_async.TrtllmAsyncGenerationWorkerImpl method) (nemo_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) slice() (nemo_rl.data.multimodal_utils.PackedTensor method) (nemo_rl.data_plane.interfaces.KVBatchMeta method) (nemo_rl.distributed.batched_data_dict.BatchedDataDict method) slice_sparse_projection_cols() (in module nemo_rl.algorithms.x_token.loss_utils) slice_sparse_projection_rows() (in module nemo_rl.algorithms.x_token.loss_utils) SlicedDataDict (class in nemo_rl.distributed.batched_data_dict) snapshot() (nemo_rl.data_plane.observability.MetricsDataPlaneClient method) (nemo_rl.models.generation.dynamo.metrics.DynamoMetricsSampler method) (nemo_rl.models.generation.fleet_health.GenerationFleetHealth method) snapshot_baseline() (nemo_rl.utils.weight_transfer_sparse_codec.DeltaCompressionTracker method) snapshot_start_of_stage() (nemo_rl.utils.memory_tracker.MemoryTracker method) snapshot_step_metrics() (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) solution_key (nemo_rl.data.LocalMathEvalDataConfig attribute) source_max (nemo_rl.algorithms.grpo.RewardScalingConfig attribute) source_min (nemo_rl.algorithms.grpo.RewardScalingConfig attribute) span_groups (nemo_rl.telemetry.config.TelemetryConfig attribute) sparse (nemo_rl.models.generation.vllm.config.VllmRefitConfig attribute) sparse_bucket_size_bytes (nemo_rl.models.generation.vllm.config.VllmDeltaCompressionConfig attribute) sparse_export_chunk_size() (in module nemo_rl.utils.weight_transfer_stream) sparse_locations_for_item() (in module nemo_rl.utils.weight_transfer_sparse_codec) sparse_operation() (in module nemo_rl.utils.weight_transfer_sparse_codec) sparse_payload_checksum() (in module nemo_rl.utils.weight_transfer_stream) SparseInfo (in module nemo_rl.utils.weight_transfer_sparse_codec) SparseItem (in module nemo_rl.utils.weight_transfer_sparse_codec) SparseOperation (in module nemo_rl.utils.weight_transfer_sparse_codec) SparseRefitTransport (class in nemo_rl.utils.weight_transfer_stream) specs (nemo_rl.weight_sync.nccl_reshard_utils.HFToLocalParamMap attribute) spinup_nemo_gym_actor() (in module nemo_rl.environments.nemo_gym) split (nemo_rl.data.DailyOmniEvalDataConfig attribute) (nemo_rl.data.LocalMathEvalDataConfig attribute) (nemo_rl.data.MMAUEvalDataConfig attribute) (nemo_rl.data.PreferenceDatasetConfig attribute) (nemo_rl.data.ResponseDatasetConfig attribute) split_output_tensor() (nemo_rl.algorithms.loss.loss_functions.PreferenceLossFn method) split_train_validation() (nemo_rl.data.datasets.raw_dataset.RawDataset method) split_validation_size (nemo_rl.data.ResponseDatasetConfig attribute) split_weight_chunks() (in module nemo_rl.utils.checkpoint_engines.base) SquadDataset (class in nemo_rl.data.datasets.response_datasets.squad) squeeze_trailing_unit_dim() (in module nemo_rl.algorithms.single_controller_utils.utils) stack_or_nest() (in module nemo_rl.data_plane.codec) stage (nemo_rl.utils.memory_tracker.MemoryTrackerDataPoint attribute) staging_buffer_size (nemo_rl.data_plane.interfaces.MooncakeCpuConfig attribute) staging_dir (nemo_rl.models.generation.vllm.config.VllmRefitStorageConfig attribute) STALE (nemo_rl.models.generation.fleet_health.ShardState attribute) stall_action (nemo_rl.algorithms.single_controller_utils.config.WatchdogConfig attribute) stall_timeout_s (nemo_rl.algorithms.single_controller_utils.config.WatchdogConfig attribute) stall_watchdog (nemo_rl.algorithms.single_controller_utils.config.AsyncRLConfig attribute) stamp_tags() (nemo_rl.data_plane.interfaces.KVBatchMeta method) stand_down() (nemo_rl.distributed.refit_watchdog.RefitAbortWatchdog method) stand_down_armed_watchdogs() (in module nemo_rl.distributed.refit_watchdog) stand_down_refit_watchdog() (nemo_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker method) start() (nemo_rl.models.generation.dynamo.managed_runtime.ManagedDynamoRuntime method) (nemo_rl.models.generation.dynamo.metrics.DynamoMetricsSampler method) (nemo_rl.models.generation.dynamo.token_wrapper.DynamoTokenWrapperServer method) (nemo_rl.models.generation.dynamo.worker_pool.FixedDynamoWorkerPool method) (nemo_rl.models.generation.vllm.vllm_sparse_delta._SparseWeightLoadMode method) (nemo_rl.utils.logger.RayGpuMonitorLogger method) (nemo_rl.utils.timer.ThreadSafeTimer method) (nemo_rl.utils.timer.Timer method) (nemo_rl.utils.weight_transfer_zmq.ZmqSparseRefitServer method) start_baseline() (nemo_rl.weight_sync.vllm_remote_sparse_weight_synchronizer.VllmRemoteSparseWeightSynchronizer static method) start_collection() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector method) start_gpu_profiling() (nemo_rl.models.generation.megatron.megatron_generation.MegatronGeneration method) (nemo_rl.models.generation.trtllm.trtllm_backend.NcclExtension method) (nemo_rl.models.generation.trtllm.trtllm_generation.TrtllmGeneration method) (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) (nemo_rl.models.generation.vllm.vllm_worker.BaseVllmGenerationWorker method) (nemo_rl.models.policy.lm_policy.Policy method) (nemo_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker method) (nemo_rl.utils.nsys.ProfilablePolicy method) start_gpu_profiling_async() (nemo_rl.models.generation.trtllm.trtllm_worker_async.TrtllmAsyncGenerationWorkerImpl method) start_http_server() (nemo_rl.models.generation.trtllm.trtllm_worker_async.TrtllmAsyncGenerationWorkerImpl method) start_iterations() (nemo_rl.utils.timer.TimeoutChecker method) start_server() (in module nemo_rl.models.generation.trtllm.trtllm_http_server) start_sync_server() (nemo_rl.models.generation.vllm.vllm_sparse_refit.VllmSparseRefitReceiver method) start_weight (nemo_rl.algorithms.async_utils.replay_buffer.TQReplayGroupMetadata attribute) start_weight_decay (nemo_rl.models.policy.MegatronSchedulerConfig attribute) start_weight_version (nemo_rl.experience.rollout_recovery.PromptGroupRecoveryRecord attribute) (nemo_rl.experience.rollout_recovery.PromptGroupRecoveryState attribute) start_zmq_sparse_refit_relay() (nemo_rl.models.generation.vllm.vllm_sparse_refit.VllmSparseRefitReceiver method) (nemo_rl.models.generation.vllm.vllm_worker.BaseVllmGenerationWorker method) started_at (nemo_rl.models.generation.vllm.vllm_sparse_refit._StagedSparsePayload attribute) startup_timeout_s (nemo_rl.models.generation.dynamo.config.DynamoCfg attribute) state (nemo_rl.models.generation.fleet_health.ShardHealth attribute) (nemo_rl.models.megatron.config.ModelAndOptimizerState attribute) state_before_partial (nemo_rl.models.generation.fleet_health.ShardHealth attribute) state_dict() (nemo_rl.algorithms.async_utils.interfaces.ReplayBufferProtocol method) (nemo_rl.algorithms.async_utils.replay_buffer.ReplayBufferImpl method) (nemo_rl.data.dataloader.CyclingDataLoader method) (nemo_rl.experience.rollout_recovery.RolloutRecoveryLedger method) (nemo_rl.utils.native_checkpoint.ModelState method) (nemo_rl.utils.native_checkpoint.OptimizerState method) state_of() (nemo_rl.models.generation.fleet_health.GenerationFleetHealth method) StateDict (in module nemo_rl.models.megatron.draft.utils) StatelessProcessGroup (class in nemo_rl.distributed.stateless_process_group) stats (nemo_rl.experience.rollout_manager.RolloutManager property) status (nemo_rl.data_plane.observability.DataPlaneEvent attribute) step (nemo_rl.algorithms.dpo.DPOSaveState attribute) (nemo_rl.algorithms.rm.RMSaveState attribute) (nemo_rl.algorithms.sft.SFTSaveState attribute) (nemo_rl.utils.logger.GpuMetricSnapshot attribute) step() (nemo_rl.environments.code_environment.CodeEnvironment method) (nemo_rl.environments.code_jaccard_environment.CodeJaccardEnvironment method) (nemo_rl.environments.interfaces.EnvironmentInterface method) (nemo_rl.environments.math_environment.MathEnvironment method) (nemo_rl.environments.math_environment.MathMultiRewardEnvironment method) (nemo_rl.environments.nemo_gym.NemoGym method) (nemo_rl.environments.reward_model_environment.RewardModelEnvironment method) (nemo_rl.environments.vlm_environment.VLMEnvironment method) STEP_WINDOW (in module nemo_rl.telemetry.metrics) STEP_WINDOW_WALL_CLOCK_CATEGORIES (in module nemo_rl.algorithms.utils) stop() (nemo_rl.utils.logger.RayGpuMonitorLogger method) (nemo_rl.utils.timer.ThreadSafeTimer method) (nemo_rl.utils.timer.Timer method) stop_at_validation_metric (nemo_rl.algorithms.grpo.GRPOConfig attribute) stop_at_validation_threshold (nemo_rl.algorithms.grpo.GRPOConfig attribute) stop_gpu_profiling() (nemo_rl.models.generation.megatron.megatron_generation.MegatronGeneration method) (nemo_rl.models.generation.trtllm.trtllm_backend.NcclExtension method) (nemo_rl.models.generation.trtllm.trtllm_generation.TrtllmGeneration method) (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) (nemo_rl.models.generation.vllm.vllm_worker.BaseVllmGenerationWorker method) (nemo_rl.models.policy.lm_policy.Policy method) (nemo_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker method) (nemo_rl.utils.nsys.ProfilablePolicy method) stop_gpu_profiling_async() (nemo_rl.models.generation.trtllm.trtllm_worker_async.TrtllmAsyncGenerationWorkerImpl method) stop_http_server() (nemo_rl.models.generation.trtllm.trtllm_worker_async.TrtllmAsyncGenerationWorkerImpl method) stop_properly_penalty_coef (nemo_rl.algorithms.reward_functions.RewardShapingConfig attribute) stop_strings (nemo_rl.data.interfaces.DatumSpec attribute) (nemo_rl.environments.code_jaccard_environment.CodeJaccardEnvConfig attribute) (nemo_rl.environments.math_environment.MathEnvConfig attribute) (nemo_rl.environments.vlm_environment.VLMEnvConfig attribute) (nemo_rl.models.generation.interfaces.GenerationConfig attribute) (nemo_rl.models.generation.interfaces.GenerationDatumSpec attribute) stop_token_ids (nemo_rl.models.generation.interfaces.GenerationConfig attribute) stop_zmq_sparse_refit_relay() (nemo_rl.models.generation.vllm.vllm_sparse_refit.VllmSparseRefitReceiver method) (nemo_rl.models.generation.vllm.vllm_worker.BaseVllmGenerationWorker method) storage (nemo_rl.algorithms.async_utils.replay_buffer.TQReplayMetadataState attribute) (nemo_rl.models.generation.vllm.config.VllmSparseRefitConfig attribute) storage_capacity (nemo_rl.data_plane.interfaces.SimpleStorageConfig attribute) stream() (nemo_rl.models.policy.workers.megatron_remote_sparse_refit.MegatronRemoteSparseRefit method) stream_remote_sparse_weights() (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) stream_sparse_delta_payloads() (in module nemo_rl.utils.weight_transfer_stream) stream_sparse_delta_payloads_via_s3_manifest() (in module nemo_rl.utils.weight_transfer_stream) stream_sparse_delta_payloads_via_zmq() (in module nemo_rl.utils.weight_transfer_zmq) stream_weights_via_ipc_zmq() (nemo_rl.models.policy.interfaces.ColocatablePolicyInterface method) (nemo_rl.models.policy.lm_policy.Policy method) (nemo_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) (nemo_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl method) (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) stream_weights_via_ipc_zmq_impl() (in module nemo_rl.models.policy.utils) strict_agent_name_match (nemo_rl.algorithms.opd.OnPolicyDistillationConfig attribute) STRING_TO_DTYPE (in module nemo_rl.models.automodel.setup) STRONG_AC_VALUE (in module nemo_rl.data.datasets.response_datasets.audiomcq) structural_tag_schema (nemo_rl.models.generation.dynamo.config.DynamoWorkerArgs attribute) structural_tag_scope (nemo_rl.models.generation.dynamo.config.DynamoWorkerArgs attribute) student_chunk_id (nemo_rl.algorithms.x_token.loss_utils.LocalizedAlignment attribute) (nemo_rl.algorithms.x_token.token_aligner.AlignmentBatch attribute) student_exact_partition_mask (nemo_rl.algorithms.x_token.token_aligner.AlignmentBatch attribute) student_input_ids (nemo_rl.algorithms.x_token.loss_utils.LocalizedAlignment attribute) student_logits (nemo_rl.algorithms.loss.loss_functions.DraftCrossEntropyLossDataDict attribute) student_next_token_ce() (in module nemo_rl.algorithms.x_token.loss_utils) student_token_mask (nemo_rl.algorithms.x_token.loss_utils.LocalizedAlignment attribute) student_vocab_indices (nemo_rl.algorithms.loss.loss_functions.DraftCrossEntropyLossDataDict attribute) student_vocab_size (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute) subset (nemo_rl.data.PreferenceDatasetConfig attribute) (nemo_rl.data.ResponseDatasetConfig attribute) subset() (nemo_rl.data_plane.interfaces.KVBatchMeta method) SUBSTITUTIONS (in module nemo_rl.environments.dapo_math_verifier) sum_weights_metric (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute) supports_buffer_checkpoint (nemo_rl.algorithms.async_utils.staleness_sampler.BaseSampler attribute) (nemo_rl.algorithms.async_utils.staleness_sampler.InOrderSampler attribute) (nemo_rl.algorithms.async_utils.staleness_sampler.PromptGroupSampler attribute) (nemo_rl.algorithms.async_utils.staleness_sampler.ReadyFirstSampler attribute) (nemo_rl.algorithms.async_utils.staleness_sampler.WeightFifoSampler attribute) (nemo_rl.algorithms.async_utils.staleness_sampler.WindowedSampler attribute) surpress_user_warnings() (in module nemo_rl.algorithms.utils) surviving_shards (nemo_rl.weight_sync.membership.RefitMembership property) SUSPECT (nemo_rl.models.generation.fleet_health.ShardState attribute) suspected_shards() (nemo_rl.models.generation.fleet_health.GenerationFleetHealth method) suspend_activation_offload_for_forward_only() (in module nemo_rl.models.megatron.train) suspend_for_refit() (nemo_rl.models.generation.megatron.megatron_generation.MegatronGeneration method) (nemo_rl.models.generation.megatron.megatron_worker.MegatronGenerationRefitMixin method) swanlab (nemo_rl.utils.logger.LoggerConfig attribute) swanlab_enabled (nemo_rl.utils.logger.LoggerConfig attribute) SwanlabConfig (class in nemo_rl.utils.logger) SwanlabLogger (class in nemo_rl.utils.logger) swap_weights_via_reshard() (nemo_rl.models.generation.megatron.megatron_worker.MegatronGenerationRefitMixin method) (nemo_rl.models.policy.lm_policy.Policy method) symlink_pre_quantized_model() (in module nemo_rl.modelopt.models.policy.workers.utils) sync_stream_within() (in module nemo_rl.distributed.refit_watchdog) sync_weights() (nemo_rl.weight_sync.checkpoint_engine_weight_synchronizer.CheckpointEngineWeightSynchronizer method) (nemo_rl.weight_sync.collective_weight_synchronizer.CollectiveWeightSynchronizer method) (nemo_rl.weight_sync.interfaces.WeightSynchronizer method) (nemo_rl.weight_sync.ipc_weight_synchronizer.IPCWeightSynchronizer method) (nemo_rl.weight_sync.megatron_weight_synchronizer.MegatronWeightSynchronizer method) (nemo_rl.weight_sync.nccl_reshard_weight_synchronizer.NcclReshardWeightSynchronizer method) (nemo_rl.weight_sync.sglang_weight_synchronizer.SGLangColocatedWeightSynchronizer method) (nemo_rl.weight_sync.sglang_weight_synchronizer.SGLangDisaggregatedWeightSynchronizer method) (nemo_rl.weight_sync.vllm_remote_sparse_weight_synchronizer.VllmRemoteSparseWeightSynchronizer method) synchronize_device() (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) (nemo_rl.models.generation.vllm.vllm_sparse_delta.VllmSparseDeltaApplier method) SyncRolloutActor (class in nemo_rl.experience.sync_rollout_actor) SYSTEM (nemo_rl.distributed.virtual_cluster.PY_EXECUTABLES attribute) SYSTEM_PROMPT (in module nemo_rl.data.datasets.response_datasets.oasst) system_prompt_file (nemo_rl.data.AIMEEvalDataConfig attribute) (nemo_rl.data.DailyOmniEvalDataConfig attribute) (nemo_rl.data.GPQAEvalDataConfig attribute) (nemo_rl.data.interfaces.TaskDataSpec attribute) (nemo_rl.data.LocalMathEvalDataConfig attribute) (nemo_rl.data.MathEvalDataConfig attribute) (nemo_rl.data.MMAUEvalDataConfig attribute) (nemo_rl.data.MMLUEvalDataConfig attribute) (nemo_rl.data.MMLUProEvalDataConfig attribute) (nemo_rl.data.PreferenceDatasetConfig attribute) (nemo_rl.data.ResponseDatasetConfig attribute) system_url (nemo_rl.models.generation.dynamo.dynamo_worker.DynamoVllmWorker property) (nemo_rl.models.generation.dynamo.refit.DynamoWorkerEndpoint attribute) T T (in module nemo_rl.distributed.collectives) t_end (nemo_rl.algorithms.x_token.token_aligner.AlignmentPair attribute) t_start (nemo_rl.algorithms.x_token.token_aligner.AlignmentPair attribute) t_tokens (nemo_rl.algorithms.x_token.token_aligner.AlignmentPair attribute) tags (nemo_rl.data_plane.adapters.noop._Partition attribute) (nemo_rl.data_plane.interfaces.KVBatchMeta attribute) target (nemo_rl.algorithms.async_utils.staleness_sampler.CustomSamplerConfig attribute) target_max (nemo_rl.algorithms.grpo.RewardScalingConfig attribute) target_min (nemo_rl.algorithms.grpo.RewardScalingConfig attribute) target_modules (nemo_rl.models.policy.LoRAConfig attribute) (nemo_rl.models.policy.MegatronPeftConfig attribute) TARGET_SAMPLE_RATE (in module nemo_rl.data.datasets.response_datasets.audiomcq) target_step (nemo_rl.algorithms.async_utils.replay_buffer.TQReplayGroupMetadata attribute) (nemo_rl.experience.rollout_recovery.PromptGroupRecoveryRecord attribute) (nemo_rl.experience.rollout_recovery.PromptGroupRecoveryState attribute) task_index (nemo_rl.experience.rollouts.NemoGymRolloutResult attribute) (nemo_rl.experience.rollouts.RolloutGroupResult attribute) task_name (nemo_rl.data.datasets.response_datasets.audiomcq.AudioMCQDataset attribute) (nemo_rl.data.datasets.response_datasets.avqa.AVQADataset attribute) (nemo_rl.data.datasets.response_datasets.daily_omni.DailyOmniDataset attribute) (nemo_rl.data.datasets.response_datasets.general_conversations_dataset.GeneralConversationsJsonlDataset attribute) (nemo_rl.data.interfaces.DatumSpec attribute) (nemo_rl.data.interfaces.TaskDataSpec attribute) (nemo_rl.data_plane.interfaces.KVBatchMeta attribute) (nemo_rl.experience.rollout_recovery.PromptRef attribute) (nemo_rl.experience.rollout_recovery.PromptRefState attribute) task_spec (nemo_rl.data.datasets.raw_dataset.RawDataset attribute) TaskDataPreProcessFnCallable (class in nemo_rl.data.interfaces) TaskDataProcessFnCallable (class in nemo_rl.data.interfaces) TaskDataSpec (class in nemo_rl.data.interfaces) teacher (nemo_rl.algorithms.distillation.MasterConfig attribute) teacher_chunk_id (nemo_rl.algorithms.x_token.loss_utils.LocalizedAlignment attribute) (nemo_rl.algorithms.x_token.token_aligner.AlignmentBatch attribute) teacher_exact_partition_mask (nemo_rl.algorithms.x_token.token_aligner.AlignmentBatch attribute) teacher_gold_loss (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute) teacher_init_time_s (nemo_rl.algorithms.metric_utils.SetupTimingMetrics attribute) teacher_logits (nemo_rl.algorithms.loss.loss_functions.DraftCrossEntropyLossDataDict attribute) teacher_logprobs_field (nemo_rl.algorithms.opd.TQTeacherLogprobCoordinator attribute) (nemo_rl.algorithms.single_controller_utils.config.AdvantageConfig attribute) TEACHER_LP_FIELDS (in module nemo_rl.data_plane.schema) teacher_model_by_agent_name (nemo_rl.algorithms.opd.OnPolicyDistillationConfig attribute) teacher_model_init_time_s (nemo_rl.algorithms.metric_utils.SetupTimingMetrics attribute) teacher_overrides (nemo_rl.algorithms.opd.NonColocatedTeachersConfig attribute) teacher_reservation_time_s (nemo_rl.algorithms.metric_utils.SetupTimingMetrics attribute) teacher_seq_pad_multiple() (in module nemo_rl.algorithms.opd) teacher_topk_indices (nemo_rl.algorithms.loss.loss_functions.DistillationLossDataDict attribute) teacher_topk_logits (nemo_rl.algorithms.loss.loss_functions.DistillationLossDataDict attribute) teacher_vocab_sizes (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute) teacher_weights (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute) teacher_worker_groups (nemo_rl.algorithms.single_controller_utils.setup.SingleControllerActorArgs attribute) teacher_xtoken_loss (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute) TeacherConfig (class in nemo_rl.algorithms.xtoken_off_policy_distillation) (class in nemo_rl.models.policy.teacher_worker_group) TeacherResourceConfig (class in nemo_rl.algorithms.opd) teachers (nemo_rl.algorithms.xtoken_off_policy_distillation.MasterConfig attribute) TeacherWorkerGroup (class in nemo_rl.models.policy.teacher_worker_group) tee_rl_metrics_to_otel() (in module nemo_rl.telemetry.metrics) telemetry (nemo_rl.algorithms.distillation.MasterConfig attribute) (nemo_rl.algorithms.dpo.MasterConfig attribute) (nemo_rl.algorithms.grpo.MasterConfig attribute) (nemo_rl.algorithms.ppo.MasterConfig attribute) (nemo_rl.algorithms.rm.MasterConfig attribute) (nemo_rl.algorithms.sft.MasterConfig attribute) telemetry_enabled_in_env() (in module nemo_rl.telemetry.setup) TelemetryConfig (class in nemo_rl.telemetry.config) temperature (nemo_rl.algorithms.logits_sampling_utils.TrainingSamplingParams attribute) (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute) (nemo_rl.models.generation.interfaces.GenerationConfig attribute) (nemo_rl.models.generation.interfaces.GenerationSamplingParams attribute) temporal_patch_size (nemo_rl.models.generation.vllm.config.VllmVideoConfig attribute) Tensor (in module nemo_rl.algorithms.loss.loss_functions) (in module nemo_rl.algorithms.loss.wrapper) (in module nemo_rl.algorithms.reward_functions) (in module nemo_rl.data.llm_message_utils) (in module nemo_rl.models.huggingface.common) tensor_field() (in module nemo_rl.algorithms.single_controller_utils.utils) tensor_model_parallel_size (nemo_rl.algorithms.opd.TeacherResourceConfig attribute) (nemo_rl.models.policy.MegatronConfig attribute) (nemo_rl.models.policy.teacher_worker_group.TeacherConfig attribute) tensor_parallel_size (nemo_rl.models.generation.dynamo.config.DynamoVllmConfig attribute) (nemo_rl.models.generation.trtllm.config.TrtllmSpecificArgs attribute) (nemo_rl.models.generation.vllm.config.VllmSpecificArgs attribute) (nemo_rl.models.policy.DTensorConfig attribute) TensorBatch (in module nemo_rl.utils.weight_transfer_sparse_codec) tensorboard (nemo_rl.utils.logger.LoggerConfig attribute) tensorboard_enabled (nemo_rl.utils.logger.LoggerConfig attribute) TensorboardConfig (class in nemo_rl.utils.logger) TensorboardLogger (class in nemo_rl.utils.logger) TensorMeta (class in nemo_rl.utils.checkpoint_engines.base) TensorPayload (in module nemo_rl.utils.weight_transfer_sparse_codec) terminate_on_evaluation (nemo_rl.environments.code_environment.CodeEnvConfig attribute) terminateds (nemo_rl.environments.interfaces.EnvironmentReturn attribute) THEORETICAL_TFLOPS (in module nemo_rl.utils.flops_tracker) think_close (nemo_rl.algorithms.grpo.RewardPenaltyTokenIdsConfig attribute) think_open (nemo_rl.algorithms.grpo.RewardPenaltyTokenIdsConfig attribute) thinking_tags (nemo_rl.environments.nemo_gym.NemoGymConfig attribute) THREAD_ACCUMULATED_EFFICIENCY_CATEGORIES (in module nemo_rl.algorithms.utils) THREAD_SECONDS_MEASUREMENT (in module nemo_rl.telemetry.metrics) ThreadSafeTimer (class in nemo_rl.utils.timer) time() (nemo_rl.utils.timer.ThreadSafeTimer method) (nemo_rl.utils.timer.Timer method) TimeoutChecker (class in nemo_rl.utils.timer) Timer (class in nemo_rl.utils.timer) to() (nemo_rl.data.multimodal_utils.PackedTensor method) (nemo_rl.distributed.batched_data_dict.BatchedDataDict method) to_dtype() (nemo_rl.data.multimodal_utils.PackedTensor method) to_local_if_dtensor() (in module nemo_rl.models.dtensor.parallelize) to_metrics_dict() (nemo_rl.algorithms.metric_utils.SetupTimingMetrics method) to_nested_by_length() (in module nemo_rl.data_plane.codec) to_torch_dtype() (in module nemo_rl.models.megatron.community_import) TOKEN_ALIGNED_FIELDS (in module nemo_rl.data_plane.column_io) token_ids (nemo_rl.algorithms.grpo.RewardPenaltyConfig attribute) TOKEN_LEVEL (nemo_rl.algorithms.loss.interfaces.LossType attribute) token_level_loss (nemo_rl.algorithms.loss.loss_functions.ClippedPGLossConfig attribute) token_mask (nemo_rl.algorithms.loss.loss_functions.ClippedPGLossDataDict attribute) (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossDataDict attribute) (nemo_rl.algorithms.loss.loss_functions.DistillationLossDataDict attribute) (nemo_rl.algorithms.loss.loss_functions.DPOLossDataDict attribute) (nemo_rl.algorithms.loss.loss_functions.DraftCrossEntropyLossDataDict attribute) (nemo_rl.algorithms.loss.loss_functions.PreferenceLossDataDict attribute) token_mask_field (nemo_rl.algorithms.single_controller_utils.config.AdvantageConfig attribute) TokenAligner (class in nemo_rl.algorithms.x_token.token_aligner) tokenizer (nemo_rl.evals.eval.MasterConfig attribute) (nemo_rl.models.generation.dynamo.config.DynamoFrontendArgs attribute) (nemo_rl.models.policy.PolicyConfig attribute) (nemo_rl.models.value.config.ValueConfig attribute) tokenizer_cache (nemo_rl.models.generation.dynamo.config.DynamoFrontendArgs attribute) tokenizer_cache_bytes (nemo_rl.models.generation.dynamo.config.DynamoFrontendArgs attribute) tokenizer_config (nemo_rl.environments.nemo_gym.NemoGymConfig attribute) tokenizer_kwargs (nemo_rl.models.policy.TokenizerConfig attribute) TokenizerConfig (class in nemo_rl.models.policy) TokenizerType (in module nemo_rl.algorithms.async_utils.trajectory_collector) (in module nemo_rl.algorithms.distillation) (in module nemo_rl.algorithms.grpo) (in module nemo_rl.algorithms.ppo) (in module nemo_rl.data.collate_fn) (in module nemo_rl.data.datasets.processed_dataset) (in module nemo_rl.data.datasets.utils) (in module nemo_rl.data.interfaces) (in module nemo_rl.data.llm_message_utils) (in module nemo_rl.data.processors) (in module nemo_rl.experience.rollout_manager) (in module nemo_rl.experience.rollouts) (in module nemo_rl.models.generation) (in module nemo_rl.models.megatron.setup) (in module nemo_rl.models.policy.workers.megatron_policy_worker) (in module nemo_rl.models.value.workers.megatron_value_worker) TOKENS (nemo_rl.algorithms.loss.interfaces.MetricNormalizer attribute) tool_call_parser (nemo_rl.models.generation.dynamo.config.DynamoWorkerArgs attribute) tool_parser (nemo_rl.models.generation.trtllm.config.TrtllmSpecificArgs attribute) tool_parser_plugin (nemo_rl.models.generation.vllm.config.VllmSpecificArgs attribute) top_k (nemo_rl.algorithms.logits_sampling_utils.TrainingSamplingParams attribute) (nemo_rl.models.generation.interfaces.GenerationConfig attribute) (nemo_rl.models.generation.interfaces.GenerationSamplingParams attribute) TOP_K_TOP_P_CHUNK_SIZE (in module nemo_rl.algorithms.logits_sampling_utils) top_p (nemo_rl.algorithms.logits_sampling_utils.TrainingSamplingParams attribute) (nemo_rl.models.generation.interfaces.GenerationConfig attribute) (nemo_rl.models.generation.interfaces.GenerationSamplingParams attribute) topk_indices (nemo_rl.models.policy.interfaces.TopkLogitsOutputSpec attribute) topk_logits (nemo_rl.models.policy.interfaces.TopkLogitsOutputSpec attribute) topk_logits_k (nemo_rl.algorithms.distillation.DistillationConfig attribute) TopkLogitsOutputSpec (class in nemo_rl.models.policy.interfaces) TopkLogitsPostProcessor (class in nemo_rl.models.automodel.train) (class in nemo_rl.models.megatron.train) TOPO_RANK_KEY (in module nemo_rl.distributed.virtual_cluster) TOPO_RANK_UNKNOWN (in module nemo_rl.distributed.virtual_cluster) total_bytes (nemo_rl.data_plane.observability.DataPlaneStats attribute) total_keys (nemo_rl.data_plane.observability.DataPlaneStats attribute) total_ops (nemo_rl.data_plane.observability.DataPlaneStats attribute) total_setup_time_s (nemo_rl.algorithms.metric_utils.SetupTimingMetrics attribute) total_steps (nemo_rl.algorithms.distillation.DistillationSaveState attribute) (nemo_rl.algorithms.dpo.DPOSaveState attribute) (nemo_rl.algorithms.grpo.GRPOSaveState attribute) (nemo_rl.algorithms.ppo.PPOSaveState attribute) (nemo_rl.algorithms.rm.RMSaveState attribute) (nemo_rl.algorithms.sft.SFTSaveState attribute) (nemo_rl.algorithms.xtoken_off_policy_distillation.OffPolicyDistillationSaveState attribute) total_valid_tokens (nemo_rl.algorithms.distillation.DistillationSaveState attribute) (nemo_rl.algorithms.dpo.DPOSaveState attribute) (nemo_rl.algorithms.grpo.GRPOSaveState attribute) (nemo_rl.algorithms.ppo.PPOSaveState attribute) (nemo_rl.algorithms.rm.RMSaveState attribute) (nemo_rl.algorithms.sft.SFTSaveState attribute) (nemo_rl.algorithms.xtoken_off_policy_distillation.OffPolicyDistillationSaveState attribute) tp_rank (nemo_rl.models.generation.vllm.refit_layout.VllmExpertParamLayout attribute) tp_shard_dim (nemo_rl.models.generation.vllm.refit_layout.HfExpertWeight attribute) tp_size (nemo_rl.models.automodel.config.DistributedContext attribute) (nemo_rl.models.generation.vllm.refit_layout.VllmExpertParamLayout attribute) tq_buffer (nemo_rl.algorithms.single_controller_utils.setup.SingleControllerActorArgs attribute) TQDataPlaneClient (class in nemo_rl.data_plane.adapters.transfer_queue) TQDriverMixin (class in nemo_rl.data_plane.driver_mixin) TQPolicy (class in nemo_rl.models.policy.tq_policy) TQReplayBuffer (class in nemo_rl.algorithms.async_utils.replay_buffer) TQReplayGroupMetadata (class in nemo_rl.algorithms.async_utils.replay_buffer) TQReplayMetadataState (class in nemo_rl.algorithms.async_utils.replay_buffer) TQTeacherLogprobCoordinator (class in nemo_rl.algorithms.opd) TQValue (class in nemo_rl.models.value.tq_value) TQWorkerMixin (class in nemo_rl.data_plane.worker_mixin) trace_cp_routed_experts() (in module nemo_rl.utils.r3_trace) trace_fn() (in module nemo_rl.telemetry.instrumentation) trace_rollout_payload() (in module nemo_rl.utils.r3_trace) trace_router_replay_action() (in module nemo_rl.utils.r3_trace) trace_router_replay_assignment() (in module nemo_rl.utils.r3_trace) trace_tq_fetch_payload() (in module nemo_rl.utils.r3_trace) traces_enabled (nemo_rl.telemetry.config.TelemetryConfig attribute) track() (nemo_rl.utils.flops_tracker.FLOPTracker method) track_batch() (nemo_rl.utils.flops_tracker.FLOPTracker method) tracking_uri (nemo_rl.utils.logger.MLflowConfig attribute) train (nemo_rl.data.DataConfig attribute) train() (nemo_rl.models.policy.interfaces.PolicyInterface method) (nemo_rl.models.policy.lm_policy.Policy method) (nemo_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) (nemo_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl method) (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) (nemo_rl.models.value.interfaces.ValueInterface method) (nemo_rl.models.value.lm_value.Value method) (nemo_rl.models.value.workers.dtensor_value_worker_v2.DTensorValueWorkerV2Impl method) (nemo_rl.models.value.workers.megatron_value_worker.MegatronValueWorkerImpl method) train_cluster (nemo_rl.algorithms.single_controller_utils.setup.SingleControllerActorArgs attribute) train_context() (nemo_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl static method) train_from_meta() (nemo_rl.models.policy.tq_policy.TQPolicy method) (nemo_rl.models.value.tq_value.TQValue method) train_global_batch_size (nemo_rl.models.policy.PolicyConfig attribute) (nemo_rl.models.value.config.ValueConfig attribute) train_mb_tokens (nemo_rl.models.policy.DynamicBatchingConfig attribute) (nemo_rl.models.policy.SequencePackingConfig attribute) train_micro_batch_size (nemo_rl.models.policy.PolicyConfig attribute) (nemo_rl.models.value.config.ValueConfig attribute) train_microbatch() (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) train_microbatch_presharded() (nemo_rl.data_plane.worker_mixin.TQWorkerMixin method) train_microbatches_from_meta() (nemo_rl.models.policy.tq_policy.TQPolicy method) train_presharded() (nemo_rl.data_plane.worker_mixin.TQWorkerMixin method) train_world_size (nemo_rl.weight_sync.membership.RefitMembership attribute) TRAINED_TASK_INDICES_KEY (in module nemo_rl.experience.interfaces) trainer_handle (nemo_rl.algorithms.single_controller_utils.setup.SingleControllerActorArgs attribute) trainer_version (nemo_rl.algorithms.grpo.GRPOSaveState attribute) TrainingSamplingParams (class in nemo_rl.algorithms.logits_sampling_utils) TransactionalAdmissionSampler (class in nemo_rl.algorithms.async_utils.staleness_sampler) transfer_workers (nemo_rl.models.generation.vllm.config.VllmRefitTuningConfig attribute) (nemo_rl.utils.weight_transfer_stream.SparseRefitTransport attribute) transformer() (in module nemo_rl.utils.flops_formulas) transformer_impl (nemo_rl.models.policy.MegatronConfig attribute) translate_parallel_style() (in module nemo_rl.models.dtensor.parallelize) TRTLLM (nemo_rl.distributed.virtual_cluster.PY_EXECUTABLES attribute) trtllm_cfg (nemo_rl.models.generation.trtllm.config.TrtllmConfig attribute) TRTLLM_EXECUTABLE (in module nemo_rl.distributed.ray_actor_environment_registry) trtllm_kwargs (nemo_rl.models.generation.trtllm.config.TrtllmConfig attribute) TrtllmAsyncGenerationWorker (class in nemo_rl.models.generation.trtllm.trtllm_worker_async) TrtllmAsyncGenerationWorkerImpl (class in nemo_rl.models.generation.trtllm.trtllm_worker_async) TrtllmConfig (class in nemo_rl.models.generation.trtllm.config) TrtllmGeneration (class in nemo_rl.models.generation.trtllm.trtllm_generation) TrtllmSpecificArgs (class in nemo_rl.models.generation.trtllm.config) truncate_tensors() (nemo_rl.distributed.batched_data_dict.BatchedDataDict method) truncated (nemo_rl.experience.interfaces.Completion attribute) (nemo_rl.models.generation.interfaces.GenerationOutputSpec attribute) truncated_importance_sampling_ratio (nemo_rl.algorithms.loss.loss_functions.ClippedPGLossConfig attribute) truncated_importance_sampling_ratio_min (nemo_rl.algorithms.loss.loss_functions.ClippedPGLossConfig attribute) truncated_importance_sampling_type (nemo_rl.algorithms.loss.loss_functions.ClippedPGLossConfig attribute) Tulu3PreferenceDataset (class in nemo_rl.data.datasets.preference_datasets.tulu3) Tulu3SftMixtureDataset (class in nemo_rl.data.datasets.response_datasets.tulu3) tuning (nemo_rl.models.generation.vllm.config.VllmSparseRefitConfig attribute) U UMBRELLA_GROUPS (in module nemo_rl.telemetry.instrumentation) uncommon_topk (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute) unhealthy_threshold (nemo_rl.algorithms.single_controller_utils.config.FleetHealthConfig attribute) (nemo_rl.models.generation.fleet_health.FleetHealthPolicy attribute) unpack_tensor() (in module nemo_rl.models.huggingface.common) unpadded_sequence_lengths (nemo_rl.models.generation.interfaces.GenerationOutputSpec attribute) unshard_fsdp2_model() (in module nemo_rl.models.policy.workers.dtensor_policy_worker) UnsupportedNativeRefitTransport (in module nemo_rl.models.generation.vllm.vllm_backend) unwanted (nemo_rl.algorithms.grpo.RewardPenaltyTokenIdsConfig attribute) unwrap_wire_stripped_payload() (in module nemo_rl.data_plane.codec) up_weight (nemo_rl.models.megatron.draft.utils._PendingLayerWeights attribute) update() (nemo_rl.data.packing.metrics.PackingMetrics method) update_checkpointer_config() (nemo_rl.models.automodel.checkpoint.AutomodelCheckpointManager method) update_single_dataset_config() (in module nemo_rl.data.datasets.utils) update_weights() (nemo_rl.models.generation.dynamo.refit.DynamoRefitChannel method) update_weights_bucket_memory_ratio (nemo_rl.models.generation.interfaces.CheckpointEngineConfig attribute) (nemo_rl.models.generation.vllm.config.VllmCheckpointEnginePluginConfig attribute) (nemo_rl.models.generation.vllm.config.VllmNixlRefitConfig attribute) update_weights_from_checkpoint_engine() (nemo_rl.models.generation.vllm.checkpoint_engine.VllmCheckpointEngineMixin method) update_weights_from_collective() (nemo_rl.models.generation.dynamo.dynamo_generation.DynamoGeneration method) (nemo_rl.models.generation.interfaces.GenerationInterface method) (nemo_rl.models.generation.megatron.megatron_generation.MegatronGeneration method) (nemo_rl.models.generation.trtllm.trtllm_backend.NcclExtension method) (nemo_rl.models.generation.trtllm.trtllm_generation.TrtllmGeneration method) (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) (nemo_rl.models.generation.vllm.vllm_worker.VllmGenerationWorkerImpl method) update_weights_from_collective_async() (nemo_rl.models.generation.trtllm.trtllm_worker_async.TrtllmAsyncGenerationWorkerImpl method) (nemo_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) update_weights_from_decoded_sparse_payload() (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) (nemo_rl.models.generation.vllm.vllm_sparse_delta.VllmSparseDeltaApplier method) update_weights_from_serialized_sparse_payloads() (nemo_rl.models.generation.vllm.vllm_sparse_refit.VllmSparseRefitReceiver method) update_weights_from_staged_sparse_payloads() (nemo_rl.models.generation.vllm.vllm_sparse_refit.VllmSparseRefitReceiver method) update_weights_to_sglang_colocated() (nemo_rl.models.policy.interfaces.ColocatablePolicyInterface method) (nemo_rl.models.policy.lm_policy.Policy method) (nemo_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl method) (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) update_weights_to_sglang_distributed() (nemo_rl.models.policy.interfaces.ColocatablePolicyInterface method) (nemo_rl.models.policy.lm_policy.Policy method) (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) update_weights_via_ipc_zmq() (nemo_rl.models.generation.dynamo.dynamo_generation.DynamoGeneration method) (nemo_rl.models.generation.interfaces.GenerationInterface method) (nemo_rl.models.generation.trtllm.trtllm_backend.NcclExtension method) (nemo_rl.models.generation.trtllm.trtllm_generation.TrtllmGeneration method) (nemo_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) (nemo_rl.models.generation.vllm.vllm_generation.VllmGeneration method) (nemo_rl.models.generation.vllm.vllm_worker.VllmGenerationWorkerImpl method) update_weights_via_ipc_zmq_async() (nemo_rl.models.generation.trtllm.trtllm_worker_async.TrtllmAsyncGenerationWorkerImpl method) (nemo_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) use_async_rollouts (nemo_rl.models.generation.interfaces.GenerationConfig attribute) use_cispo (nemo_rl.algorithms.loss.loss_functions.ClippedPGLossConfig attribute) use_cuda_graphs_for_non_decode_steps (nemo_rl.models.generation.megatron.config.MCoreGenerationSpecificArgs attribute) use_custom_fsdp (nemo_rl.models.policy.MegatronDDPConfig attribute) use_distributed_optimizer (nemo_rl.models.policy.MegatronOptimizerConfig attribute) use_dynamic_sampling (nemo_rl.algorithms.grpo.GRPOConfig attribute) (nemo_rl.algorithms.ppo.PPOConfig attribute) use_fastokens (nemo_rl.environments.nemo_gym.NemoGymConfig attribute) (nemo_rl.models.policy.TokenizerConfig attribute) use_fused_linear_logprobs (nemo_rl.models.policy.MegatronConfig attribute) use_fused_weighted_squared_relu (nemo_rl.models.policy.MegatronConfig attribute) use_gloo_process_groups (nemo_rl.models.policy.MegatronConfig attribute) use_importance_sampling_correction (nemo_rl.algorithms.loss.loss_functions.ClippedPGLossConfig attribute) use_kl_in_reward (nemo_rl.algorithms.loss.loss_functions.ClippedPGLossConfig attribute) use_leave_one_out_baseline (nemo_rl.algorithms.advantage_estimator.AdvEstimatorConfig attribute) (nemo_rl.algorithms.grpo.GRPOConfig attribute) use_liger_kernel (nemo_rl.models.policy.AutomodelKwargs attribute) use_multiple_dataloader (nemo_rl.data.DataConfig attribute) use_on_policy_kl_approximation (nemo_rl.algorithms.loss.loss_functions.ClippedPGLossConfig attribute) use_precision_aware_optimizer (nemo_rl.models.policy.MegatronOptimizerConfig attribute) use_processor (nemo_rl.models.policy.TokenizerConfig attribute) use_reference_model() (nemo_rl.modelopt.models.policy.workers.megatron_quant_policy_worker.MegatronQuantPolicyWorker method) (nemo_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) (nemo_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl method) (nemo_rl.models.policy.workers.megatron_policy_worker.MegatronPolicyWorkerImpl method) USE_SYSTEM_EXECUTABLE (in module nemo_rl.distributed.ray_actor_environment_registry) (in module nemo_rl.modelopt.registry) use_tqdm (nemo_rl.models.generation.vllm.config.VllmSpecificArgs attribute) use_triton (nemo_rl.models.policy.LoRAConfig attribute) uses_image_placeholder() (in module nemo_rl.data.multimodal_utils) V v_head_dim (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) v_weight (nemo_rl.models.megatron.draft.utils._PendingLayerWeights attribute) val_at_end (nemo_rl.algorithms.distillation.DistillationConfig attribute) (nemo_rl.algorithms.dpo.DPOConfig attribute) (nemo_rl.algorithms.grpo.GRPOConfig attribute) (nemo_rl.algorithms.ppo.PPOConfig attribute) (nemo_rl.algorithms.rm.RMConfig attribute) (nemo_rl.algorithms.sft.SFTConfig attribute) (nemo_rl.algorithms.xtoken_off_policy_distillation.OffPolicyDistillationConfig attribute) val_at_start (nemo_rl.algorithms.distillation.DistillationConfig attribute) (nemo_rl.algorithms.dpo.DPOConfig attribute) (nemo_rl.algorithms.grpo.GRPOConfig attribute) (nemo_rl.algorithms.ppo.PPOConfig attribute) (nemo_rl.algorithms.rm.RMConfig attribute) (nemo_rl.algorithms.sft.SFTConfig attribute) (nemo_rl.algorithms.xtoken_off_policy_distillation.OffPolicyDistillationConfig attribute) val_batch_size (nemo_rl.algorithms.distillation.DistillationConfig attribute) (nemo_rl.algorithms.grpo.GRPOConfig attribute) (nemo_rl.algorithms.ppo.PPOConfig attribute) val_batches (nemo_rl.algorithms.dpo.DPOConfig attribute) (nemo_rl.algorithms.rm.RMConfig attribute) (nemo_rl.algorithms.sft.SFTConfig attribute) val_dataset (nemo_rl.data.datasets.raw_dataset.RawDataset attribute) val_global_batch_size (nemo_rl.algorithms.dpo.DPOConfig attribute) (nemo_rl.algorithms.rm.RMConfig attribute) (nemo_rl.algorithms.sft.SFTConfig attribute) val_loss (nemo_rl.algorithms.xtoken_off_policy_distillation.OffPolicyDistillationSaveState attribute) val_micro_batch_size (nemo_rl.algorithms.dpo.DPOConfig attribute) (nemo_rl.algorithms.rm.RMConfig attribute) (nemo_rl.algorithms.sft.SFTConfig attribute) val_num_generations_per_prompt (nemo_rl.algorithms.grpo.GRPOConfig attribute) val_period (nemo_rl.algorithms.distillation.DistillationConfig attribute) (nemo_rl.algorithms.dpo.DPOConfig attribute) (nemo_rl.algorithms.grpo.GRPOConfig attribute) (nemo_rl.algorithms.ppo.PPOConfig attribute) (nemo_rl.algorithms.rm.RMConfig attribute) (nemo_rl.algorithms.sft.SFTConfig attribute) (nemo_rl.algorithms.xtoken_off_policy_distillation.OffPolicyDistillationConfig attribute) val_reward (nemo_rl.algorithms.distillation.DistillationSaveState attribute) (nemo_rl.algorithms.grpo.GRPOSaveState attribute) (nemo_rl.algorithms.ppo.PPOSaveState attribute) val_start_at (nemo_rl.algorithms.grpo.GRPOConfig attribute) val_temperature (nemo_rl.models.generation.interfaces.GenerationConfig attribute) val_top_k (nemo_rl.models.generation.interfaces.GenerationConfig attribute) val_top_p (nemo_rl.models.generation.interfaces.GenerationConfig attribute) valid_chunk_mask() (in module nemo_rl.algorithms.x_token.loss_utils) validate() (in module nemo_rl.algorithms.distillation) (in module nemo_rl.algorithms.dpo) (in module nemo_rl.algorithms.grpo) (in module nemo_rl.algorithms.ppo) (in module nemo_rl.algorithms.rm) (in module nemo_rl.algorithms.sft) (in module nemo_rl.algorithms.xtoken_off_policy_distillation) (nemo_rl.models.generation.dynamo.worker_pool.FixedDynamoWorkerPool method) validate_algorithm_block() (nemo_rl.algorithms.single_controller_utils.config.MasterConfig method) validate_and_prepare_config() (in module nemo_rl.models.automodel.setup) validate_and_set_config() (in module nemo_rl.models.megatron.setup) validate_async_warmup_settings() (nemo_rl.algorithms.ppo.PPOConfig method) validate_batch() (nemo_rl.models.generation.vllm.vllm_backend._IPCWeightManifest method) validate_model_paths() (in module nemo_rl.models.megatron.setup) validate_one_dataset() (in module nemo_rl.algorithms.dpo) (in module nemo_rl.algorithms.rm) validate_reward_components_match_scalar() (in module nemo_rl.environments.nemo_gym) validate_router_replay_config() (in module nemo_rl.models.megatron.router_replay) validate_sampler_buffer_capacity() (in module nemo_rl.algorithms.single_controller_utils.config) validate_settings() (nemo_rl.algorithms.ppo.AsyncPPOConfig method) (nemo_rl.models.generation.interfaces.GenerationInterface class method) (nemo_rl.models.generation.megatron.megatron_generation.MegatronGeneration class method) validate_single_controller_config() (in module nemo_rl.algorithms.single_controller_utils.config) validate_sync() (in module nemo_rl.algorithms.grpo_sync) validate_vllm_remote_sparse_refit() (in module nemo_rl.weight_sync.vllm_remote_sparse_weight_synchronizer) validate_warm_start_checkpoint() (in module nemo_rl.utils.checkpoint) validate_warmup_lookahead() (nemo_rl.algorithms.async_utils.staleness_sampler.InOrderSamplerConfig method) validate_workers() (nemo_rl.models.generation.dynamo.managed_runtime.ManagedDynamoRuntime method) validation (nemo_rl.data.DataConfig attribute) Value (class in nemo_rl.models.value.lm_value) value (nemo_rl.algorithms.ppo.MasterConfig attribute) (nemo_rl.algorithms.single_controller_utils.config.MasterConfig attribute) value_handle (nemo_rl.algorithms.single_controller_utils.setup.SingleControllerActorArgs attribute) value_init_time_s (nemo_rl.algorithms.metric_utils.SetupTimingMetrics attribute) value_loss_fn (nemo_rl.algorithms.ppo.MasterConfig attribute) (nemo_rl.algorithms.single_controller_utils.config.MasterConfig attribute) (nemo_rl.algorithms.single_controller_utils.setup.SingleControllerActorArgs attribute) VALUE_SEED_FIELDS (in module nemo_rl.data_plane.schema) ValueConfig (class in nemo_rl.models.value.config) ValueInterface (class in nemo_rl.models.value.interfaces) ValueOutputSpec (class in nemo_rl.models.value.interfaces) values (nemo_rl.models.value.interfaces.ValueOutputSpec attribute) values_field (nemo_rl.algorithms.single_controller_utils.config.AdvantageConfig attribute) variables_after_stage (nemo_rl.utils.memory_tracker.MemoryTrackerDataPoint attribute) variables_before_stage (nemo_rl.utils.memory_tracker.MemoryTrackerDataPoint attribute) vec_in_dim (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) verifier_type (nemo_rl.environments.math_environment.MathEnvConfig attribute) verify() (in module nemo_rl.environments.dapo_math_verifier) (nemo_rl.environments.code_jaccard_environment.CodeJaccardVerifyWorker method) (nemo_rl.environments.math_environment.EnglishMultichoiceVerifyWorker method) (nemo_rl.environments.math_environment.HFMultiRewardVerifyWorker method) (nemo_rl.environments.math_environment.HFVerifyWorker method) (nemo_rl.environments.math_environment.MultilingualMultichoiceVerifyWorker method) (nemo_rl.environments.vlm_environment.VLMVerifyWorker method) verify_right_padding() (in module nemo_rl.models.generation.interfaces) verify_samples_per_payload (nemo_rl.models.generation.vllm.config.VllmSparseRefitConfig attribute) verify_served_address() (nemo_rl.models.generation.megatron.megatron_generation.MegatronGeneration class method) VERSION (in module nemo_rl.package_info) video (nemo_rl.models.generation.vllm.config.VllmSpecificArgs attribute) (nemo_rl.models.policy.TokenizerConfig attribute) VIDEO_CONTENT_TYPES (in module nemo_rl.data.multimodal_utils) video_maintain_aspect_ratio (nemo_rl.data.interfaces.TaskDataSpec attribute) (nemo_rl.data.PreferenceDatasetConfig attribute) (nemo_rl.data.ResponseDatasetConfig attribute) video_sampling_style (nemo_rl.data.interfaces.TaskDataSpec attribute) (nemo_rl.data.PreferenceDatasetConfig attribute) (nemo_rl.data.ResponseDatasetConfig attribute) video_target_num_patches (nemo_rl.data.interfaces.TaskDataSpec attribute) (nemo_rl.data.PreferenceDatasetConfig attribute) (nemo_rl.data.ResponseDatasetConfig attribute) video_temporal_patch_size (nemo_rl.data.interfaces.TaskDataSpec attribute) (nemo_rl.data.PreferenceDatasetConfig attribute) (nemo_rl.data.ResponseDatasetConfig attribute) VideoSamplingStyle (in module nemo_rl.models.generation.vllm.video_utils) VIOLATION_TAG_KEYS (in module nemo_rl.experience.payload) VISUAL_BYTE_MAP (in module nemo_rl.algorithms.x_token.token_aligner) VLLM (nemo_rl.distributed.virtual_cluster.PY_EXECUTABLES attribute) VLLM_BACKEND (in module nemo_rl.models.generation.constants) vllm_cfg (nemo_rl.models.generation.dynamo.config.DynamoConfig attribute) (nemo_rl.models.generation.vllm.config.VllmConfig attribute) vllm_checkpoint_engine_init_time_s (nemo_rl.algorithms.metric_utils.SetupTimingMetrics attribute) VLLM_EXECUTABLE (in module nemo_rl.distributed.ray_actor_environment_registry) vllm_kwargs (nemo_rl.models.generation.dynamo.config.DynamoConfig attribute) (nemo_rl.models.generation.vllm.config.VllmConfig attribute) VLLM_LOAD_FORMAT_AUTO (nemo_rl.models.huggingface.common.ModelFlag attribute) VLLM_LOGPROB_FLOOR (in module nemo_rl.models.generation.vllm.utils) vllm_metrics_logger_interval (nemo_rl.models.generation.dynamo.config.DynamoVllmConfig attribute) VLLM_MULTIMODAL_DATA_KEYS (in module nemo_rl.data.multimodal_utils) vllm_native_tracing (nemo_rl.telemetry.config.TelemetryConfig attribute) vllm_native_tracing_requested() (in module nemo_rl.telemetry.setup) VLLM_PACKED_BUFFER_SIZE_BYTES (in module nemo_rl.models.generation.dynamo.config) VLLM_PACKED_NUM_BUFFERS (in module nemo_rl.models.generation.dynamo.config) vllm_refit_api_key() (in module nemo_rl.utils.weight_transfer_http) vllm_refit_endpoints() (in module nemo_rl.utils.weight_transfer_http) VLLM_SPARSE_REFIT_TRANSPORTS (in module nemo_rl.models.generation.vllm.config) VllmAsyncCheckpointEngineRpcMixin (class in nemo_rl.models.generation.vllm.checkpoint_engine) VllmAsyncGenerationWorker (class in nemo_rl.models.generation.vllm.vllm_worker_async) VllmAsyncGenerationWorkerImpl (class in nemo_rl.models.generation.vllm.vllm_worker_async) VllmCheckpointEngineMixin (class in nemo_rl.models.generation.vllm.checkpoint_engine) VllmCheckpointEnginePluginConfig (class in nemo_rl.models.generation.vllm.config) VllmCheckpointEngineRpcMixin (class in nemo_rl.models.generation.vllm.checkpoint_engine) VllmConfig (class in nemo_rl.models.generation.vllm.config) VllmDeltaCompressionConfig (class in nemo_rl.models.generation.vllm.config) VllmExpertParamLayout (class in nemo_rl.models.generation.vllm.refit_layout) VllmGeneration (class in nemo_rl.models.generation.vllm.vllm_generation) VllmGenerationWorker (class in nemo_rl.models.generation.vllm.vllm_worker) VllmGenerationWorkerImpl (class in nemo_rl.models.generation.vllm.vllm_worker) VllmInternalWorkerExtension (class in nemo_rl.models.generation.vllm.vllm_backend) VllmInternalWorkerExtensionWithCheckpointEngine (class in nemo_rl.models.generation.vllm.vllm_backend) VllmNixlRefitConfig (class in nemo_rl.models.generation.vllm.config) VllmQuantAsyncGenerationWorker (class in nemo_rl.modelopt.models.generation.vllm_quant_worker) VllmQuantGenerationWorker (class in nemo_rl.modelopt.models.generation.vllm_quant_worker) VllmQuantInternalWorkerExtension (class in nemo_rl.modelopt.models.generation.vllm_quant_backend) VllmQuantInternalWorkerExtensionWithCheckpointEngine (class in nemo_rl.modelopt.models.generation.vllm_quant_backend) VllmRefitBaselineConfig (class in nemo_rl.models.generation.vllm.config) VllmRefitConfig (class in nemo_rl.models.generation.vllm.config) VllmRefitSelector (in module nemo_rl.models.generation.vllm.config) VllmRefitStorageConfig (class in nemo_rl.models.generation.vllm.config) VllmRefitTransportName (in module nemo_rl.models.generation.vllm.config) VllmRefitTuningConfig (class in nemo_rl.models.generation.vllm.config) VllmRemoteSparseWeightSynchronizer (class in nemo_rl.weight_sync.vllm_remote_sparse_weight_synchronizer) VllmShardedExpertRefitMixin (class in nemo_rl.models.generation.vllm.refit_loader) VllmSparseDeltaApplier (class in nemo_rl.models.generation.vllm.vllm_sparse_delta) VllmSparseRefitConfig (class in nemo_rl.models.generation.vllm.config) VllmSparseRefitReceiver (class in nemo_rl.models.generation.vllm.vllm_sparse_refit) VllmSpecificArgs (class in nemo_rl.models.generation.vllm.config) VllmVideoConfig (class in nemo_rl.models.generation.vllm.config) VllmWeightLayout (class in nemo_rl.models.generation.vllm.refit_layout) vlm_hf_data_processor() (in module nemo_rl.data.processors) vlm_kwargs (nemo_rl.models.automodel.data.ProcessedInputs attribute) VLMEnvConfig (class in nemo_rl.environments.vlm_environment) VLMEnvironment (class in nemo_rl.environments.vlm_environment) VLMEnvironmentMetadata (class in nemo_rl.environments.vlm_environment) VLMMessageLogType (in module nemo_rl.data.interfaces) VLMVerifyWorker (class in nemo_rl.environments.vlm_environment) vocab_parallel_argmax() (in module nemo_rl.distributed.model_utils) vocab_parallel_full_log_softmax() (in module nemo_rl.distributed.model_utils) vocab_parallel_gather_columns() (in module nemo_rl.distributed.model_utils) vocab_parallel_group (nemo_rl.algorithms.loss.loss_functions.DraftCrossEntropyLossConfig attribute) vocab_parallel_log_softmax() (in module nemo_rl.distributed.model_utils) vocab_size (nemo_rl.utils.flops_formulas.FLOPSConfig attribute) vocab_topk (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute) W wait_for_pending_generations() (nemo_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollector method) wait_notification() (nemo_rl.utils.checkpoint_engines.nixl.NixlAgent method) wait_until_admissible() (nemo_rl.algorithms.async_utils.staleness_sampler._GatedSampler method) (nemo_rl.algorithms.async_utils.staleness_sampler.TransactionalAdmissionSampler method) (nemo_rl.algorithms.async_utils.staleness_sampler.WindowedSampler method) wake_carries_weight_updates() (nemo_rl.models.generation.interfaces.GenerationInterface method) (nemo_rl.models.generation.megatron.megatron_generation.MegatronGeneration method) wake_up() (nemo_rl.models.generation.vllm.vllm_worker.VllmGenerationWorkerImpl method) wake_up_async() (nemo_rl.models.generation.trtllm.trtllm_worker_async.TrtllmAsyncGenerationWorkerImpl method) (nemo_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) WALL_CLOCK_EFFICIENCY_CATEGORIES (in module nemo_rl.algorithms.utils) WALL_CLOCK_MEASUREMENT (in module nemo_rl.telemetry.metrics) wall_ms (nemo_rl.data_plane.observability.DataPlaneEvent attribute) wandb (nemo_rl.utils.logger.LoggerConfig attribute) wandb_enabled (nemo_rl.utils.logger.LoggerConfig attribute) WandbConfig (class in nemo_rl.utils.logger) WandbLogger (class in nemo_rl.utils.logger) warm_start_value_checkpoint (nemo_rl.algorithms.ppo.PPOConfig attribute) warmup_generation_lead_steps (nemo_rl.algorithms.ppo.AsyncPPOConfig attribute) warmup_lookahead_versions (nemo_rl.algorithms.async_utils.staleness_sampler.InOrderSamplerConfig attribute) warn_if_inf_grad_norm() (in module nemo_rl.utils.grad_norm) warn_on_unsupported_dataset_config_keys() (in module nemo_rl.data.datasets.utils) warn_once() (in module nemo_rl.telemetry.metrics) WASTED (nemo_rl.telemetry.instrumentation.Bucket attribute) WatchdogConfig (class in nemo_rl.algorithms.single_controller_utils.config) weight (nemo_rl.algorithms.xtoken_off_policy_distillation.TeacherConfig attribute) weight_decay (nemo_rl.models.policy.MegatronOptimizerConfig attribute) weight_decay_incr_style (nemo_rl.models.policy.MegatronSchedulerConfig attribute) weight_sync_time_s (nemo_rl.algorithms.metric_utils.SetupTimingMetrics attribute) weight_synchronizer (nemo_rl.algorithms.single_controller_utils.setup.SingleControllerActorArgs attribute) weight_version (nemo_rl.models.generation.fleet_health.ShardHealth attribute) WeightFifoSampler (class in nemo_rl.algorithms.async_utils.staleness_sampler) WeightFifoSamplerConfig (class in nemo_rl.algorithms.async_utils.staleness_sampler) WeightSynchronizer (class in nemo_rl.weight_sync.interfaces) WeightUpdateFinalizer (in module nemo_rl.models.generation.vllm.vllm_backend) WeightUpdateTransport (in module nemo_rl.models.generation.vllm.vllm_backend) WindowedSampler (class in nemo_rl.algorithms.async_utils.staleness_sampler) WindowedSamplerConfig (class in nemo_rl.algorithms.async_utils.staleness_sampler) with_fields() (nemo_rl.data_plane.interfaces.KVBatchMeta method) without_model_config() (nemo_rl.modelopt.models.policy.workers.megatron_quant_policy_worker.MegatronQuantPolicyWorker method) worker_args (nemo_rl.models.generation.dynamo.config.DynamoCfg attribute) WORKER_CLASS_DICT (nemo_rl.environments.math_environment.BaseMathEnvironment attribute) (nemo_rl.environments.math_environment.MathEnvironment attribute) (nemo_rl.environments.math_environment.MathMultiRewardEnvironment attribute) worker_group (nemo_rl.models.generation.megatron.megatron_generation.MegatronGeneration property) worker_metadata (nemo_rl.distributed.worker_groups.RayWorkerGroup property) worker_setup_time_s (nemo_rl.algorithms.metric_utils.SetupTimingMetrics attribute) workers (nemo_rl.distributed.worker_groups.RayWorkerGroup property) workers_per_shard (nemo_rl.weight_sync.membership.RefitMembership attribute) working_dir (nemo_rl.environments.code_environment.CodeEnvMetadata attribute) world_size (nemo_rl.weight_sync.membership.RefitMembership attribute) world_size() (nemo_rl.distributed.virtual_cluster.RayVirtualCluster method) wrap_loss_fn_with_input_preparation() (in module nemo_rl.algorithms.loss.wrapper) wrap_outer_model (nemo_rl.models.policy.MoEParallelizerOptions attribute) wrap_with_nvtx_name() (in module nemo_rl.utils.nsys) write_columns() (in module nemo_rl.data_plane.column_io) write_to_dataplane() (nemo_rl.data_plane.driver_mixin.TQDriverMixin method) X xferdtensor() (in module nemo_rl.weight_sync.xferdtensor) xferdtensor_golden() (in module nemo_rl.weight_sync.xferdtensor) xferdtensor_python_impl() (in module nemo_rl.weight_sync.xferdtensor_python) xtoken_loss (nemo_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute) (nemo_rl.algorithms.xtoken_off_policy_distillation.TeacherConfig attribute) xtoken_non_student_seq_keys() (in module nemo_rl.algorithms.xtoken_off_policy_distillation) xtoken_off_policy_distillation_train() (in module nemo_rl.algorithms.xtoken_off_policy_distillation) Z zero_outside_topk (nemo_rl.algorithms.loss.loss_functions.DistillationLossConfig attribute) zmq_refit_server_port (nemo_rl.models.generation.vllm.config.VllmSpecificArgs attribute) zmq_relay_forward_workers (nemo_rl.models.generation.vllm.config.VllmRefitTuningConfig attribute) zmq_relay_payload_workers (nemo_rl.models.generation.vllm.config.VllmRefitTuningConfig attribute) zmq_retries (nemo_rl.models.generation.vllm.config.VllmRefitTuningConfig attribute) ZmqSparseRefitClient (class in nemo_rl.utils.weight_transfer_zmq) ZmqSparseRefitServer (class in nemo_rl.utils.weight_transfer_zmq) zstd_compress() (in module nemo_rl.utils.weight_transfer_stream) zstd_threads (nemo_rl.models.generation.vllm.config.VllmDeltaCompressionConfig attribute)
