Index

_

__abs__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __add__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __all__ (in module nemo_automodel) (in module nemo_automodel._diffusers) (in module nemo_automodel._transformers) (in module nemo_automodel._transformers.auto_tokenizer) (in module nemo_automodel.autonvtx) (in module nemo_automodel.components.checkpoint._backports._fsspec_filesystem) (in module nemo_automodel.components.checkpoint._backports.default_planner) (in module nemo_automodel.components.checkpoint._backports.filesystem) (in module nemo_automodel.components.checkpoint._backports.hf_storage) (in module nemo_automodel.components.datasets.llm) (in module nemo_automodel.components.datasets.llm.nanogpt_dataset) (in module nemo_automodel.components.datasets.vlm) (in module nemo_automodel.components.distributed) (in module nemo_automodel.components.distributed.pipelining) (in module nemo_automodel.components.models) (in module nemo_automodel.components.models.common) (in module nemo_automodel.components.models.common.bidirectional) (in module nemo_automodel.components.models.common.combined_projection) (in module nemo_automodel.components.models.common.utils) (in module nemo_automodel.components.models.gpt2) (in module nemo_automodel.components.models.llama.rope_utils) (in module nemo_automodel.components.models.llama_bidirectional) (in module nemo_automodel.components.models.llama_bidirectional.model) (in module nemo_automodel.components.models.qwen2) (in module nemo_automodel.components.models.qwen2.model) (in module nemo_automodel.components.optim) (in module nemo_automodel.components.quantization) (in module nemo_automodel.components.quantization.qat) (in module nemo_automodel.shared) __call__() (nemo_automodel.components.datasets.llm.retrieval_collator.RetrievalBiencoderCollator method) (nemo_automodel.components.distributed.pipelining.functional.ParallelizeFnProtocol method) (nemo_automodel.components.loss.te_parallel_ce.TEParallelCrossEntropy method) (nemo_automodel.components.models.kimi_k25_vl.model.DeepSeekV3RotaryEmbeddingAdapter method) (nemo_automodel.components.models.kimivl.model.DeepSeekV3RotaryEmbeddingAdapter method) (nemo_automodel.components.training.timers.Timers method) (nemo_automodel.shared.import_utils.UnavailableMeta method) __contains__() (nemo_automodel.components.config.loader.ConfigNode method) __del__() (nemo_automodel.components.datasets.llm.megatron.indexed_dataset._IndexReader method) (nemo_automodel.components.datasets.llm.megatron.indexed_dataset._MMapBinReader method) __delete__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __delitem__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __dir__() (in module nemo_automodel) (in module nemo_automodel._transformers) (in module nemo_automodel._transformers.auto_tokenizer) __divmod__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __enter__() (nemo_automodel.components.datasets.llm.megatron.indexed_dataset._IndexWriter method) (nemo_automodel.components.distributed.utils.FirstRankPerNode method) (nemo_automodel.components.loggers.metric_logger.MetricLogger method) (nemo_automodel.components.loggers.mlflow_utils.MLflowLogger method) (nemo_automodel.components.training.rng.ScopedRNG method) (nemo_automodel.components.training.signal_handler.DistributedSignalHandler method) (nemo_automodel.components.training.timers.TimerBase method) (nemo_automodel.components.training.utils.ScopedModuleOffloading method) (nemo_automodel.shared.import_utils.UnavailableMeta method) __eq__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __exit__() (nemo_automodel.components.datasets.llm.megatron.indexed_dataset._IndexWriter method) (nemo_automodel.components.distributed.utils.FirstRankPerNode method) (nemo_automodel.components.loggers.metric_logger.MetricLogger method) (nemo_automodel.components.loggers.mlflow_utils.MLflowLogger method) (nemo_automodel.components.training.rng.ScopedRNG method) (nemo_automodel.components.training.signal_handler.DistributedSignalHandler method) (nemo_automodel.components.training.timers.TimerBase method) (nemo_automodel.components.training.utils.ScopedModuleOffloading method) __floordiv__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __ge__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __get__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __getattr__() (in module nemo_automodel) (in module nemo_automodel._transformers) (in module nemo_automodel._transformers.auto_tokenizer) (nemo_automodel.components.config.loader.ConfigNode method) (nemo_automodel.shared.import_utils.UnavailableMeta method) __getitem__() (nemo_automodel.components.datasets.llm.chat_dataset.ChatDataset method) (nemo_automodel.components.datasets.llm.column_mapped_text_instruction_dataset.ColumnMappedTextInstructionDataset method) (nemo_automodel.components.datasets.llm.column_mapped_text_instruction_iterable_dataset.ColumnMappedTextInstructionIterableDataset method) (nemo_automodel.components.datasets.llm.delta_lake_dataset.DeltaLakeDataset method) (nemo_automodel.components.datasets.llm.hellaswag.HellaSwag method) (nemo_automodel.components.datasets.llm.megatron.builder.BlendedDataset method) (nemo_automodel.components.datasets.llm.megatron.gpt_dataset.GPTDataset method) (nemo_automodel.components.datasets.llm.megatron.indexed_dataset._IndexReader method) (nemo_automodel.components.datasets.llm.megatron.indexed_dataset.IndexedDataset method) (nemo_automodel.components.datasets.llm.nanogpt_dataset.NanogptDataset method) (nemo_automodel.components.datasets.llm.seq_cls.GLUE_MRPC method) (nemo_automodel.components.datasets.reservoir_sampler.ReservoirSampler method) __getstate__() (nemo_automodel.components.checkpoint._backports.filesystem._StorageInfo method) (nemo_automodel.components.checkpoint._backports.hf_utils._HFStorageInfo method) __gt__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __hash__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __iadd__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __ifloordiv__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __ilshift__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __imul__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __index__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __invert__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __ipow__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __irshift__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __isub__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __iter__() (nemo_automodel.components.datasets.llm.column_mapped_text_instruction_dataset.ColumnMappedTextInstructionDataset method) (nemo_automodel.components.datasets.llm.column_mapped_text_instruction_iterable_dataset.ColumnMappedTextInstructionIterableDataset method) (nemo_automodel.components.datasets.llm.delta_lake_dataset._LimitedDeltaLakeDataset method) (nemo_automodel.components.datasets.llm.delta_lake_dataset.DeltaLakeDataset method) (nemo_automodel.components.datasets.llm.delta_lake_dataset.DeltaLakeIterator method) (nemo_automodel.components.datasets.llm.megatron.sampler.BaseMegatronSampler method) (nemo_automodel.components.datasets.llm.megatron.sampler.MegatronPretrainingRandomSampler method) (nemo_automodel.components.datasets.llm.megatron.sampler.MegatronPretrainingSampler method) (nemo_automodel.components.datasets.llm.mock_iterable_dataset.MockIterableDataset method) (nemo_automodel.components.datasets.llm.nanogpt_dataset.NanogptDataset method) (nemo_automodel.components.datasets.reservoir_sampler.ReservoirSampler method) (nemo_automodel.components.training.step_scheduler.StepScheduler method) (nemo_automodel.shared.import_utils.UnavailableMeta method) __itruediv__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __le__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __len__() (nemo_automodel.components.datasets.llm.chat_dataset.ChatDataset method) (nemo_automodel.components.datasets.llm.column_mapped_text_instruction_dataset.ColumnMappedTextInstructionDataset method) (nemo_automodel.components.datasets.llm.column_mapped_text_instruction_iterable_dataset.ColumnMappedTextInstructionIterableDataset method) (nemo_automodel.components.datasets.llm.delta_lake_dataset.DeltaLakeDataset method) (nemo_automodel.components.datasets.llm.hellaswag.HellaSwag method) (nemo_automodel.components.datasets.llm.megatron.builder.BlendedDataset method) (nemo_automodel.components.datasets.llm.megatron.gpt_dataset.GPTDataset method) (nemo_automodel.components.datasets.llm.megatron.indexed_dataset._IndexReader method) (nemo_automodel.components.datasets.llm.megatron.indexed_dataset.IndexedDataset method) (nemo_automodel.components.datasets.llm.megatron.sampler.BaseMegatronSampler method) (nemo_automodel.components.datasets.llm.megatron.sampler.MegatronPretrainingRandomSampler method) (nemo_automodel.components.datasets.llm.mock_iterable_dataset.MockIterableDataset method) (nemo_automodel.components.datasets.llm.nanogpt_dataset.NanogptDataset method) (nemo_automodel.components.datasets.llm.seq_cls.GLUE_MRPC method) (nemo_automodel.components.datasets.reservoir_sampler.ReservoirSampler method) (nemo_automodel.shared.import_utils.UnavailableMeta method) __lshift__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __lt__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __mul__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __ne__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __neg__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __new__() (nemo_automodel.components.config.loader._OrigValueStr method) (nemo_automodel.shared.import_utils.UnavailableMeta method) __post_init__() (nemo_automodel._transformers.registry._ModelRegistry method) (nemo_automodel.components._peft.module_matcher.ModuleMatcher method) (nemo_automodel.components.checkpoint.checkpointing.CheckpointingConfig method) (nemo_automodel.components.datasets.llm.megatron.gpt_dataset.BlendedMegatronDatasetConfig method) (nemo_automodel.components.datasets.llm.megatron.gpt_dataset.GPTDatasetConfig method) (nemo_automodel.components.distributed.config.FSDP2Config method) (nemo_automodel.components.distributed.config.MegatronFSDPConfig method) (nemo_automodel.components.distributed.mesh.MeshContext method) (nemo_automodel.components.loggers.metric_logger.MetricsSample method) (nemo_automodel.components.models.common.utils.BackendConfig method) (nemo_automodel.components.moe.config.MoEConfig method) __pow__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __radd__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __rdivmod__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __repr__() (nemo_automodel.components.config.loader.ConfigNode method) (nemo_automodel.components.optim.scheduler.OptimizerParamScheduler method) __rfloordiv__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __rlshift__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __rmul__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __rpow__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __rrshift__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __rshift__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __rsub__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __rtruediv__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __setitem__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __str__() (nemo_automodel.components.config.loader.ConfigNode method) __sub__() (nemo_automodel.shared.import_utils.UnavailableMeta method) __truediv__() (nemo_automodel.shared.import_utils.UnavailableMeta method) _add_env_storage_options() (nemo_automodel.components.datasets.llm.delta_lake_dataset.DeltaLakeIterator method) _add_nvtx_hooks() (in module nemo_automodel.autonvtx) _add_outer_prefix() (in module nemo_automodel.components.checkpoint.stateful_wrappers) _add_pad_token() (in module nemo_automodel.components.datasets.llm.formatting_utils) _add_quantization_scale_inv_tensors() (nemo_automodel.components.models.deepseek_v32.state_dict_adapter.DeepSeekV32StateDictAdapter method) _aggregate_stats() (in module nemo_automodel.components.moe.load_balance_metrics) _AllGatherConcatVarlenFn (class in nemo_automodel.components.moe.experts) _apply() (in module nemo_automodel.components.checkpoint.checkpointing) (nemo_automodel.components.models.qwen3_5_moe.model.Fp32SafeQwen3_5MoeTextRotaryEmbedding method) (nemo_automodel.components.models.qwen3_5_moe.model.Fp32SafeQwen3_5MoeVisionRotaryEmbedding method) (nemo_automodel.components.models.qwen3_vl_moe.model.Fp32SafeQwen3VLMoeTextRotaryEmbedding method) (nemo_automodel.components.models.qwen3_vl_moe.model.Fp32SafeQwen3VLMoeVisionRotaryEmbedding method) (nemo_automodel.components.moe.parallelizer.ExpertParallel method) _apply_bias() (in module nemo_automodel.components.moe.experts) _apply_key_mapping() (in module nemo_automodel.components.checkpoint.checkpointing) (nemo_automodel.components.models.gpt_oss.state_dict_adapter.GPTOSSStateDictAdapter method) (nemo_automodel.components.models.qwen3_5_moe.state_dict_adapter.Qwen3_5MoeStateDictAdapter method) (nemo_automodel.components.models.qwen3_next.state_dict_adapter.Qwen3NextStateDictAdapter method) _apply_peft_and_lower_precision() (in module nemo_automodel._transformers.infrastructure) _apply_preload_overrides() (in module nemo_automodel._transformers.kernel_patches) _apply_rope_vision() (in module nemo_automodel.components.models.kimi_k25_vl.model) (in module nemo_automodel.components.models.kimivl.model) _apply_tokenizer() (nemo_automodel.components.datasets.llm.column_mapped_text_instruction_dataset.ColumnMappedTextInstructionDataset method) _as_iter() (in module nemo_automodel.components.datasets.llm.chat_dataset) _assert_colwise_parallel() (in module nemo_automodel.components.models.common.combined_projection.combined_qkv) _assert_same_signature() (in module nemo_automodel._transformers.kernel_patches) _AsyncSaveContext (class in nemo_automodel.components.checkpoint.checkpointing) _barrier_with_timeout() (in module nemo_automodel.components.distributed.utils) _base_non_quantized_keys (nemo_automodel.components.models.deepseek_v32.state_dict_adapter.DeepSeekV32StateDictAdapter attribute) _BaseNeMoAutoModelClass (class in nemo_automodel._transformers.auto_model) _BinReader (class in nemo_automodel.components.datasets.llm.megatron.indexed_dataset) _buffer (in module nemo_automodel.components.moe.megatron.fused_a2a) _build_blended_dataset_splits() (nemo_automodel.components.datasets.llm.megatron.builder.BlendedMegatronDatasetBuilder method) _build_cache() (nemo_automodel.components.models.llama.rope_utils.LlamaRotaryEmbedding method) _build_document_index() (in module nemo_automodel.components.datasets.llm.megatron.gpt_dataset) _build_document_sample_shuffle_indices() (nemo_automodel.components.datasets.llm.megatron.gpt_dataset.GPTDataset method) _build_indices() (nemo_automodel.components.datasets.llm.megatron.builder.BlendedDataset method) _build_megatron_dataset_splits() (nemo_automodel.components.datasets.llm.megatron.builder.BlendedMegatronDatasetBuilder method) _build_megatron_datasets_parallel() (nemo_automodel.components.datasets.llm.megatron.builder.BlendedMegatronDatasetBuilder method) _build_model() (nemo_automodel._transformers.auto_model._BaseNeMoAutoModelClass class method) _build_position_ids() (in module nemo_automodel.components.distributed.cp_utils) _build_shuffle_index() (in module nemo_automodel.components.datasets.llm.megatron.gpt_dataset) _build_sparse_mask() (nemo_automodel.components.models.deepseek_v32.layers.DeepseekV32MLA method) _build_uc_table_fqn() (in module nemo_automodel.components.datasets.llm.delta_lake_dataset) _calculate_max_contiguous_elements() (in module nemo_automodel.components.checkpoint._backports.consolidate_hf_safetensors) _calculate_max_steps() (in module nemo_automodel.components.training.step_scheduler) _calculate_num_epochs() (in module nemo_automodel.components.training.step_scheduler) _can_compile_fullgraph (nemo_automodel.components.models.llama.model.LlamaPreTrainedModel attribute) (nemo_automodel.components.models.mistral3.model.Ministral3PreTrainedModel attribute) (nemo_automodel.components.models.qwen2.model.Qwen2PreTrainedModel attribute) _can_record_outputs (nemo_automodel.components.models.llama.model.LlamaPreTrainedModel attribute) (nemo_automodel.components.models.mistral3.model.Ministral3PreTrainedModel attribute) (nemo_automodel.components.models.qwen2.model.Qwen2PreTrainedModel attribute) _check_all_values_equal_length() (in module nemo_automodel.components.datasets.llm.column_mapped_text_instruction_dataset) _check_and_set() (nemo_automodel.components.optim.scheduler.OptimizerParamScheduler method) _check_box_bounds() (in module nemo_automodel.components.checkpoint._backports.default_planner) _check_box_overlap() (in module nemo_automodel.components.checkpoint._backports.default_planner) _check_databricks_sql_available() (in module nemo_automodel.components.datasets.llm.delta_lake_dataset) _check_delta_reader_available() (in module nemo_automodel.components.datasets.llm.delta_lake_dataset) _check_deltalake_available() (in module nemo_automodel.components.datasets.llm.delta_lake_dataset) _check_pyspark_available() (in module nemo_automodel.components.datasets.llm.delta_lake_dataset) _choose_device() (in module nemo_automodel._diffusers.auto_diffusion_pipeline) _clip_grad_norm_impl() (in module nemo_automodel.components.training.utils) _coerce_to_list() (in module nemo_automodel.components.datasets.llm.retrieval_dataset_inline) _COLOR_RESET (in module nemo_automodel.components.loggers.log_utils) _compiled_compute_cross_entropy (in module nemo_automodel.components.loss.chunked_ce) _compute_aux_loss() (nemo_automodel.components.moe.layers.Gate method) _compute_concentration_and_inv_freq() (nemo_automodel.components.models.gpt_oss.rope_utils.RotaryEmbedding method) _compute_cos_sin() (nemo_automodel.components.models.gpt_oss.rope_utils.RotaryEmbedding method) _compute_dataset_max_len() (nemo_automodel.components.datasets.utils.SFTSingleTurnPreprocessor method) _compute_default_inv_freq() (in module nemo_automodel.components.models.llama.rope_utils) _compute_expert_load() (nemo_automodel.components.moe.layers.Gate method) _compute_expert_utilization() (in module nemo_automodel.components.moe.load_balance_metrics) _compute_inv_freq() (nemo_automodel.components.models.step3p5.layers.Step3p5RotaryEmbedding method) _compute_llama3_inv_freq() (in module nemo_automodel.components.models.llama.rope_utils) _compute_num_image_tokens_from_grid() (nemo_automodel.components.models.kimi_k25_vl.model.KimiK25VLModel method) _compute_per_layer_stats() (in module nemo_automodel.components.moe.load_balance_metrics) _compute_utilization_aggregates() (in module nemo_automodel.components.moe.load_balance_metrics) _concatenate_expert_weights() (nemo_automodel.components.moe.state_dict_mixin.MoESplitExpertsStateDictMixin method) _configure_fsdp_module() (in module nemo_automodel.components.moe.fsdp_mixin) _consolidate_safetensors_files() (in module nemo_automodel.components.checkpoint._backports.consolidate_hf_safetensors) _consume_config_overrides() (in module nemo_automodel._transformers.model_init) _contains_usable_plan() (in module nemo_automodel.components.checkpoint._backports.planner_helpers) _convert_checkpoint_with_transformers() (in module nemo_automodel.components.checkpoint.checkpointing) _convert_dict_to_list() (nemo_automodel.components.datasets.llm.retrieval_collator.RetrievalBiencoderCollator method) _convert_lora_expert_to_hf() (nemo_automodel.components.moe.state_dict_mixin.MoESplitExpertsStateDictMixin method) _convert_moe_packed_tensors() (nemo_automodel.components.models.gpt_oss.state_dict_adapter.GPTOSSStateDictAdapter method) _convert_native_to_hf() (nemo_automodel.components.models.step3p5.state_dict_adapter.Step3p5StateDictAdapter method) _convert_single_merged_expert_to_hf_split_experts() (nemo_automodel.components.moe.state_dict_mixin.MoESplitExpertsStateDictMixin method) _convert_to_tensors() (in module nemo_automodel.components.datasets.llm.packed_sequence) _convert_tool_calls() (in module nemo_automodel.components.datasets.llm.xlam) _convert_tools() (in module nemo_automodel.components.datasets.llm.xlam) _count_parameters() (nemo_automodel.components.distributed.pipelining.autopipeline.AutoPipeline static method) _CP_STREAM (in module nemo_automodel.components.moe.parallelizer) _create_default_local_metadata() (in module nemo_automodel.components.checkpoint._backports.default_planner) _create_dtensor_from_local_or_reference() (in module nemo_automodel.components.models.step3p5.state_dict_adapter) _create_fsdp2_device_mesh() (in module nemo_automodel.components.distributed.mesh_utils) _create_global_plan() (nemo_automodel.components.checkpoint._backports.default_planner.DefaultSavePlanner method) _create_global_plan_with_caching() (nemo_automodel.components.checkpoint._backports.default_planner.DefaultSavePlanner method) _create_gloo_group() (in module nemo_automodel.components.distributed.utils) _create_megatron_fsdp_device_mesh() (in module nemo_automodel.components.distributed.mesh_utils) _create_moe_mesh() (in module nemo_automodel.components.distributed.mesh_utils) _create_transform_func() (in module nemo_automodel.components.datasets.llm.retrieval_dataset) (in module nemo_automodel.components.datasets.llm.retrieval_dataset_inline) _DATABRICKS_SQL_AVAILABLE (in module nemo_automodel.components.datasets.llm.delta_lake_dataset) _decode_single_token() (in module nemo_automodel.components.datasets.vlm.collate_fns) _dedup_save_plans() (nemo_automodel.components.checkpoint._backports.default_planner.DefaultSavePlanner method) _DeepepManager (class in nemo_automodel.components.moe.megatron.token_dispatcher) _deepstack_process() (nemo_automodel.components.models.qwen3_omni_moe.model.Qwen3OmniMoeThinkerTextModel method) (nemo_automodel.components.models.qwen3_vl_moe.model.Qwen3VLMoeTextModelBackend method) _DEFAULT_STRATEGY (in module nemo_automodel.components.distributed.parallelizer) _deinterleave_gate_up() (nemo_automodel.components.models.common.combined_projection.state_dict_adapter.CombinedProjectionStateDictAdapter method) _deinterleave_qkv() (nemo_automodel.components.models.common.combined_projection.state_dict_adapter.CombinedProjectionStateDictAdapter method) _DELTALAKE_AVAILABLE (in module nemo_automodel.components.datasets.llm.delta_lake_dataset) _dequantize() (nemo_automodel.components.models.deepseek_v3.state_dict_adapter.DeepSeekV3StateDictAdapter method) (nemo_automodel.components.models.minimax_m2.state_dict_adapter.MiniMaxM2StateDictAdapter method) _dequantize_block_scale_tensors() (nemo_automodel.components.models.gpt_oss.state_dict_adapter.GPTOSSStateDictAdapter method) _dequantize_with_torch() (in module nemo_automodel.components.models.deepseek_v3.state_dict_adapter) _dequantize_with_triton() (in module nemo_automodel.components.models.deepseek_v3.state_dict_adapter) _derived_version (in module nemo_automodel.components.checkpoint._backports._version) _DISABLE_FN_BY_MODE (in module nemo_automodel.components.quantization.qat) _disable_fsdp_for_moe_module() (in module nemo_automodel.components.moe.fsdp_mixin) _DispatchManager (class in nemo_automodel.components.moe.megatron.token_dispatcher) _distribute_param() (in module nemo_automodel.components.distributed.parallel_styles) _do_load() (nemo_automodel.components.checkpoint.checkpointing.Checkpointer method) _do_save() (nemo_automodel.components.checkpoint.checkpointing.Checkpointer method) _done (nemo_automodel.components.checkpoint._backports.filesystem._OverlappingCpuLoader property) _dora_weight_norm() (nemo_automodel.components._peft.lora.LinearLoRA method) _download_model_weights() (in module nemo_automodel._transformers.model_init) _dp_axis_names() (nemo_automodel.components.distributed.mesh.MeshContext method) _drain() (nemo_automodel.components.checkpoint._backports.filesystem._OverlappingCpuLoader method) _drop_outer_prefix() (in module nemo_automodel.components.checkpoint.stateful_wrappers) _DTYPE_TO_SAFETENSORS_DTYPE (in module nemo_automodel.components.checkpoint._backports.filesystem) _EmptyStateDictLoadPlanner (class in nemo_automodel.components.checkpoint._backports.default_planner) _ENABLE_FN_BY_MODE (in module nemo_automodel.components.quantization.qat) _encode() (nemo_automodel._transformers.biencoder.BiencoderModel method) _ensure_dirs() (in module nemo_automodel.components.checkpoint.checkpointing) _ensure_rgb() (in module nemo_automodel.components.datasets.vlm.collate_fns) _ensure_root_handler_with_formatter() (in module nemo_automodel.components.loggers.log_utils) _enum_representer() (in module nemo_automodel.components.utils.yaml_utils) _equally_divide_layers() (in module nemo_automodel.components.checkpoint.checkpointing) _expand_image_tokens() (in module nemo_automodel.components.datasets.vlm.collate_fns) _expand_quantized_keys() (nemo_automodel.components.models.kimi_k25_vl.state_dict_adapter.KimiK25VLStateDictAdapter method) _expert_path_segment (nemo_automodel.components.models.minimax_m2.state_dict_adapter.MiniMaxM2StateDictAdapter property) (nemo_automodel.components.models.nemotron_v3.state_dict_adapter.NemotronV3StateDictAdapter property) (nemo_automodel.components.moe.state_dict_mixin.MoESplitExpertsStateDictMixin property) _extract_assistant_text() (in module nemo_automodel.components.datasets.vlm.collate_fns) _extract_file_index() (in module nemo_automodel.components.checkpoint._backports.hf_storage) _extract_image_features() (nemo_automodel.components.models.kimi_k25_vl.model.KimiK25VLModel method) (nemo_automodel.components.models.kimivl.model.KimiVLModel method) _extract_model_layers() (in module nemo_automodel.components.distributed.parallelizer) _extract_target_modules() (in module nemo_automodel.components.checkpoint.addons) _FileBinReader (class in nemo_automodel.components.datasets.llm.megatron.indexed_dataset) _FileSystemWriter (class in nemo_automodel.components.checkpoint._backports.filesystem) _fill_labels_with_cross_entropy_ignore_idx() (in module nemo_automodel.components.datasets.llm.packed_sequence) _filter_kwargs_for_init() (in module nemo_automodel._transformers.model_init) _filter_meta_device_from_init_context() (in module nemo_automodel._transformers.model_init) _find_largest_module_list() (in module nemo_automodel.components.distributed.parallelizer) _find_next_bos_with_index() (in module nemo_automodel.components.datasets.llm.nanogpt_dataset) _find_pattern_indices() (in module nemo_automodel.components.datasets.vlm.collate_fns) _finish() (nemo_automodel.components.checkpoint._backports.filesystem._OverlappingCpuLoader method) _finish_plan_with_caching() (nemo_automodel.components.checkpoint._backports.default_planner.DefaultSavePlanner method) _fixed_block_mask_mod() (nemo_automodel.components.attention.flex_attention.FlexAttention static method) _FLASH_ATTENTION_FIX_APPLIED (in module nemo_automodel.components.utils.compile_utils) _format_example() (in module nemo_automodel.components.datasets.llm.xlam) _formatting_prompts_func() (in module nemo_automodel.components.datasets.llm.squad) _formatting_prompts_func_with_chat_template() (in module nemo_automodel.components.datasets.llm.squad) _forward_grouped_mm() (nemo_automodel.components.moe.experts.GroupedExperts method) _forward_loop() (nemo_automodel.components.moe.experts.GroupedExperts method) _FqnData (class in nemo_automodel.components.checkpoint._backports.consolidate_hf_safetensors) _freeze_module_by_attribute_and_patterns() (in module nemo_automodel.components.utils.model_utils) _from_config_parent_class() (nemo_automodel._transformers.auto_model._BaseNeMoAutoModelClass class method) _from_hf_w_merged_experts() (nemo_automodel.components.moe.state_dict_mixin.MoESplitExpertsStateDictMixin method) _from_pretrained_parent_class() (nemo_automodel._transformers.auto_model._BaseNeMoAutoModelClass class method) _fully_shard() (in module nemo_automodel.components.distributed.parallelizer_utils) _function_representer() (in module nemo_automodel.components.utils.yaml_utils) _gather_1d_if_needed() (nemo_automodel.components.models.common.combined_projection.state_dict_adapter.CombinedProjectionStateDictAdapter static method) _gen_file_name() (in module nemo_automodel.components.checkpoint._backports.hf_utils) _generate_uuid() (in module nemo_automodel.components.checkpoint._backports.filesystem) _generation_config_representer() (in module nemo_automodel.components.utils.yaml_utils) _get_active_ranges() (in module nemo_automodel.autonvtx)

_get_all_ranks_time_string() (nemo_automodel.components.training.timers.Timers method) _get_automodel_peft_metadata() (in module nemo_automodel.components.checkpoint.addons) _get_axis_size() (in module nemo_automodel.components.distributed.mesh) _get_base_model_state_dict() (nemo_automodel.components.checkpoint.stateful_wrappers.ModelState method) _get_block_causal_mask_mod() (nemo_automodel.components.attention.flex_attention.FlexAttention static method) _get_causal_mask_mod() (nemo_automodel.components.attention.flex_attention.FlexAttention static method) _get_cp_stream() (in module nemo_automodel.components.moe.parallelizer) _get_dcp_custom_metadata() (in module nemo_automodel.components.checkpoint._backports.hf_utils) _get_dion_mesh() (in module nemo_automodel.components.optim.utils) _get_dtype() (in module nemo_automodel.components.checkpoint._backports.hf_utils) _get_dtype_from_val() (in module nemo_automodel.components.datasets.llm.nanogpt_dataset) _get_elapsed_time_all_ranks() (nemo_automodel.components.training.timers.Timers method) _get_file_iterator() (nemo_automodel.components.datasets.llm.nanogpt_dataset.NanogptDataset method) _get_global_min_max_time() (nemo_automodel.components.training.timers.Timers method) _get_global_min_max_time_string() (nemo_automodel.components.training.timers.Timers method) _get_hf_meta_device_disabled() (in module nemo_automodel._transformers.model_init) _get_hf_peft_config() (in module nemo_automodel.components.checkpoint.addons) _get_init_param_names() (in module nemo_automodel._transformers.model_init) _get_key_renaming_mapping() (in module nemo_automodel.components.checkpoint._backports.hf_storage) _get_llama_4_attn_scale() (in module nemo_automodel.components.models.mistral3.model) _get_lm_head_weight_and_name() (in module nemo_automodel.components.checkpoint.stateful_wrappers) _get_logical_numel() (in module nemo_automodel.components.utils.model_utils) _get_ltor_masks_and_position_ids() (in module nemo_automodel.components.datasets.llm.megatron.gpt_dataset) _get_mixin_wrapped_class() (in module nemo_automodel._transformers.model_init) _get_model_param_stats() (in module nemo_automodel.components.utils.model_utils) _get_model_type() (in module nemo_automodel._transformers.auto_tokenizer) _get_module_from_path() (in module nemo_automodel.components.distributed.parallelizer_utils) _get_next_bos_position() (in module nemo_automodel.components.datasets.llm.nanogpt_dataset) _get_next_fallback_attn() (in module nemo_automodel._transformers.kernel_patches) _get_num_epochs() (nemo_automodel.components.datasets.llm.megatron.gpt_dataset.GPTDataset method) _get_num_tokens_per_epoch() (nemo_automodel.components.datasets.llm.megatron.gpt_dataset.GPTDataset method) _get_original_model_path() (nemo_automodel.components.checkpoint.checkpointing.Checkpointer method) _get_parallel_plan() (in module nemo_automodel.components.distributed.parallelizer) _get_peft_state_dict() (in module nemo_automodel.components.checkpoint.stateful_wrappers) _get_right_trailing_pad_mask() (in module nemo_automodel.components.datasets.llm.formatting_utils) _get_rng_state() (in module nemo_automodel.components.training.rng) _get_rope_config() (in module nemo_automodel.components.models.llama.rope_utils) _get_safetensors_file_metadata() (in module nemo_automodel.components.checkpoint._backports.hf_utils) _get_size_per_split_per_dataset() (in module nemo_automodel.components.datasets.llm.megatron.builder) _get_sliding_window_mask_mod() (nemo_automodel.components.attention.flex_attention.FlexAttention static method) _get_spark_session() (in module nemo_automodel.components.datasets.llm.delta_lake_dataset) _get_stacked_bias() (nemo_automodel.components.moe.experts.GroupedExpertsTE method) _get_stacked_weight() (nemo_automodel.components.moe.experts.GroupedExpertsTE method) _get_start_end_pos_single_file() (in module nemo_automodel.components.datasets.llm.nanogpt_dataset) _get_storage_reader() (nemo_automodel.components.checkpoint.checkpointing.Checkpointer method) _get_storage_writer() (nemo_automodel.components.checkpoint.checkpointing.Checkpointer method) _get_tokenizer_registry() (in module nemo_automodel._transformers.auto_tokenizer) _get_worker_id_and_total_workers() (in module nemo_automodel.components.datasets.llm.nanogpt_dataset) _group_params_by_dtype() (in module nemo_automodel.components.distributed.parallelizer_utils) _has_chat_template() (in module nemo_automodel.components.datasets.llm.formatting_utils) _has_cuda_capability() (in module nemo_automodel.components.quantization.fp8) _has_expert_parallelism() (in module nemo_automodel.components.checkpoint.stateful_wrappers) _has_quantized_params() (in module nemo_automodel.components.checkpoint.stateful_wrappers) _hf_key_to_native() (nemo_automodel.components.models.minimax_m2.state_dict_adapter.MiniMaxM2StateDictAdapter method) _hf_meta_device_disabled (in module nemo_automodel._transformers.model_init) _hf_prefix (nemo_automodel.components.models.nemotron_v3.state_dict_adapter.NemotronV3StateDictAdapter property) (nemo_automodel.components.models.step3p5.state_dict_adapter.Step3p5StateDictAdapter property) (nemo_automodel.components.moe.state_dict_mixin.MoESplitExpertsStateDictMixin property) _HFStorageInfo (class in nemo_automodel.components.checkpoint._backports.hf_utils) _HuggingFaceStorageReader (class in nemo_automodel.components.checkpoint._backports.hf_storage) _HuggingFaceStorageWriter (class in nemo_automodel.components.checkpoint._backports.hf_storage) _hybrid_model_flops() (in module nemo_automodel.components.utils.flops_utils) _import_error (in module nemo_automodel.components.optim.utils) _INDEX_HEADER (in module nemo_automodel.components.datasets.llm.megatron.indexed_dataset) _indexer_non_quantized_keys (nemo_automodel.components.models.deepseek_v32.state_dict_adapter.DeepSeekV32StateDictAdapter attribute) _IndexReader (class in nemo_automodel.components.datasets.llm.megatron.indexed_dataset) _IndexWriter (class in nemo_automodel.components.datasets.llm.megatron.indexed_dataset) _indices_to_multihot() (nemo_automodel.components.moe.megatron.token_dispatcher._DeepepManager method) _indices_to_multihot_kernel() (in module nemo_automodel.components.moe.megatron.fused_indices_converter) _init_adapter() (nemo_automodel.components._peft.lora.LinearLoRA static method) (nemo_automodel.components._peft.lora_moe.GroupedExpertsDeepEPLoRA static method) (nemo_automodel.components._peft.lora_moe.GroupedExpertsLoRA static method) _init_model() (in module nemo_automodel._transformers.model_init) _init_peft_adapters() (in module nemo_automodel.components.checkpoint.checkpointing) _init_weights() (in module nemo_automodel.components.moe.experts) (in module nemo_automodel.components.moe.layers) (nemo_automodel.components.models.gpt2.GPT2LMHeadModel method) (nemo_automodel.components.models.nemotron_parse.model.NemotronParsePreTrainedModel method) _initialize_metadata() (nemo_automodel.components.moe.megatron.token_dispatcher.MoEFlexTokenDispatcher method) _InputFileData (class in nemo_automodel.components.checkpoint._backports.consolidate_hf_safetensors) _instantiate_distributed() (in module nemo_automodel._transformers.infrastructure) _instantiate_pipeline() (in module nemo_automodel._transformers.infrastructure) _instantiate_qat() (in module nemo_automodel._transformers.infrastructure) _instantiate_value() (nemo_automodel.components.config.loader.ConfigNode method) _interleave_gate_up() (nemo_automodel.components.models.common.combined_projection.state_dict_adapter.CombinedProjectionStateDictAdapter method) _interleave_qkv() (nemo_automodel.components.models.common.combined_projection.state_dict_adapter.CombinedProjectionStateDictAdapter method) _is_allowed_module() (in module nemo_automodel.components.config.loader) _is_config_compatible_with_custom_model() (in module nemo_automodel._transformers.model_init) _is_custom_model() (in module nemo_automodel.components.checkpoint.checkpointing) _is_deletion_vectors_error() (in module nemo_automodel.components.datasets.llm.delta_lake_dataset) _is_enabled_index() (nemo_automodel.components.datasets.llm.megatron.builder.BlendedMegatronDatasetBuilder method) _is_gated_moe (nemo_automodel.components.moe.state_dict_mixin.MoESplitExpertsStateDictMixin property) _is_geq_torch_2_9() (in module nemo_automodel.components.checkpoint.checkpointing) _is_hf_repo_id() (in module nemo_automodel.components.datasets.llm.chat_dataset) _is_linear_module() (in module nemo_automodel.components._peft.module_matcher) _is_location_overlap_error() (in module nemo_automodel.components.datasets.llm.delta_lake_dataset) _is_nvshmem_available() (in module nemo_automodel.components.moe.megatron.fused_a2a) _is_quantized_expert_key() (nemo_automodel.components.models.kimi_k25_vl.state_dict_adapter.KimiK25VLStateDictAdapter method) _is_quantized_module() (in module nemo_automodel.components.checkpoint.stateful_wrappers) _is_safe_attr() (in module nemo_automodel.components.config.loader) _is_safe_path() (in module nemo_automodel.components.config.loader) _is_safetensors_checkpoint() (in module nemo_automodel.components.checkpoint.checkpointing) _is_stateful (nemo_automodel.components.models.nemotron_v3.model.NemotronHForCausalLM attribute) _is_transformers_v5_or_higher() (in module nemo_automodel.components.distributed.parallelizer) _is_unity_catalog_path() (in module nemo_automodel.components.datasets.llm.delta_lake_dataset) _item_size() (in module nemo_automodel.components.checkpoint._backports.filesystem) _iter_all_rows() (nemo_automodel.components.datasets.llm.delta_lake_dataset.DeltaLakeIterator method) _iter_fsdp_modules() (in module nemo_automodel.components.moe.fsdp_mixin) _iter_pipeline_modules() (in module nemo_automodel._diffusers.auto_diffusion_pipeline) _iter_with_databricks_sql() (nemo_automodel.components.datasets.llm.delta_lake_dataset.DeltaLakeIterator method) _iter_with_deltalake() (nemo_automodel.components.datasets.llm.delta_lake_dataset.DeltaLakeIterator method) _iter_with_spark() (nemo_automodel.components.datasets.llm.delta_lake_dataset.DeltaLakeIterator method) _json_load_if_str() (in module nemo_automodel.components.datasets.llm.xlam) _key_config_attributes() (nemo_automodel.components.datasets.llm.megatron.gpt_dataset.GPTDataset static method) _LAZY_ATTRS (in module nemo_automodel) (in module nemo_automodel._transformers) _LEVEL_TO_COLOR (in module nemo_automodel.components.loggers.log_utils) _LimitedDeltaLakeDataset (class in nemo_automodel.components.datasets.llm.delta_lake_dataset) _load_bos_index() (in module nemo_automodel.components.datasets.llm.nanogpt_dataset) _load_dataset() (in module nemo_automodel.components.datasets.llm.column_mapped_text_instruction_dataset) _load_from_state_dict() (nemo_automodel.components.moe.experts.GroupedExpertsTE method) _load_full_state_dict_into_model() (in module nemo_automodel.components.checkpoint.checkpointing) _load_hf_checkpoint_preserving_dtype() (in module nemo_automodel.components.checkpoint.checkpointing) _load_json_or_jsonl() (in module nemo_automodel.components.datasets.llm.retrieval_dataset_inline) _load_openai_messages() (in module nemo_automodel.components.datasets.llm.chat_dataset) _load_streaming_dataset() (in module nemo_automodel.components.datasets.llm.column_mapped_text_instruction_iterable_dataset) _logger (in module nemo_automodel.shared.torch_patches) _make_causal_mask() (nemo_automodel.components.models.nemotron_v3.model.NemotronHForCausalLM static method) _make_lazy_te_patcher() (in module nemo_automodel.components.models.common.utils) _make_missing() (in module nemo_automodel.components.models.qwen3_5_moe.model) _mamba_layer_flops() (in module nemo_automodel.components.utils.flops_utils) _mapping_model_arch_name_to_cls() (nemo_automodel._transformers.registry._ModelRegistry method) _masked_split_matrix() (nemo_automodel.components.datasets.llm.megatron.builder.BlendedMegatronDatasetBuilder method) _maybe_adapt_state_dict_from_hf() (in module nemo_automodel.components.checkpoint.checkpointing) _maybe_adapt_state_dict_to_hf() (in module nemo_automodel.components.checkpoint.checkpointing) _maybe_build_consolidated_index() (nemo_automodel.components.checkpoint.checkpointing.Checkpointer method) _maybe_save_custom_model_code() (in module nemo_automodel.components.checkpoint.addons) _merge_batch_dict() (nemo_automodel.components.datasets.llm.retrieval_collator.RetrievalBiencoderCollator method) _merge_input_ids_with_image_features() (nemo_automodel.components.models.kimi_k25_vl.model.KimiK25VLModel method) _merge_with_image_features() (nemo_automodel.components.models.kimivl.model.KimiVLModel method) _metadata_fn (in module nemo_automodel.components.checkpoint._backports.filesystem) (in module nemo_automodel.components.checkpoint._backports.hf_utils) _mlp() (nemo_automodel.components.models.deepseek_v3.model.Block method) (nemo_automodel.components.models.glm4_moe.model.Block method) (nemo_automodel.components.models.glm4_moe_lite.model.Block method) (nemo_automodel.components.models.gpt_oss.model.Block method) (nemo_automodel.components.models.qwen3_moe.model.Block method) (nemo_automodel.components.models.qwen3_next.model.Block method) _MMapBinReader (class in nemo_automodel.components.datasets.llm.megatron.indexed_dataset) _model_has_dtensors() (in module nemo_automodel.components.checkpoint.checkpointing) _ModelRegistry (class in nemo_automodel._transformers.registry) _module_filter_fn() (in module nemo_automodel.components.quantization.fp8) _move_module_to_device() (in module nemo_automodel._diffusers.auto_diffusion_pipeline) _move_to_cpu() (nemo_automodel.components.loggers.metric_logger.MetricLogger method) _multihot_to_indices_kernel() (in module nemo_automodel.components.moe.megatron.fused_indices_converter) _native_key_to_hf() (nemo_automodel.components.models.minimax_m2.state_dict_adapter.MiniMaxM2StateDictAdapter method) _nemotronh_mlp_layer_flops() (in module nemo_automodel.components.utils.flops_utils) _nemotronh_moe_layer_flops() (in module nemo_automodel.components.utils.flops_utils) _no_split_modules (nemo_automodel.components.models.kimi_k25_vl.model.KimiK25VLForConditionalGeneration attribute) (nemo_automodel.components.models.llama.model.LlamaPreTrainedModel attribute) (nemo_automodel.components.models.mistral3.model.Ministral3PreTrainedModel attribute) (nemo_automodel.components.models.nemotron_parse.model.NemotronParsePreTrainedModel attribute) (nemo_automodel.components.models.qwen2.model.Qwen2PreTrainedModel attribute) _non_mla_attn_layer_flops() (in module nemo_automodel.components.utils.flops_utils) _non_quantized_keys (nemo_automodel.components.models.deepseek_v32.state_dict_adapter.DeepSeekV32StateDictAdapter property) _norm() (nemo_automodel.components.models.qwen3_next.layers.Qwen3NextRMSNorm method) _normalize_delta_path() (in module nemo_automodel.components.datasets.llm.delta_lake_dataset) _normalize_inline_doc() (in module nemo_automodel.components.datasets.llm.retrieval_dataset_inline) _normalize_messages() (in module nemo_automodel.components.datasets.llm.chat_dataset) _normalize_moe_mesh() (nemo_automodel.components.moe.experts.GroupedExpertsTE method) _normalize_prefix() (in module nemo_automodel.components.datasets.llm.megatron.indexed_dataset) _nvshmem_available (in module nemo_automodel.components.moe.megatron.fused_a2a) _optional_axis() (in module nemo_automodel.components.distributed.mesh) _original_get_init_context (in module nemo_automodel._transformers.model_init) _OrigValueStr (class in nemo_automodel.components.config.loader) _OutputFileData (class in nemo_automodel.components.checkpoint._backports.consolidate_hf_safetensors) _OverlappingCpuLoader (class in nemo_automodel.components.checkpoint._backports.filesystem) _package_tokenized_example() (in module nemo_automodel.components.datasets.llm.formatting_utils) _pad_function() (nemo_automodel.components.datasets.utils.SFTSingleTurnPreprocessor method) _pad_pack() (in module nemo_automodel.components.datasets.llm.packed_sequence) _pad_to_seq_length() (in module nemo_automodel.components.datasets.llm.formatting_utils) _PAD_TOKEN_ID (in module nemo_automodel.components.datasets.llm.megatron.gpt_dataset) _parallelize_gemma3() (in module nemo_automodel.components.distributed.optimized_tp_plans) _parallelize_llama() (in module nemo_automodel.components.distributed.optimized_tp_plans) _parallelize_ministral3() (in module nemo_automodel.components.distributed.optimized_tp_plans) _parallelize_phi3() (in module nemo_automodel.components.distributed.optimized_tp_plans) _parallelize_qwen() (in module nemo_automodel.components.distributed.optimized_tp_plans) _parallelize_qwen_classification() (in module nemo_automodel.components.distributed.optimized_tp_plans) _parse_args() (in module nemo_automodel.components.models.llama_bidirectional.export_onnx) _parse_input_metadata() (in module nemo_automodel.components.checkpoint._backports.consolidate_hf_safetensors) _parse_unity_storage_ids() (in module nemo_automodel.components.datasets.llm.delta_lake_dataset) _partial_representer() (in module nemo_automodel.components.utils.yaml_utils) _partition_embedding_fn() (nemo_automodel.components.distributed.parallel_styles.ColwiseParallelLora method) (nemo_automodel.components.distributed.parallel_styles.RowwiseParallelLora method) _partition_fn() (nemo_automodel.components.moe.parallelizer.ExpertParallel method) _partition_linear_fn() (nemo_automodel.components.distributed.parallel_styles.ColwiseParallelLora method) (nemo_automodel.components.distributed.parallel_styles.RowwiseParallelLora method) _patch_attention() (in module nemo_automodel._transformers.kernel_patches) _patch_bytes_to_unicode() (in module nemo_automodel._transformers.utils) _patch_liger_kernel() (in module nemo_automodel._transformers.kernel_patches) _patch_special_tokens_pattern() (in module nemo_automodel._transformers.utils) _patch_te_modules (in module nemo_automodel.components.models.common.utils) _patched_get_init_context() (in module nemo_automodel._transformers.model_init) _peek_num_tokens() (in module nemo_automodel.components.datasets.llm.nanogpt_dataset) _PEFT_PREFIX (nemo_automodel.components.models.common.bidirectional.BiencoderStateDictAdapter attribute) _permute_tokens_for_grouped_mm() (in module nemo_automodel.components.moe.experts) _pool() (in module nemo_automodel.components.models.llama_bidirectional.model) _Pooling (class in nemo_automodel.components.models.llama_bidirectional.export_onnx) _pp_plan (nemo_automodel.components.models.llama.model.LlamaForCausalLM attribute) (nemo_automodel.components.models.mistral3.model.Ministral3ForCausalLM attribute) (nemo_automodel.components.models.qwen2.model.Qwen2ForCausalLM attribute) _pre_shard_combined_projections() (in module nemo_automodel.components.distributed.parallelizer) _precompute_freqs_cis() (nemo_automodel.components.models.kimi_k25_vl.model.Rope2DPosEmbRepeated method) (nemo_automodel.components.models.kimivl.model.Rope2DPosEmb method) _precompute_stage_shapes() (in module nemo_automodel.components.distributed.pipelining.functional) _PREFIX (in module nemo_automodel.components.checkpoint.stateful_wrappers) _prepare_input_fn() (nemo_automodel.components.distributed.optimized_tp_plans.RotaryEmbedParallel static method) (nemo_automodel.components.distributed.optimized_tp_plans.VocabParallelEmbedding static method) _prepare_output_fn() (nemo_automodel.components.distributed.optimized_tp_plans.RotaryEmbedParallel static method) (nemo_automodel.components.distributed.optimized_tp_plans.SequenceParallelAllGatherActivation static method) (nemo_automodel.components.distributed.optimized_tp_plans.VocabParallelEmbedding static method) _process_file_tokens() (nemo_automodel.components.datasets.llm.nanogpt_dataset.NanogptDataset method) _process_output_file() (in module nemo_automodel.components.checkpoint._backports.consolidate_hf_safetensors) _PYSPARK_AVAILABLE (in module nemo_automodel.components.datasets.llm.delta_lake_dataset) _QUANTIZER_TO_MODE (in module nemo_automodel.components.quantization.qat) _query_document_sample_shuffle_indices() (nemo_automodel.components.datasets.llm.megatron.gpt_dataset.GPTDataset method) _quote_sql_ident() (in module nemo_automodel.components.datasets.llm.delta_lake_dataset) _read_tensor_data_mmap() (in module nemo_automodel.components.checkpoint._backports.consolidate_hf_safetensors) _recombine_lora_expert_keys() (nemo_automodel.components.moe.state_dict_mixin.MoESplitExpertsStateDictMixin method) _recombine_split_projection_keys() (nemo_automodel.components.models.common.combined_projection.state_dict_adapter.CombinedProjectionStateDictAdapter method) _redact() (in module nemo_automodel.components.config.loader) _refill() (nemo_automodel.components.checkpoint._backports.filesystem._OverlappingCpuLoader method) _register_kimi_k25_vl_with_transformers() (in module nemo_automodel.components.models.kimi_k25_vl.model) _register_kimi_vl_with_transformers() (in module nemo_automodel.components.models.kimivl.model) _register_ministral3_with_transformers() (in module nemo_automodel.components.models.mistral3.model) _registry (nemo_automodel._transformers.auto_tokenizer.NeMoAutoTokenizer attribute) _reinit_rope_buffers() (in module nemo_automodel.components.checkpoint.checkpointing) _rename_dora_keys_from_hf() (in module nemo_automodel.components.checkpoint.stateful_wrappers) _rename_dora_keys_to_hf() (in module nemo_automodel.components.checkpoint.stateful_wrappers) _reorder_cache() (nemo_automodel.components.models.nemotron_parse.model.NemotronParseForConditionalGeneration method) _replicate_module_fn() (nemo_automodel.components.distributed.parallel_styles.SequenceParallelLora method) _repr_value() (nemo_automodel.components.config.loader.ConfigNode method) _resolve_doc_to_example() (in module nemo_automodel.components.datasets.llm.retrieval_dataset_inline) _resolve_target() (in module nemo_automodel.components.config.loader) _resolve_uc_table_from_unity_storage_path() (in module nemo_automodel.components.datasets.llm.delta_lake_dataset) _restore_rng_state() (in module nemo_automodel.components.training.rng) _rotate_activation() (in module nemo_automodel.components.models.deepseek_v32.layers) _run_post_backward_for_moe_module() (in module nemo_automodel.components.moe.fsdp_mixin) _run_post_backward_hooks() (in module nemo_automodel.components.moe.fsdp_mixin) _safe_object_representer() (in module nemo_automodel.components.utils.yaml_utils) _save() (nemo_automodel.components.loggers.metric_logger.MetricLogger method) _separate_param_groups() (in module nemo_automodel.components.optim.utils) _sequence_pointers() (nemo_automodel.components.datasets.llm.megatron.indexed_dataset._IndexWriter method) _SerialCpuLoader (class in nemo_automodel.components.checkpoint._backports.filesystem) _set_base_model_state_dict() (nemo_automodel.components.checkpoint.stateful_wrappers.ModelState method) _set_peft_state_dict() (in module nemo_automodel.components.checkpoint.stateful_wrappers) _set_stacked_bias() (nemo_automodel.components.moe.experts.GroupedExpertsTE method) _set_stacked_weight() (nemo_automodel.components.moe.experts.GroupedExpertsTE method) _setup_distributed() (nemo_automodel.components.distributed.ddp.DDPManager method) _setup_worker_context() (nemo_automodel.components.datasets.llm.nanogpt_dataset.NanogptDataset method) _shard_ep_fsdp() (in module nemo_automodel._transformers.infrastructure) _shard_pp() (in module nemo_automodel._transformers.infrastructure) _shard_thd_chunk_for_te() (in module nemo_automodel.components.distributed.cp_utils) _shared_experts_stream (in module nemo_automodel.components.moe.layers) _should_include_key() (nemo_automodel.components.checkpoint._backports.default_planner._EmptyStateDictLoadPlanner method) _should_load_before_shard() (in module nemo_automodel._transformers.utils) _should_stop_packing() (in module nemo_automodel.components.datasets.llm.packed_sequence) _should_write_consolidated_safetensors() (nemo_automodel.components.checkpoint.checkpointing.Checkpointer method) _should_write_hf_metadata() (nemo_automodel.components.checkpoint.checkpointing.Checkpointer method) _skip_keys_device_placement (nemo_automodel.components.models.llama.model.LlamaPreTrainedModel attribute) (nemo_automodel.components.models.mistral3.model.Ministral3PreTrainedModel attribute) (nemo_automodel.components.models.nemotron_parse.model.NemotronParsePreTrainedModel attribute) (nemo_automodel.components.models.qwen2.model.Qwen2PreTrainedModel attribute) _slice_file() (nemo_automodel.components.checkpoint._backports.filesystem.FileSystemReader method) _slice_scale_for_dtensor() (in module nemo_automodel.components.models.deepseek_v3.state_dict_adapter) _split_and_add_pack() (in module nemo_automodel.components.datasets.llm.packed_sequence) _split_by_size_and_type() (in module nemo_automodel.components.checkpoint._backports.filesystem) _split_by_storage_plan() (nemo_automodel.components.checkpoint._backports.hf_storage._HuggingFaceStorageWriter method) _split_experts_weights() (nemo_automodel.components.moe.state_dict_mixin.MoESplitExpertsStateDictMixin method) _split_remaining_combined_projection_keys() (nemo_automodel.components.models.common.combined_projection.state_dict_adapter.CombinedProjectionStateDictAdapter method) _StorageInfo (class in nemo_automodel.components.checkpoint._backports.filesystem) _StoragePrefix (class in nemo_automodel.components.checkpoint._backports.filesystem) _StorageReaderTransforms (class in nemo_automodel.components.checkpoint._backports.filesystem) _StorageWriterTransforms (class in nemo_automodel.components.checkpoint._backports.filesystem) _str_is_hf_repo_id() (in module nemo_automodel.components.datasets.llm.column_mapped_text_instruction_dataset) _stream_supports_color() (nemo_automodel.components.loggers.log_utils.ColorFormatter method) _SUBMODULES (in module nemo_automodel) _supports_attention_backend (nemo_automodel.components.models.llama.model.LlamaPreTrainedModel attribute) (nemo_automodel.components.models.mistral3.model.Ministral3PreTrainedModel attribute) (nemo_automodel.components.models.qwen2.model.Qwen2PreTrainedModel attribute) _supports_flash_attn (nemo_automodel.components.models.llama.model.LlamaPreTrainedModel attribute) (nemo_automodel.components.models.mistral3.model.Ministral3PreTrainedModel attribute) (nemo_automodel.components.models.qwen2.model.Qwen2PreTrainedModel attribute) _supports_flex_attn (nemo_automodel.components.models.llama.model.LlamaPreTrainedModel attribute) (nemo_automodel.components.models.mistral3.model.Ministral3PreTrainedModel attribute) (nemo_automodel.components.models.qwen2.model.Qwen2PreTrainedModel attribute) _supports_logits_to_keep() (in module nemo_automodel.components.utils.model_utils) _supports_sdpa (nemo_automodel.components.models.llama.model.LlamaPreTrainedModel attribute) (nemo_automodel.components.models.mistral3.model.Ministral3PreTrainedModel attribute) (nemo_automodel.components.models.qwen2.model.Qwen2PreTrainedModel attribute) _supports_seq_lens() (in module nemo_automodel.components.utils.model_utils) _swap_key() (nemo_automodel.components.models.common.bidirectional.BiencoderStateDictAdapter static method) _TE_EXPERT_PARAM_PATTERN (in module nemo_automodel.components.training.utils) _tensorize_and_pad_pack() (in module nemo_automodel.components.datasets.llm.packed_sequence) _TensorLoader (class in nemo_automodel.components.checkpoint._backports.filesystem) _thread_local (in module nemo_automodel.autonvtx) _tied_weights_keys (nemo_automodel.components.models.llama.model.LlamaForCausalLM attribute) (nemo_automodel.components.models.mistral3.model.Ministral3ForCausalLM attribute) (nemo_automodel.components.models.qwen2.model.Qwen2ForCausalLM attribute) _to_dotted_path() (nemo_automodel.components.config.loader.ConfigNode method) _to_ep_dtensor() (nemo_automodel.components.moe.experts.GroupedExpertsTE method) _to_hf_w_split_experts() (nemo_automodel.components.moe.state_dict_mixin.MoESplitExpertsStateDictMixin method) _to_safetensors_dtype_str() (in module nemo_automodel.components.checkpoint._backports.filesystem) _tokenize_function() (nemo_automodel.components.datasets.utils.SFTSingleTurnPreprocessor method) _torch_dtype_representer() (in module nemo_automodel.components.utils.yaml_utils) _torch_mm_experts_fwd() (in module nemo_automodel.components.moe.experts) _TORCH_PATCHES_APPLIED (in module nemo_automodel.shared.torch_patches) _tp_plan (nemo_automodel.components.models.llama.model.LlamaForCausalLM attribute) (nemo_automodel.components.models.mistral3.model.Ministral3ForCausalLM attribute) (nemo_automodel.components.models.qwen2.model.Qwen2ForCausalLM attribute) _transform_func() (in module nemo_automodel.components.datasets.llm.retrieval_dataset) (in module nemo_automodel.components.datasets.llm.retrieval_dataset_inline) _TRANSFORMERS_AVAILABLE (in module nemo_automodel.components.checkpoint.conversion_mapping) _try_resolve_uc_table_from_system_tables() (in module nemo_automodel.components.datasets.llm.delta_lake_dataset) _TYPE_MAP (in module nemo_automodel.components.datasets.llm.xlam) _UNITY_STORAGE_TABLE_PATH_RE (in module nemo_automodel.components.datasets.llm.delta_lake_dataset) _unpack_doc_values() (in module nemo_automodel.components.datasets.llm.retrieval_collator) _unwrap() (nemo_automodel.components.config.loader.ConfigNode method) _update_causal_mask() (nemo_automodel.components.models.llama_bidirectional.model.LlamaBidirectionalModel method) _VALID_AXIS_NAMES (in module nemo_automodel.components.distributed.mesh) _validate_distributed_setup() (in module nemo_automodel.components.distributed.mesh) _validate_expert_availability() (nemo_automodel.components.moe.state_dict_mixin.MoESplitExpertsStateDictMixin method) _validate_global_plan() (in module nemo_automodel.components.checkpoint._backports.default_planner) _validate_mesh_dim_names() (in module nemo_automodel.components.distributed.mesh) _VERBOSE (in module nemo_automodel.components.datasets.llm.megatron.builder) _verify_sdpa_support() (in module nemo_automodel._transformers.kernel_patches) _warned_add_pad_token (in module nemo_automodel.components.datasets.llm.formatting_utils) _wrap() (nemo_automodel.components.config.loader.ConfigNode method) _write_data() (in module nemo_automodel.components.checkpoint._backports.consolidate_hf_safetensors) (nemo_automodel.components.checkpoint._backports.filesystem._FileSystemWriter method) _write_files_from_queue() (in module nemo_automodel.components.checkpoint._backports.filesystem) _write_item() (in module nemo_automodel.components.checkpoint._backports.filesystem) _write_metadata() (in module nemo_automodel.components.checkpoint._backports.consolidate_hf_safetensors) _write_overall_metadata_file() (in module nemo_automodel.components.checkpoint._backports.consolidate_hf_safetensors) _write_overall_metadata_file_from_shards() (in module nemo_automodel.components.checkpoint._backports.consolidate_hf_safetensors) _write_sub_tensor_to_file_optimized() (in module nemo_automodel.components.checkpoint._backports.consolidate_hf_safetensors)

A

AbstractDataset (class in nemo_automodel.components.datasets.llm.retrieval_dataset) activation_alpha (nemo_automodel.components.moe.config.MoEConfig attribute) activation_checkpointing (nemo_automodel.components.distributed.config.DDPConfig attribute) (nemo_automodel.components.distributed.config.FSDP2Config attribute) (nemo_automodel.components.distributed.config.MegatronFSDPConfig attribute) (nemo_automodel.components.distributed.mesh.MeshContext attribute) activation_limit (nemo_automodel.components.moe.config.MoEConfig attribute) active_time() (nemo_automodel.components.training.timers.DummyTimer method) (nemo_automodel.components.training.timers.Timer method) add() (nemo_automodel.components.checkpoint._backports.filesystem._OverlappingCpuLoader method) (nemo_automodel.components.checkpoint._backports.filesystem._SerialCpuLoader method) (nemo_automodel.components.checkpoint._backports.filesystem._TensorLoader method) add_causal_masks_to_batch() (in module nemo_automodel.components.datasets.utils) add_corpus() (in module nemo_automodel.components.datasets.llm.retrieval_dataset) add_document() (nemo_automodel.components.datasets.llm.megatron.indexed_dataset.IndexedDatasetBuilder method) add_extra_token_to_sequence (nemo_automodel.components.datasets.llm.megatron.gpt_dataset.GPTDatasetConfig attribute) add_filter_to_all_loggers() (in module nemo_automodel.components.loggers.log_utils) add_index() (nemo_automodel.components.datasets.llm.megatron.indexed_dataset.IndexedDatasetBuilder method) add_item() (nemo_automodel.components.datasets.llm.megatron.indexed_dataset.IndexedDatasetBuilder method) all_gather_item() (in module nemo_automodel.components.training.signal_handler) ALLOWED_IMPORT_PREFIXES (in module nemo_automodel.components.config.loader) alpha (nemo_automodel.components._peft.lora.PeftConfig attribute) Answer (nemo_automodel.components.datasets.llm.column_mapped_text_instruction_dataset.ColumnTypes attribute) apply_ac() (in module nemo_automodel.components.moe.parallelizer) apply_async_checkpoint_patch() (in module nemo_automodel.components.checkpoint._torch_backports)

apply_cache_compatibility_patches() (in module nemo_automodel._transformers.utils) apply_cp() (in module nemo_automodel.components.moe.parallelizer) apply_ep() (in module nemo_automodel.components.moe.parallelizer) apply_flash_attention_compile_fix() (in module nemo_automodel.components.utils.compile_utils) apply_fp8_to_model() (in module nemo_automodel.components.quantization.fp8) apply_fsdp() (in module nemo_automodel.components.moe.parallelizer) apply_fsdp2_sharding_recursively() (in module nemo_automodel.components.distributed.parallelizer) apply_lora_to_linear_modules() (in module nemo_automodel.components._peft.lora) apply_model_infrastructure() (in module nemo_automodel._transformers.infrastructure) apply_parameter_freezing() (in module nemo_automodel.components.utils.model_utils) apply_patches() (in module nemo_automodel.components.checkpoint._torch_backports) apply_qwen3_omni_config_patch() (in module nemo_automodel._transformers.utils) apply_rotary_emb() (in module nemo_automodel.components.models.deepseek_v3.rope_utils) (in module nemo_automodel.components.models.gpt_oss.rope_utils) apply_rotary_emb_qk() (in module nemo_automodel.components.models.deepseek_v3.rope_utils) (in module nemo_automodel.components.models.gpt_oss.rope_utils) apply_rotary_pos_emb() (in module nemo_automodel.components.models.llama.rope_utils) (in module nemo_automodel.components.models.mistral3.model) apply_torch_patches() (in module nemo_automodel.shared.torch_patches) async_wait() (nemo_automodel.components.checkpoint.checkpointing.Checkpointer method) attention_flops_calculator() (in module nemo_automodel.components.utils.flops_utils) attn (nemo_automodel.components.models.common.utils.BackendConfig attribute) AutoPipeline (class in nemo_automodel.components.distributed.pipelining.autopipeline) aux_loss_coeff (nemo_automodel.components.moe.config.MoEConfig attribute) average_in_collective (nemo_automodel.components.distributed.config.MegatronFSDPConfig attribute)

B

backend (nemo_automodel.components.distributed.config.DDPConfig attribute) (nemo_automodel.components.distributed.config.FSDP2Config attribute) (nemo_automodel.components.distributed.config.MegatronFSDPConfig attribute) (nemo_automodel.components.distributed.init_utils.DistInfo attribute) (nemo_automodel.components.utils.compile_utils.CompileConfig attribute) BackendConfig (class in nemo_automodel.components.models.common.utils) backward() (nemo_automodel.components._peft.lora.LoRATritonFunction static method) (nemo_automodel.components.loss.te_parallel_ce.CrossEntropyFunction static method) (nemo_automodel.components.moe.experts._AllGatherConcatVarlenFn static method) (nemo_automodel.components.moe.megatron.fused_a2a.FusedCombine static method) (nemo_automodel.components.moe.megatron.fused_a2a.FusedDispatch static method) (nemo_automodel.components.moe.megatron.fused_indices_converter.IndicesToMultihot static method) (nemo_automodel.components.moe.megatron.moe_utils.MoEAuxLossAutoScaler static method) (nemo_automodel.components.moe.megatron.moe_utils.WeightedBiasQuickGeGLUFunction static method) (nemo_automodel.components.moe.megatron.moe_utils.WeightedQuickGeGLUFunction static method) (nemo_automodel.components.moe.megatron.moe_utils.WeightedSwiGLUFunction static method) barrier_and_log() (in module nemo_automodel.components.distributed.utils) base_model_pp_plan (nemo_automodel.components.models.mistral3.model.Ministral3Config attribute) base_model_prefix (nemo_automodel.components.models.kimi_k25_vl.model.KimiK25VLForConditionalGeneration attribute) (nemo_automodel.components.models.llama.model.LlamaPreTrainedModel attribute) (nemo_automodel.components.models.mistral3.model.Ministral3PreTrainedModel attribute) (nemo_automodel.components.models.nemotron_parse.model.NemotronParsePreTrainedModel attribute) (nemo_automodel.components.models.qwen2.model.Qwen2PreTrainedModel attribute) base_model_tp_plan (nemo_automodel.components.models.mistral3.model.Ministral3Config attribute) BaseMegatronSampler (class in nemo_automodel.components.datasets.llm.megatron.sampler) batchify() (in module nemo_automodel.components.datasets.utils) bert_flops() (in module nemo_automodel.components.utils.flops_utils) BiencoderModel (class in nemo_automodel._transformers.biencoder) BiencoderStateDictAdapter (class in nemo_automodel.components.models.common.bidirectional) blend (nemo_automodel.components.datasets.llm.megatron.gpt_dataset.BlendedMegatronDatasetConfig attribute) blend_per_split (nemo_automodel.components.datasets.llm.megatron.gpt_dataset.BlendedMegatronDatasetConfig attribute)

BlendedDataset (class in nemo_automodel.components.datasets.llm.megatron.builder) BlendedMegatronDatasetBuilder (class in nemo_automodel.components.datasets.llm.megatron.builder) BlendedMegatronDatasetConfig (class in nemo_automodel.components.datasets.llm.megatron.gpt_dataset) Block (class in nemo_automodel.components.models.deepseek_v3.model) (class in nemo_automodel.components.models.glm4_moe.model) (class in nemo_automodel.components.models.glm4_moe_lite.model) (class in nemo_automodel.components.models.gpt_oss.model) (class in nemo_automodel.components.models.minimax_m2.model) (class in nemo_automodel.components.models.qwen3_moe.model) (class in nemo_automodel.components.models.qwen3_next.model) (class in nemo_automodel.components.models.step3p5.model) block_masks (nemo_automodel.components.attention.flex_attention.FlexAttention attribute) BLOCK_SIZE (in module nemo_automodel.components.models.deepseek_v3.state_dict_adapter) block_vector_mul() (in module nemo_automodel.components._peft.lora_kernel) build() (nemo_automodel._transformers.biencoder.BiencoderModel class method) (nemo_automodel.components.datasets.llm.megatron.builder.BlendedMegatronDatasetBuilder method) (nemo_automodel.components.datasets.llm.megatron_dataset.MegatronPretraining method) (nemo_automodel.components.distributed.pipelining.autopipeline.AutoPipeline method) build_compile_config() (in module nemo_automodel.components.utils.compile_utils) build_dion_optimizer() (in module nemo_automodel.components.optim.utils) build_fp8_config() (in module nemo_automodel.components.quantization.fp8) build_generic_dataset() (nemo_automodel.components.datasets.llm.megatron.builder.BlendedMegatronDatasetBuilder static method) build_gpt2_model() (in module nemo_automodel.components.models.gpt2) build_labels() (in module nemo_automodel.components.datasets.vlm.collate_fns) build_low_level_dataset() (nemo_automodel.components.datasets.llm.megatron.gpt_dataset.GPTDataset static method) build_metric_logger() (in module nemo_automodel.components.loggers.metric_logger) build_mlflow() (in module nemo_automodel.components.loggers.mlflow_utils) build_packed_dataset() (in module nemo_automodel.components.datasets.llm.mock_packed) build_pipeline_schedule() (in module nemo_automodel.components.distributed.pipelining.functional) build_recipe() (nemo_automodel.components.models.common.utils.TEFp8Config method) build_sample_idx() (in module nemo_automodel.components.datasets.llm.megatron.helpers) build_unpacked_dataset() (in module nemo_automodel.components.datasets.llm.mock)

C

calculate_mfu() (in module nemo_automodel.components.utils.flops_utils) calculate_per_token_loss (nemo_automodel.components.distributed.config.MegatronFSDPConfig attribute) calculate_scale_shape() (in module nemo_automodel.components.models.deepseek_v3.state_dict_adapter) calculate_virtual_stages() (in module nemo_automodel.components.distributed.pipelining.functional) CausalSelfAttention (class in nemo_automodel.components.models.gpt2) ChatDataset (class in nemo_automodel.components.datasets.llm.chat_dataset) check_for_nan_in_grad (nemo_automodel.components.distributed.config.MegatronFSDPConfig attribute) check_model_inputs (in module nemo_automodel.components.models.llama.model) (in module nemo_automodel.components.models.qwen2.model) checkpoint_dir (nemo_automodel.components.checkpoint.checkpointing.CheckpointingConfig attribute) checkpoint_id (nemo_automodel.components.checkpoint._backports.filesystem._FileSystemWriter property) (nemo_automodel.components.checkpoint._backports.filesystem.FileSystemReader property) CheckpointAddon (class in nemo_automodel.components.checkpoint.addons) Checkpointer (class in nemo_automodel.components.checkpoint.checkpointing) CheckpointingConfig (class in nemo_automodel.components.checkpoint.checkpointing) ChunkedCrossEntropy (class in nemo_automodel.components.loss.chunked_ce) clip_grad_by_total_norm_() (in module nemo_automodel.components.distributed.grad_utils) clip_grad_norm() (in module nemo_automodel.components.training.utils) clip_vit_l_flops() (in module nemo_automodel.components.utils.flops_utils) close() (nemo_automodel.components.checkpoint.checkpointing.Checkpointer method) (nemo_automodel.components.loggers.metric_logger.MetricLogger method) code_from_dtype() (nemo_automodel.components.datasets.llm.megatron.indexed_dataset.DType class method) COLLATE_FNS (in module nemo_automodel.components.datasets.vlm.collate_fns) collect() (nemo_automodel.components.training.garbage_collection.GarbageCollection static method) collect_expert_loads() (in module nemo_automodel.components.moe.load_balance_metrics) ColorFormatter (class in nemo_automodel.components.loggers.log_utils) ColumnMappedTextInstructionDataset (class in nemo_automodel.components.datasets.llm.column_mapped_text_instruction_dataset) ColumnMappedTextInstructionIterableDataset (class in nemo_automodel.components.datasets.llm.column_mapped_text_instruction_iterable_dataset) ColumnTypes (class in nemo_automodel.components.datasets.llm.column_mapped_text_instruction_dataset) ColwiseParallelLora (class in nemo_automodel.components.distributed.parallel_styles) combine() (nemo_automodel.components.moe.megatron.token_dispatcher._DeepepManager method) (nemo_automodel.components.moe.megatron.token_dispatcher._DispatchManager method) combine_all_to_all() (nemo_automodel.components.moe.megatron.token_dispatcher.MoEFlexTokenDispatcher method) combine_postprocess() (nemo_automodel.components.moe.megatron.token_dispatcher.MoEFlexTokenDispatcher method) combine_preprocess() (nemo_automodel.components.moe.megatron.token_dispatcher.MoEFlexTokenDispatcher method) CombinedGateUpMLP (class in nemo_automodel.components.models.common.combined_projection.combined_mlp) CombinedProjectionStateDictAdapter (class in nemo_automodel.components.models.common.combined_projection.state_dict_adapter) CombinedQKVAttentionMixin (class in nemo_automodel.components.models.common.combined_projection.combined_qkv) commit_tensor() (nemo_automodel.components.checkpoint._backports.default_planner.DefaultLoadPlanner method) compile_helper() (in module nemo_automodel.components.datasets.llm.megatron.megatron_utils) compile_model() (in module nemo_automodel.components.utils.compile_utils) CompileConfig (class in nemo_automodel.components.utils.compile_utils) compute_brief_metrics() (in module nemo_automodel.components.moe.load_balance_metrics) compute_cross_entropy() (in module nemo_automodel.components.loss.chunked_ce) compute_default_rope_parameters() (nemo_automodel.components.models.mistral3.model.Ministral3RotaryEmbedding static method) compute_detailed_metrics() (in module nemo_automodel.components.moe.load_balance_metrics) compute_expanded_seq_length() (in module nemo_automodel.components.models.kimi_k25_vl.model) compute_qkv() (nemo_automodel.components.models.common.combined_projection.combined_qkv.CombinedQKVAttentionMixin method) concat_path() (nemo_automodel.components.checkpoint._backports._fsspec_filesystem.FileSystem method) (nemo_automodel.components.checkpoint._backports.filesystem.FileSystem method) (nemo_automodel.components.checkpoint._backports.filesystem.FileSystemBase method) config (nemo_automodel.components.models.mistral3.model.Ministral3PreTrainedModel attribute) config_class (nemo_automodel.components.models.kimi_k25_vl.model.KimiK25VLForConditionalGeneration attribute) (nemo_automodel.components.models.llama.model.LlamaPreTrainedModel attribute) (nemo_automodel.components.models.llama_bidirectional.model.LlamaBidirectionalForSequenceClassification attribute) (nemo_automodel.components.models.llama_bidirectional.model.LlamaBidirectionalModel attribute) (nemo_automodel.components.models.nemotron_parse.model.NemotronParsePreTrainedModel attribute) (nemo_automodel.components.models.qwen2.model.Qwen2PreTrainedModel attribute)

config_to_yaml_str() (in module nemo_automodel.components.config.loader) ConfigNode (class in nemo_automodel.components.config.loader) configure_torch_dynamo() (in module nemo_automodel.components.utils.compile_utils) consolidate_safetensors_files() (in module nemo_automodel.components.checkpoint._backports.consolidate_hf_safetensors) consolidate_safetensors_files_on_every_rank() (in module nemo_automodel.components.checkpoint._backports.consolidate_hf_safetensors) ConsolidatedHFAddon (class in nemo_automodel.components.checkpoint.addons) Context (nemo_automodel.components.datasets.llm.column_mapped_text_instruction_dataset.ColumnTypes attribute) convert_single_tensor_to_hf() (nemo_automodel.components.checkpoint.state_dict_adapter.StateDictAdapter method) (nemo_automodel.components.models.common.bidirectional.BiencoderStateDictAdapter method) (nemo_automodel.components.models.deepseek_v3.state_dict_adapter.DeepSeekV3StateDictAdapter method) (nemo_automodel.components.models.deepseek_v32.state_dict_adapter.DeepSeekV32StateDictAdapter method) (nemo_automodel.components.models.glm4_moe.state_dict_adapter.Glm4MoeStateDictAdapter method) (nemo_automodel.components.models.gpt_oss.state_dict_adapter.GPTOSSStateDictAdapter method) (nemo_automodel.components.models.kimi_k25_vl.state_dict_adapter.KimiK25VLStateDictAdapter method) (nemo_automodel.components.models.minimax_m2.state_dict_adapter.MiniMaxM2StateDictAdapter method) (nemo_automodel.components.models.nemotron_v3.state_dict_adapter.NemotronV3StateDictAdapter method) (nemo_automodel.components.models.qwen3_5_moe.state_dict_adapter.Qwen3_5MoeStateDictAdapter method) (nemo_automodel.components.models.qwen3_moe.state_dict_adapter.Qwen3MoeStateDictAdapter method) (nemo_automodel.components.models.qwen3_next.state_dict_adapter.Qwen3NextStateDictAdapter method) (nemo_automodel.components.models.qwen3_omni_moe.state_dict_adapter.Qwen3OmniMoeStateDictAdapter method) (nemo_automodel.components.models.qwen3_vl_moe.state_dict_adapter.Qwen3VLMoeStateDictAdapter method) (nemo_automodel.components.models.step3p5.state_dict_adapter.Step3p5StateDictAdapter method) convert_split_vector_to_split_matrix() (in module nemo_automodel.components.datasets.llm.megatron.gpt_dataset) corpus (nemo_automodel.components.datasets.llm.retrieval_dataset.CorpusInfo attribute) corpus_id (nemo_automodel.components.datasets.llm.retrieval_dataset.CorpusInfo property) CorpusInfo (class in nemo_automodel.components.datasets.llm.retrieval_dataset) count_tail_padding() (in module nemo_automodel.components.training.utils) CP (nemo_automodel.components.distributed.mesh.MeshAxisName attribute) cp_size (nemo_automodel.components.distributed.mesh.MeshContext property) create_attention_mask (nemo_automodel.components.datasets.llm.megatron.gpt_dataset.GPTDatasetConfig attribute) create_block_causal_mask() (in module nemo_automodel.components.datasets.llm.packed_sequence) create_bnb_config() (in module nemo_automodel.components.quantization.qlora) create_causal_mask_mapping() (in module nemo_automodel.components.datasets.utils) create_compile_config_from_dict() (in module nemo_automodel.components.utils.compile_utils) create_context_parallel_ctx() (in module nemo_automodel.components.distributed.cp_utils) create_default_global_load_plan() (in module nemo_automodel.components.checkpoint._backports.default_planner) create_default_global_save_plan() (in module nemo_automodel.components.checkpoint._backports.default_planner) create_default_local_load_plan() (in module nemo_automodel.components.checkpoint._backports.default_planner) create_default_local_save_plan() (in module nemo_automodel.components.checkpoint._backports.default_planner) create_device_mesh() (in module nemo_automodel.components.distributed.mesh_utils) create_dtensor_from_local() (in module nemo_automodel.components.moe.state_dict_utils) create_fp8_config_from_dict() (in module nemo_automodel.components.quantization.fp8) create_global_plan() (nemo_automodel.components.checkpoint._backports.default_planner.DefaultLoadPlanner method) (nemo_automodel.components.checkpoint._backports.default_planner.DefaultSavePlanner method) create_local_plan() (nemo_automodel.components.checkpoint._backports.default_planner.DefaultLoadPlanner method) (nemo_automodel.components.checkpoint._backports.default_planner.DefaultSavePlanner method) create_megatron_sampler() (in module nemo_automodel.components.datasets.llm.megatron.sampler) create_pipeline_forward_causal_lm() (in module nemo_automodel.components.distributed.pipelining.hf_utils) create_pipeline_forward_inner() (in module nemo_automodel.components.distributed.pipelining.hf_utils) create_quantizer() (nemo_automodel.components.quantization.qat.QATConfig method) create_scale_inv_for_weight() (in module nemo_automodel.components.models.deepseek_v3.state_dict_adapter) create_stream() (nemo_automodel.components.checkpoint._backports._fsspec_filesystem.FileSystem method) (nemo_automodel.components.checkpoint._backports.filesystem.FileSystem method) (nemo_automodel.components.checkpoint._backports.filesystem.FileSystemBase method) CROSS_ENTROPY_IGNORE_IDX (in module nemo_automodel.components.datasets.llm.packed_sequence) CrossEntropyFunction (class in nemo_automodel.components.loss.te_parallel_ce) cuda_rng_state (nemo_automodel.components.training.rng.RNGState attribute) CUSTOM_METADATA_KEY (in module nemo_automodel.components.checkpoint._backports.hf_utils)

D

da_dx_autotune_configs() (in module nemo_automodel.components._peft.lora_kernel) DATA_KEY (in module nemo_automodel.components.checkpoint._backports.hf_utils) DATA_OFFSETS_KEY (in module nemo_automodel.components.checkpoint._backports.hf_utils) DATASETS (in module nemo_automodel.components.datasets.llm.retrieval_dataset) db_autotune_configs() (in module nemo_automodel.components._peft.lora_kernel) DCP_SHARDING_INFO_KEY (in module nemo_automodel.components.checkpoint._backports.hf_utils) DCP_VERSION_KEY (in module nemo_automodel.components.checkpoint._backports.hf_utils) DDPConfig (class in nemo_automodel.components.distributed.config) DDPManager (class in nemo_automodel.components.distributed.ddp) debug_summary() (nemo_automodel.components.distributed.pipelining.autopipeline.AutoPipeline method) DeepseekV32Block (class in nemo_automodel.components.models.deepseek_v32.model) DeepseekV32Config (class in nemo_automodel.components.models.deepseek_v32.config) DeepseekV32ForCausalLM (class in nemo_automodel.components.models.deepseek_v32.model) DeepseekV32Indexer (class in nemo_automodel.components.models.deepseek_v32.layers) DeepseekV32MLA (class in nemo_automodel.components.models.deepseek_v32.layers) DeepseekV32Model (class in nemo_automodel.components.models.deepseek_v32.model) DeepSeekV32StateDictAdapter (class in nemo_automodel.components.models.deepseek_v32.state_dict_adapter) deepseekv3_flops() (in module nemo_automodel.components.utils.flops_utils) DeepseekV3ForCausalLM (class in nemo_automodel.components.models.deepseek_v3.model) DeepseekV3Model (class in nemo_automodel.components.models.deepseek_v3.model) DeepSeekV3RotaryEmbeddingAdapter (class in nemo_automodel.components.models.kimi_k25_vl.model) (class in nemo_automodel.components.models.kimivl.model) DeepSeekV3StateDictAdapter (class in nemo_automodel.components.models.deepseek_v3.state_dict_adapter) DEFAULT_ATTN_IMPLEMENTATION (in module nemo_automodel._transformers.kernel_patches) default_collate_fn() (in module nemo_automodel.components.datasets.vlm.collate_fns) default_collater() (in module nemo_automodel.components.datasets.utils) DEFAULT_EXTRA_METADATA_KEY (in module nemo_automodel.components.checkpoint._backports.hf_utils) default_stop_tokens() (in module nemo_automodel.components.datasets.vlm.utils) DEFAULT_SUFFIX (in module nemo_automodel.components.checkpoint._backports.filesystem) DefaultLoadPlanner (class in nemo_automodel.components.checkpoint._backports.default_planner) DefaultParallelizationStrategy (class in nemo_automodel.components.distributed.parallelizer) DefaultSavePlanner (class in nemo_automodel.components.checkpoint._backports.default_planner) defer_fsdp_grad_sync (nemo_automodel.components.distributed.config.FSDP2Config attribute) DeltaLakeDataset (class in nemo_automodel.components.datasets.llm.delta_lake_dataset) DeltaLakeIterator (class in nemo_automodel.components.datasets.llm.delta_lake_dataset) dequantize_base_checkpoint (nemo_automodel.components.checkpoint.checkpointing.CheckpointingConfig attribute) dequantize_from_fp8() (in module nemo_automodel.components.models.deepseek_v3.state_dict_adapter) dequantize_int4() (in module nemo_automodel.components.models.kimi_k25_vl.state_dict_adapter) destroy_global_state() (in module nemo_automodel.components.distributed.init_utils) detailed_every_steps (nemo_automodel.components.moe.config.MoEMetricsConfig attribute) device (nemo_automodel.components.distributed.init_utils.DistInfo attribute) (nemo_automodel.components.distributed.pipelining.autopipeline.AutoPipeline property) (nemo_automodel.components.models.nemotron_v3.model.NemotronHForCausalLM property) device_mesh (nemo_automodel.components.distributed.mesh.MeshContext attribute)

dim (nemo_automodel.components._peft.lora.PeftConfig attribute) (nemo_automodel.components.moe.config.MoEConfig attribute) disable_bucketing (nemo_automodel.components.distributed.config.MegatronFSDPConfig attribute) dispatch() (nemo_automodel.components.moe.megatron.token_dispatcher._DeepepManager method) (nemo_automodel.components.moe.megatron.token_dispatcher._DispatchManager method) dispatch_all_to_all() (nemo_automodel.components.moe.megatron.token_dispatcher.MoEFlexTokenDispatcher method) dispatch_postprocess() (nemo_automodel.components.moe.megatron.token_dispatcher.MoEFlexTokenDispatcher method) dispatch_preprocess() (nemo_automodel.components.moe.megatron.token_dispatcher.MoEFlexTokenDispatcher method) dispatch_preprocess2() (nemo_automodel.components.moe.megatron.token_dispatcher.MoEFlexTokenDispatcher method) dispatcher (nemo_automodel.components.models.common.utils.BackendConfig attribute) DistInfo (class in nemo_automodel.components.distributed.init_utils) DistributedConfig (in module nemo_automodel.components.distributed.config) DistributedSignalHandler (class in nemo_automodel.components.training.signal_handler) document_indices (nemo_automodel.components.datasets.llm.megatron.indexed_dataset.IndexedDataset property) down_proj_bias (nemo_automodel.components.moe.experts.GroupedExpertsTE property) down_projs (nemo_automodel.components.moe.experts.GroupedExpertsTE property) DP (nemo_automodel.components.distributed.mesh.MeshAxisName attribute) DP_CP (nemo_automodel.components.distributed.mesh.MeshAxisName attribute) DP_REPLICATE (nemo_automodel.components.distributed.mesh.MeshAxisName attribute) dp_replicate_size (nemo_automodel.components.distributed.mesh.MeshContext property) DP_SHARD (nemo_automodel.components.distributed.mesh.MeshAxisName attribute) DP_SHARD_CP (nemo_automodel.components.distributed.mesh.MeshAxisName attribute) dp_size (nemo_automodel.components.distributed.mesh.MeshContext property) drop_last_partial_validation_sequence (nemo_automodel.components.datasets.llm.megatron.gpt_dataset.GPTDatasetConfig attribute) dropout (nemo_automodel.components._peft.lora.PeftConfig attribute) dropout_position (nemo_automodel.components._peft.lora.PeftConfig attribute) DType (class in nemo_automodel.components.datasets.llm.megatron.indexed_dataset) dtype (nemo_automodel.components.checkpoint._backports.hf_utils._HFStorageInfo attribute) (nemo_automodel.components.distributed.pipelining.config.PipelineConfig attribute) (nemo_automodel.components.models.kimi_k25_vl.model.KimiK25VLForConditionalGeneration property) (nemo_automodel.components.models.kimi_k25_vl.model.MoonViT3dPretrainedModel property) (nemo_automodel.components.models.kimivl.model.KimiVLForConditionalGeneration property) (nemo_automodel.components.models.kimivl.model.MoonVitPretrainedModel property) (nemo_automodel.components.models.nemotron_v3.model.NemotronHForCausalLM property) (nemo_automodel.components.moe.config.MoEConfig attribute) dtype_from_code() (nemo_automodel.components.datasets.llm.megatron.indexed_dataset.DType class method) dtype_from_str() (in module nemo_automodel.shared.utils) DTYPE_KEY (in module nemo_automodel.components.checkpoint._backports.hf_utils) DTYPE_MAP (in module nemo_automodel.components.checkpoint._backports.hf_utils) dtype_size (nemo_automodel.components.checkpoint._backports.consolidate_hf_safetensors._FqnData attribute) dtype_str (nemo_automodel.components.checkpoint._backports.consolidate_hf_safetensors._FqnData attribute) DummyTimer (class in nemo_automodel.components.training.timers) dynamic (nemo_automodel.components.utils.compile_utils.CompileConfig attribute) dynamo_cache_size_limit (nemo_automodel.components.utils.compile_utils.CompileConfig attribute)

E

eager_attention_forward() (in module nemo_automodel.components.models.mistral3.model) elapsed() (nemo_automodel.components.training.timers.DummyTimer method) (nemo_automodel.components.training.timers.Timer method) (nemo_automodel.components.training.timers.TimerBase method) embed_tokens (nemo_automodel.components.models.kimi_k25_vl.model.KimiK25VLLanguageModelBackend property) (nemo_automodel.components.models.kimi_k25_vl.model.KimiK25VLModel property) (nemo_automodel.components.models.kimivl.model.KimiVLLanguageModelBackend property) (nemo_automodel.components.models.kimivl.model.KimiVLModel property) (nemo_automodel.components.models.qwen3_5_moe.model.Qwen3_5MoeModel property) (nemo_automodel.components.models.qwen3_vl_moe.model.Qwen3VLMoeModel property) EmbeddingModelForExport (class in nemo_automodel.components.models.llama_bidirectional.export_onnx) emulate (nemo_automodel.components.quantization.fp8.FP8Config attribute) enable_deepep (nemo_automodel.components.models.common.utils.BackendConfig attribute) enable_fsdp_float8_all_gather (nemo_automodel.components.quantization.fp8.FP8Config attribute) enable_fsdp_optimizations (nemo_automodel.components.models.common.utils.BackendConfig attribute) enable_hf_state_dict_adapter (nemo_automodel.components.models.common.utils.BackendConfig attribute) enable_load_balance_tracking() (in module nemo_automodel.components.moe.load_balance_metrics) enable_torch_dynamo_scalar_outputs() (in module nemo_automodel.components.utils.compile_utils) ENABLE_USER_MODULES (in module nemo_automodel.components.config.loader) enabled (nemo_automodel.components.checkpoint.checkpointing.CheckpointingConfig attribute) (nemo_automodel.components.distributed.pipelining.autopipeline.PipelineInfo attribute) (nemo_automodel.components.moe.config.MoEMetricsConfig attribute) (nemo_automodel.components.quantization.fp8.FP8Config attribute) (nemo_automodel.components.utils.compile_utils.CompileConfig attribute)

encode() (nemo_automodel._transformers.biencoder.BiencoderModel method) end_document() (nemo_automodel.components.datasets.llm.megatron.indexed_dataset.IndexedDatasetBuilder method) eod_mask_loss (nemo_automodel.components.datasets.llm.megatron.gpt_dataset.GPTDatasetConfig attribute) EP (nemo_automodel.components.distributed.mesh.MeshAxisName attribute) EP_SHARD (nemo_automodel.components.distributed.mesh.MeshAxisName attribute) ep_size (nemo_automodel.components.distributed.mesh.MeshContext property) epoch (nemo_automodel.components.loggers.metric_logger.MetricsSample attribute) epochs (nemo_automodel.components.training.step_scheduler.StepScheduler property) EXAMPLE_TEMPLATE (in module nemo_automodel.components.datasets.llm.retrieval_dataset) exclude_modules (nemo_automodel.components._peft.lora.PeftConfig attribute) (nemo_automodel.components._peft.module_matcher.ModuleMatcher attribute) exists() (nemo_automodel.components.checkpoint._backports._fsspec_filesystem.FileSystem method) (nemo_automodel.components.checkpoint._backports.filesystem.FileSystem method) (nemo_automodel.components.checkpoint._backports.filesystem.FileSystemBase method) (nemo_automodel.components.datasets.llm.megatron.indexed_dataset.IndexedDataset static method) expert_activation (nemo_automodel.components.moe.config.MoEConfig attribute) expert_bias (nemo_automodel.components.moe.config.MoEConfig attribute) ExpertParallel (class in nemo_automodel.components.moe.parallelizer) experts (nemo_automodel.components.models.common.utils.BackendConfig attribute) export_to_onnx() (in module nemo_automodel.components.models.llama_bidirectional.export_onnx) extra_repr() (nemo_automodel.components.models.qwen3_next.layers.Qwen3NextRMSNorm method) extract_key_from_dicts() (in module nemo_automodel.components.datasets.utils)

F

fake_balanced_gate (nemo_automodel.components.models.common.utils.BackendConfig attribute) fake_gate_noise (nemo_automodel.components.models.common.utils.BackendConfig attribute) FakeBalancedGate (class in nemo_automodel.components.moe.layers) FILE_NAME (in module nemo_automodel.components.checkpoint._backports.hf_utils) FileSystem (class in nemo_automodel.components.checkpoint._backports._fsspec_filesystem) (class in nemo_automodel.components.checkpoint._backports.filesystem) FileSystemBase (class in nemo_automodel.components.checkpoint._backports.filesystem) FileSystemReader (class in nemo_automodel.components.checkpoint._backports.filesystem) FileSystemWriter (class in nemo_automodel.components.checkpoint._backports.filesystem) filter() (nemo_automodel.components.loggers.log_utils.RankFilter method) filter_fqns (nemo_automodel.components.quantization.fp8.FP8Config attribute) finalize() (nemo_automodel.components.datasets.llm.megatron.indexed_dataset.IndexedDatasetBuilder method) find_last_non_pad_token() (in module nemo_automodel.components.datasets.utils) finish() (nemo_automodel.components.checkpoint._backports.filesystem._FileSystemWriter method) (nemo_automodel.components.checkpoint._backports.hf_storage._HuggingFaceStorageWriter method) finish_plan() (nemo_automodel.components.checkpoint._backports.default_planner.DefaultLoadPlanner method) (nemo_automodel.components.checkpoint._backports.default_planner.DefaultSavePlanner method) FirstRankPerNode (class in nemo_automodel.components.distributed.utils) flex_attn (nemo_automodel.components.attention.flex_attention.FlexAttention attribute) FLEX_ATTN_MASK_T (in module nemo_automodel.components.attention.flex_attention) FlexAttention (class in nemo_automodel.components.attention.flex_attention) float32 (nemo_automodel.components.datasets.llm.megatron.indexed_dataset.DType attribute) float64 (nemo_automodel.components.datasets.llm.megatron.indexed_dataset.DType attribute) flush_block() (in module nemo_automodel.components.datasets.llm.mock_packed) flux_flops() (in module nemo_automodel.components.utils.flops_utils) force_e_score_correction_bias (nemo_automodel.components.moe.config.MoEConfig attribute) force_recompute_fp8_weight_in_bwd (nemo_automodel.components.quantization.fp8.FP8Config attribute) format() (nemo_automodel.components.loggers.log_utils.ColorFormatter method) format_chat_template() (in module nemo_automodel.components.datasets.llm.formatting_utils) format_prompt_completion() (in module nemo_automodel.components.datasets.llm.formatting_utils) forward() (nemo_automodel._transformers.biencoder.BiencoderModel method) (nemo_automodel.components._peft.lora.LinearLoRA method) (nemo_automodel.components._peft.lora.LoRATritonFunction static method) (nemo_automodel.components._peft.lora.TritonLinearLoRA method) (nemo_automodel.components._peft.lora_moe.GroupedExpertsDeepEPLoRA method) (nemo_automodel.components._peft.lora_moe.GroupedExpertsLoRA method) (nemo_automodel.components.attention.flex_attention.FlexAttention method) (nemo_automodel.components.loss.chunked_ce.ChunkedCrossEntropy method) (nemo_automodel.components.loss.kd_loss.KDLoss method) (nemo_automodel.components.loss.linear_ce.FusedLinearCrossEntropy method) (nemo_automodel.components.loss.masked_ce.MaskedCrossEntropy method) (nemo_automodel.components.loss.te_parallel_ce.CrossEntropyFunction static method) (nemo_automodel.components.models.common.combined_projection.combined_mlp.CombinedGateUpMLP method) (nemo_automodel.components.models.deepseek_v3.layers.MLA method) (nemo_automodel.components.models.deepseek_v3.model.Block method) (nemo_automodel.components.models.deepseek_v3.model.DeepseekV3ForCausalLM method) (nemo_automodel.components.models.deepseek_v3.model.DeepseekV3Model method) (nemo_automodel.components.models.deepseek_v32.layers.DeepseekV32Indexer method) (nemo_automodel.components.models.deepseek_v32.layers.DeepseekV32MLA method) (nemo_automodel.components.models.glm4_moe.layers.Glm4MoeAttention method) (nemo_automodel.components.models.glm4_moe.model.Block method) (nemo_automodel.components.models.glm4_moe.model.Glm4MoeForCausalLM method) (nemo_automodel.components.models.glm4_moe.model.Glm4MoeModel method) (nemo_automodel.components.models.glm4_moe_lite.model.Block method) (nemo_automodel.components.models.glm4_moe_lite.model.Glm4MoeLiteForCausalLM method) (nemo_automodel.components.models.glm4_moe_lite.model.Glm4MoeLiteModel method) (nemo_automodel.components.models.gpt2.CausalSelfAttention method) (nemo_automodel.components.models.gpt2.GPT2LMHeadModel method) (nemo_automodel.components.models.gpt2.MLP method) (nemo_automodel.components.models.gpt2.TransformerBlock method) (nemo_automodel.components.models.gpt_oss.layers.GptOssAttention method) (nemo_automodel.components.models.gpt_oss.model.Block method) (nemo_automodel.components.models.gpt_oss.model.GptOssForCausalLM method) (nemo_automodel.components.models.gpt_oss.model.GptOssModel method) (nemo_automodel.components.models.gpt_oss.rope_utils.RotaryEmbedding method) (nemo_automodel.components.models.kimi_k25_vl.model.KimiK25VLForConditionalGeneration method) (nemo_automodel.components.models.kimi_k25_vl.model.KimiK25VLLanguageModelBackend method) (nemo_automodel.components.models.kimi_k25_vl.model.KimiK25VLModel method) (nemo_automodel.components.models.kimi_k25_vl.model.KimiK25VLMultiModalProjector method) (nemo_automodel.components.models.kimi_k25_vl.model.Learnable2DInterpPosEmbDividedFixed method) (nemo_automodel.components.models.kimi_k25_vl.model.MoonVision3dPatchEmbed method) (nemo_automodel.components.models.kimi_k25_vl.model.MoonViT3dEncoder method) (nemo_automodel.components.models.kimi_k25_vl.model.MoonViT3dEncoderLayer method) (nemo_automodel.components.models.kimi_k25_vl.model.MoonViT3dMLP method) (nemo_automodel.components.models.kimi_k25_vl.model.MoonViT3dPretrainedModel method) (nemo_automodel.components.models.kimivl.model.KimiVLForConditionalGeneration method) (nemo_automodel.components.models.kimivl.model.KimiVLLanguageModelBackend method) (nemo_automodel.components.models.kimivl.model.KimiVLModel method) (nemo_automodel.components.models.kimivl.model.KimiVLMultiModalProjector method) (nemo_automodel.components.models.kimivl.model.Learnable2DInterpPosEmb method) (nemo_automodel.components.models.kimivl.model.MoonVisionPatchEmbed method) (nemo_automodel.components.models.kimivl.model.MoonVitEncoder method) (nemo_automodel.components.models.kimivl.model.MoonVitEncoderLayer method) (nemo_automodel.components.models.kimivl.model.MoonVitMLP method) (nemo_automodel.components.models.kimivl.model.MoonVitPretrainedModel method) (nemo_automodel.components.models.llama.model.LlamaAttention method) (nemo_automodel.components.models.llama.model.LlamaDecoderLayer method) (nemo_automodel.components.models.llama.model.LlamaForCausalLM method) (nemo_automodel.components.models.llama.model.LlamaMLP method) (nemo_automodel.components.models.llama.model.LlamaModel method) (nemo_automodel.components.models.llama.rope_utils.LlamaRotaryEmbedding method) (nemo_automodel.components.models.llama_bidirectional.export_onnx._Pooling method) (nemo_automodel.components.models.llama_bidirectional.export_onnx.EmbeddingModelForExport method) (nemo_automodel.components.models.llama_bidirectional.model.LlamaBidirectionalForSequenceClassification method) (nemo_automodel.components.models.llama_bidirectional.model.LlamaBidirectionalModel method) (nemo_automodel.components.models.minimax_m2.layers.MiniMaxM2Attention method) (nemo_automodel.components.models.minimax_m2.model.Block method) (nemo_automodel.components.models.minimax_m2.model.MiniMaxM2ForCausalLM method) (nemo_automodel.components.models.minimax_m2.model.MiniMaxM2Model method) (nemo_automodel.components.models.mistral3.model.GradientCheckpointingLayer method) (nemo_automodel.components.models.mistral3.model.Ministral3Attention method) (nemo_automodel.components.models.mistral3.model.Ministral3DecoderLayer method) (nemo_automodel.components.models.mistral3.model.Ministral3ForCausalLM method) (nemo_automodel.components.models.mistral3.model.Ministral3MLP method) (nemo_automodel.components.models.mistral3.model.Ministral3Model method) (nemo_automodel.components.models.mistral3.model.Ministral3RMSNorm method) (nemo_automodel.components.models.mistral3.model.Ministral3RotaryEmbedding method) (nemo_automodel.components.models.nemotron_parse.model.NemotronParseDecoder method) (nemo_automodel.components.models.nemotron_parse.model.NemotronParseForConditionalGeneration method) (nemo_automodel.components.models.nemotron_parse.model.RadioWithNeck method) (nemo_automodel.components.models.nemotron_parse.nemotron_parse_loss.NemotronParseLoss method) (nemo_automodel.components.models.nemotron_v3.layers.NemotronV3Attention method) (nemo_automodel.components.models.nemotron_v3.layers.NemotronV3Block method) (nemo_automodel.components.models.nemotron_v3.layers.NemotronV3Mamba2Mixer method) (nemo_automodel.components.models.nemotron_v3.layers.NemotronV3MambaRMSNormGated method) (nemo_automodel.components.models.nemotron_v3.model.NemotronHForCausalLM method) (nemo_automodel.components.models.nemotron_v3.model.NemotronV3Model method) (nemo_automodel.components.models.qwen2.model.Qwen2Attention method) (nemo_automodel.components.models.qwen2.model.Qwen2DecoderLayer method) (nemo_automodel.components.models.qwen2.model.Qwen2ForCausalLM method) (nemo_automodel.components.models.qwen2.model.Qwen2Model method) (nemo_automodel.components.models.qwen3_5_moe.model.Qwen3_5MoeForConditionalGeneration method) (nemo_automodel.components.models.qwen3_5_moe.model.Qwen3_5MoeModel method) (nemo_automodel.components.models.qwen3_5_moe.model.Qwen3_5MoeTextModelBackend method) (nemo_automodel.components.models.qwen3_moe.layers.Qwen3MoeAttention method) (nemo_automodel.components.models.qwen3_moe.model.Block method) (nemo_automodel.components.models.qwen3_moe.model.Qwen3MoeForCausalLM method) (nemo_automodel.components.models.qwen3_moe.model.Qwen3MoeModel method) (nemo_automodel.components.models.qwen3_next.layers.Qwen3NextAttention method) (nemo_automodel.components.models.qwen3_next.layers.Qwen3NextRMSNorm method) (nemo_automodel.components.models.qwen3_next.model.Block method) (nemo_automodel.components.models.qwen3_next.model.Qwen3NextForCausalLM method) (nemo_automodel.components.models.qwen3_next.model.Qwen3NextModel method) (nemo_automodel.components.models.qwen3_omni_moe.model.Qwen3OmniMoeThinkerForConditionalGeneration method) (nemo_automodel.components.models.qwen3_omni_moe.model.Qwen3OmniMoeThinkerTextModel method) (nemo_automodel.components.models.qwen3_vl_moe.model.Qwen3VLMoeBlock method) (nemo_automodel.components.models.qwen3_vl_moe.model.Qwen3VLMoeForConditionalGeneration method) (nemo_automodel.components.models.qwen3_vl_moe.model.Qwen3VLMoeModel method) (nemo_automodel.components.models.qwen3_vl_moe.model.Qwen3VLMoeTextModelBackend method) (nemo_automodel.components.models.step3p5.layers.Step3p5Attention method) (nemo_automodel.components.models.step3p5.layers.Step3p5MLP method) (nemo_automodel.components.models.step3p5.layers.Step3p5RMSNorm method) (nemo_automodel.components.models.step3p5.layers.Step3p5RotaryEmbedding method) (nemo_automodel.components.models.step3p5.model.Block method) (nemo_automodel.components.models.step3p5.model.Step3p5ForCausalLM method) (nemo_automodel.components.models.step3p5.model.Step3p5Model method) (nemo_automodel.components.moe.experts._AllGatherConcatVarlenFn static method) (nemo_automodel.components.moe.experts.GroupedExperts method) (nemo_automodel.components.moe.experts.GroupedExpertsDeepEP method) (nemo_automodel.components.moe.experts.GroupedExpertsTE method) (nemo_automodel.components.moe.layers.FakeBalancedGate method) (nemo_automodel.components.moe.layers.Gate method) (nemo_automodel.components.moe.layers.MLP method) (nemo_automodel.components.moe.layers.MoE method) (nemo_automodel.components.moe.megatron.fused_a2a.FusedCombine static method) (nemo_automodel.components.moe.megatron.fused_a2a.FusedDispatch static method) (nemo_automodel.components.moe.megatron.fused_indices_converter.IndicesToMultihot static method) (nemo_automodel.components.moe.megatron.moe_utils.MoEAuxLossAutoScaler static method) (nemo_automodel.components.moe.megatron.moe_utils.WeightedBiasQuickGeGLUFunction static method) (nemo_automodel.components.moe.megatron.moe_utils.WeightedQuickGeGLUFunction static method) (nemo_automodel.components.moe.megatron.moe_utils.WeightedSwiGLUFunction static method)

forward_autotune_configs() (in module nemo_automodel.components._peft.lora_kernel) Fp32SafeQwen3_5MoeTextRotaryEmbedding (class in nemo_automodel.components.models.qwen3_5_moe.model) Fp32SafeQwen3_5MoeVisionRotaryEmbedding (class in nemo_automodel.components.models.qwen3_5_moe.model) Fp32SafeQwen3VLMoeTextRotaryEmbedding (class in nemo_automodel.components.models.qwen3_vl_moe.model) Fp32SafeQwen3VLMoeVisionRotaryEmbedding (class in nemo_automodel.components.models.qwen3_vl_moe.model) FP4_VALUES (in module nemo_automodel.components.models.gpt_oss.state_dict_adapter) FP8Config (class in nemo_automodel.components.quantization.fp8) fqn_data (nemo_automodel.components.checkpoint._backports.consolidate_hf_safetensors._OutputFileData attribute) freqs_cis (nemo_automodel.components.models.kimi_k25_vl.model.DeepSeekV3RotaryEmbeddingAdapter property) (nemo_automodel.components.models.kimivl.model.DeepSeekV3RotaryEmbeddingAdapter property) freqs_cis_from_position_ids() (in module nemo_automodel.components.models.deepseek_v3.rope_utils) from_config() (nemo_automodel._transformers.auto_model._BaseNeMoAutoModelClass class method) (nemo_automodel.components.models.deepseek_v3.model.DeepseekV3ForCausalLM class method) (nemo_automodel.components.models.deepseek_v32.model.DeepseekV32ForCausalLM class method) (nemo_automodel.components.models.glm4_moe.model.Glm4MoeForCausalLM class method) (nemo_automodel.components.models.glm4_moe_lite.model.Glm4MoeLiteForCausalLM class method) (nemo_automodel.components.models.gpt_oss.model.GptOssForCausalLM class method) (nemo_automodel.components.models.kimi_k25_vl.model.KimiK25VLForConditionalGeneration class method) (nemo_automodel.components.models.kimivl.model.KimiVLForConditionalGeneration class method) (nemo_automodel.components.models.llama.model.LlamaForCausalLM class method) (nemo_automodel.components.models.minimax_m2.model.MiniMaxM2ForCausalLM class method) (nemo_automodel.components.models.nemotron_v3.model.NemotronHForCausalLM class method) (nemo_automodel.components.models.qwen3_5_moe.model.Qwen3_5MoeForConditionalGeneration class method) (nemo_automodel.components.models.qwen3_moe.model.Qwen3MoeForCausalLM class method) (nemo_automodel.components.models.qwen3_next.model.Qwen3NextForCausalLM class method) (nemo_automodel.components.models.qwen3_omni_moe.model.Qwen3OmniMoeThinkerForConditionalGeneration class method) (nemo_automodel.components.models.qwen3_vl_moe.model.Qwen3VLMoeForConditionalGeneration class method) (nemo_automodel.components.models.step3p5.model.Step3p5ForCausalLM class method) from_dict() (nemo_automodel.components._peft.lora.PeftConfig class method) from_hf() (nemo_automodel.components.checkpoint.state_dict_adapter.StateDictAdapter method) (nemo_automodel.components.models.common.bidirectional.BiencoderStateDictAdapter method) (nemo_automodel.components.models.common.combined_projection.state_dict_adapter.CombinedProjectionStateDictAdapter method) (nemo_automodel.components.models.deepseek_v3.state_dict_adapter.DeepSeekV3StateDictAdapter method) (nemo_automodel.components.models.glm4_moe.state_dict_adapter.Glm4MoeStateDictAdapter method) (nemo_automodel.components.models.gpt_oss.state_dict_adapter.GPTOSSStateDictAdapter method) (nemo_automodel.components.models.kimi_k25_vl.state_dict_adapter.KimiK25VLStateDictAdapter method) (nemo_automodel.components.models.kimivl.model.KimiVLStateDictAdapter method) (nemo_automodel.components.models.minimax_m2.state_dict_adapter.MiniMaxM2StateDictAdapter method) (nemo_automodel.components.models.nemotron_v3.state_dict_adapter.NemotronV3StateDictAdapter method) (nemo_automodel.components.models.qwen3_5_moe.state_dict_adapter.Qwen3_5MoeStateDictAdapter method) (nemo_automodel.components.models.qwen3_moe.state_dict_adapter.Qwen3MoeStateDictAdapter method) (nemo_automodel.components.models.qwen3_next.state_dict_adapter.Qwen3NextStateDictAdapter method) (nemo_automodel.components.models.qwen3_omni_moe.state_dict_adapter.Qwen3OmniMoeStateDictAdapter method) (nemo_automodel.components.models.qwen3_vl_moe.state_dict_adapter.Qwen3VLMoeStateDictAdapter method) (nemo_automodel.components.models.step3p5.state_dict_adapter.Step3p5StateDictAdapter method) from_meshes() (nemo_automodel.components.distributed.mesh.MeshContext class method) from_pretrained() (nemo_automodel._diffusers.auto_diffusion_pipeline.NeMoAutoDiffusionPipeline class method) (nemo_automodel._transformers.auto_model._BaseNeMoAutoModelClass class method) (nemo_automodel._transformers.auto_model.NeMoAutoModelBiencoder class method) (nemo_automodel._transformers.auto_tokenizer.NeMoAutoTokenizer class method) (nemo_automodel.components.models.deepseek_v3.model.DeepseekV3ForCausalLM class method) (nemo_automodel.components.models.deepseek_v32.model.DeepseekV32ForCausalLM class method) (nemo_automodel.components.models.glm4_moe.model.Glm4MoeForCausalLM class method) (nemo_automodel.components.models.glm4_moe_lite.model.Glm4MoeLiteForCausalLM class method) (nemo_automodel.components.models.gpt_oss.model.GptOssForCausalLM class method) (nemo_automodel.components.models.kimi_k25_vl.model.KimiK25VLForConditionalGeneration class method) (nemo_automodel.components.models.kimivl.model.KimiVLForConditionalGeneration class method) (nemo_automodel.components.models.minimax_m2.model.MiniMaxM2ForCausalLM class method) (nemo_automodel.components.models.nemotron_v3.model.NemotronHForCausalLM class method) (nemo_automodel.components.models.qwen3_5_moe.model.Qwen3_5MoeForConditionalGeneration class method) (nemo_automodel.components.models.qwen3_moe.model.Qwen3MoeForCausalLM class method) (nemo_automodel.components.models.qwen3_next.model.Qwen3NextForCausalLM class method) (nemo_automodel.components.models.qwen3_omni_moe.model.Qwen3OmniMoeThinkerForConditionalGeneration class method) (nemo_automodel.components.models.qwen3_vl_moe.model.Qwen3VLMoeForConditionalGeneration class method) (nemo_automodel.components.models.step3p5.model.Step3p5ForCausalLM class method) fsdp2_strategy_parallelize() (in module nemo_automodel.components.distributed.parallelizer) FSDP2Config (class in nemo_automodel.components.distributed.config) FSDP2Manager (class in nemo_automodel.components.distributed.fsdp2) fsdp_double_buffer (nemo_automodel.components.distributed.config.MegatronFSDPConfig attribute) FsspecReader (class in nemo_automodel.components.checkpoint._backports._fsspec_filesystem) FsspecWriter (class in nemo_automodel.components.checkpoint._backports._fsspec_filesystem) fullgraph (nemo_automodel.components.utils.compile_utils.CompileConfig attribute) fully_shard_by_dtype() (in module nemo_automodel.components.distributed.parallelizer_utils) fully_shard_optimizer() (in module nemo_automodel.components.distributed.megatron_fsdp) fused_indices_to_multihot() (in module nemo_automodel.components.moe.megatron.fused_indices_converter) FusedCombine (class in nemo_automodel.components.moe.megatron.fused_a2a) FusedDispatch (class in nemo_automodel.components.moe.megatron.fused_a2a) FusedLinearCrossEntropy (class in nemo_automodel.components.loss.linear_ce) future (nemo_automodel.components.checkpoint.checkpointing._AsyncSaveContext attribute)

G

GarbageCollection (class in nemo_automodel.components.training.garbage_collection) Gate (class in nemo_automodel.components.moe.layers) gate_and_up_projs (nemo_automodel.components.moe.experts.GroupedExpertsTE property) gate_bias_update_factor (nemo_automodel.components.moe.config.MoEConfig attribute) gate_precision (nemo_automodel.components.models.common.utils.BackendConfig attribute) gate_up_proj_bias (nemo_automodel.components.moe.experts.GroupedExpertsTE property) gen_sentence_ids() (in module nemo_automodel.components.datasets.llm.mock) (in module nemo_automodel.components.datasets.llm.mock_packed) generate_hf_model_fqn_per_model_part() (in module nemo_automodel.components.distributed.pipelining.functional) GENERATION_REGEX (in module nemo_automodel.components.datasets.llm.formatting_utils) get() (nemo_automodel.components.config.loader.ConfigNode method) (nemo_automodel.components.datasets.llm.megatron.indexed_dataset.IndexedDataset method) get_1d_sincos_pos_embed() (in module nemo_automodel.components.models.kimi_k25_vl.model) get_1d_sincos_pos_embed_from_grid() (in module nemo_automodel.components.models.kimi_k25_vl.model) get_all_ids() (nemo_automodel.components.datasets.llm.retrieval_dataset.AbstractDataset method) (nemo_automodel.components.datasets.llm.retrieval_dataset.CorpusInfo method) (nemo_automodel.components.datasets.llm.retrieval_dataset.TextQADataset method) get_all_timers_string() (nemo_automodel.components.training.timers.Timers method) get_architectures() (in module nemo_automodel._transformers.model_init) get_as_string() (nemo_automodel.components.config.loader.ConfigNode method) get_bin_path() (in module nemo_automodel.components.datasets.llm.megatron.indexed_dataset) get_blend_from_list() (in module nemo_automodel.components.datasets.llm.megatron.megatron_utils) get_buffer() (in module nemo_automodel.components.moe.megatron.fused_a2a) get_check_model_inputs_decorator() (in module nemo_automodel.shared.import_utils) get_checkpoint_conversion_mapping() (in module nemo_automodel.components.checkpoint.conversion_mapping) get_combined_key_mapping() (in module nemo_automodel.components.checkpoint.conversion_mapping) get_context() (nemo_automodel.components.datasets.llm.hellaswag.HellaSwag method) get_cpu_state_dict() (in module nemo_automodel.components.distributed.tensor_utils) get_dataset() (nemo_automodel.components.datasets.llm.megatron_dataset.MegatronPretraining method) get_decoder() (nemo_automodel.components.models.llama.model.LlamaForCausalLM method) (nemo_automodel.components.models.nemotron_parse.model.NemotronParseForConditionalGeneration method) get_device() (in module nemo_automodel.components.training.signal_handler) get_disable_fake_quant_fn() (in module nemo_automodel.components.quantization.qat) get_dispached_metadata() (nemo_automodel.components.moe.megatron.token_dispatcher._DeepepManager method) (nemo_automodel.components.moe.megatron.token_dispatcher._DispatchManager method) get_document_by_id() (nemo_automodel.components.datasets.llm.retrieval_dataset.AbstractDataset method) (nemo_automodel.components.datasets.llm.retrieval_dataset.CorpusInfo method) (nemo_automodel.components.datasets.llm.retrieval_dataset.TextQADataset method) get_enable_fake_quant_fn() (in module nemo_automodel.components.quantization.qat) get_encoder() (nemo_automodel.components.models.nemotron_parse.model.NemotronParseForConditionalGeneration method) get_expert_activation_for_deepep() (in module nemo_automodel.components.moe.experts) get_expert_activation_with_lora() (in module nemo_automodel.components._peft.lora_moe) get_expert_range_for_rank_from_mesh() (in module nemo_automodel.components.moe.state_dict_utils) get_expert_slice_for_rank() (in module nemo_automodel.components.moe.state_dict_utils) get_final_hidden_states() (in module nemo_automodel.components.training.model_output_utils) get_flops_formula_for_hf_config() (in module nemo_automodel.components.utils.flops_utils) get_fqn_to_file_index_mapping() (in module nemo_automodel.components.checkpoint._backports.hf_storage) get_freqs_cis() (nemo_automodel.components.models.kimi_k25_vl.model.Rope2DPosEmbRepeated method) (nemo_automodel.components.models.kimivl.model.Rope2DPosEmb method) get_grad_norm() (in module nemo_automodel.components.distributed.grad_utils) get_hf_config() (in module nemo_automodel._transformers.model_init) get_hf_tp_shard_plan() (in module nemo_automodel.components.distributed.parallelizer) get_hidden_bytes() (in module nemo_automodel.components.moe.megatron.fused_a2a) get_idx_path() (in module nemo_automodel.components.datasets.llm.megatron.indexed_dataset) get_input_embeddings() (nemo_automodel.components.models.deepseek_v3.model.DeepseekV3ForCausalLM method) (nemo_automodel.components.models.deepseek_v32.model.DeepseekV32ForCausalLM method) (nemo_automodel.components.models.glm4_moe.model.Glm4MoeForCausalLM method) (nemo_automodel.components.models.glm4_moe_lite.model.Glm4MoeLiteForCausalLM method) (nemo_automodel.components.models.gpt_oss.model.GptOssForCausalLM method) (nemo_automodel.components.models.kimi_k25_vl.model.KimiK25VLForConditionalGeneration method) (nemo_automodel.components.models.kimi_k25_vl.model.KimiK25VLLanguageModelBackend method) (nemo_automodel.components.models.kimivl.model.KimiVLForConditionalGeneration method) (nemo_automodel.components.models.kimivl.model.KimiVLLanguageModelBackend method) (nemo_automodel.components.models.llama.model.LlamaForCausalLM method) (nemo_automodel.components.models.minimax_m2.model.MiniMaxM2ForCausalLM method) (nemo_automodel.components.models.mistral3.model.Ministral3ForCausalLM method) (nemo_automodel.components.models.nemotron_parse.model.NemotronParseDecoder method) (nemo_automodel.components.models.nemotron_parse.model.NemotronParseForConditionalGeneration method) (nemo_automodel.components.models.nemotron_v3.model.NemotronHForCausalLM method) (nemo_automodel.components.models.qwen2.model.Qwen2ForCausalLM method) (nemo_automodel.components.models.qwen3_5_moe.model.Qwen3_5MoeTextModelBackend method) (nemo_automodel.components.models.qwen3_moe.model.Qwen3MoeForCausalLM method) (nemo_automodel.components.models.qwen3_next.model.Qwen3NextForCausalLM method) (nemo_automodel.components.models.qwen3_omni_moe.model.Qwen3OmniMoeThinkerForConditionalGeneration method) (nemo_automodel.components.models.qwen3_vl_moe.model.Qwen3VLMoeForConditionalGeneration method) (nemo_automodel.components.models.qwen3_vl_moe.model.Qwen3VLMoeTextModelBackend method) (nemo_automodel.components.models.step3p5.model.Step3p5ForCausalLM method) get_is_first_microbatch() (in module nemo_automodel.components.models.common.utils) get_is_hf_model() (in module nemo_automodel._transformers.model_init) get_is_optim_step() (in module nemo_automodel.components.models.common.utils)

get_list_of_files() (in module nemo_automodel.components.datasets.llm.megatron_dataset) get_local_rank_preinit() (in module nemo_automodel.components.distributed.init_utils) get_local_world_size_preinit() (in module nemo_automodel.components.distributed.init_utils) get_lr() (nemo_automodel.components.optim.scheduler.OptimizerParamScheduler method) get_model_cls_from_model_arch() (nemo_automodel._transformers.registry._ModelRegistry method) get_model_conversion_mapping() (in module nemo_automodel.components.checkpoint.conversion_mapping) get_number_of_tokens_per_expert() (nemo_automodel.components.moe.megatron.token_dispatcher._DeepepManager method) get_output_embeddings() (nemo_automodel.components.models.deepseek_v3.model.DeepseekV3ForCausalLM method) (nemo_automodel.components.models.deepseek_v32.model.DeepseekV32ForCausalLM method) (nemo_automodel.components.models.glm4_moe.model.Glm4MoeForCausalLM method) (nemo_automodel.components.models.glm4_moe_lite.model.Glm4MoeLiteForCausalLM method) (nemo_automodel.components.models.gpt_oss.model.GptOssForCausalLM method) (nemo_automodel.components.models.kimi_k25_vl.model.KimiK25VLForConditionalGeneration method) (nemo_automodel.components.models.kimivl.model.KimiVLForConditionalGeneration method) (nemo_automodel.components.models.llama.model.LlamaForCausalLM method) (nemo_automodel.components.models.minimax_m2.model.MiniMaxM2ForCausalLM method) (nemo_automodel.components.models.mistral3.model.Ministral3ForCausalLM method) (nemo_automodel.components.models.nemotron_parse.model.NemotronParseForConditionalGeneration method) (nemo_automodel.components.models.nemotron_v3.model.NemotronHForCausalLM method) (nemo_automodel.components.models.qwen2.model.Qwen2ForCausalLM method) (nemo_automodel.components.models.qwen3_moe.model.Qwen3MoeForCausalLM method) (nemo_automodel.components.models.qwen3_next.model.Qwen3NextForCausalLM method) (nemo_automodel.components.models.qwen3_omni_moe.model.Qwen3OmniMoeThinkerForConditionalGeneration method) (nemo_automodel.components.models.qwen3_vl_moe.model.Qwen3VLMoeForConditionalGeneration method) (nemo_automodel.components.models.step3p5.model.Step3p5ForCausalLM method) get_pad_token_from_key() (in module nemo_automodel.components.datasets.utils) get_parallelization_strategy() (in module nemo_automodel.components.distributed.parallelizer) get_permuted_hidden_states_by_experts() (nemo_automodel.components.moe.megatron.token_dispatcher._DeepepManager method) (nemo_automodel.components.moe.megatron.token_dispatcher._DispatchManager method) get_pid_coords() (in module nemo_automodel.components._peft.lora_kernel) get_quantizer_mode() (in module nemo_automodel.components.quantization.qat) get_rank_safe() (in module nemo_automodel.components.distributed.init_utils) get_registry() (in module nemo_automodel._transformers.registry) get_restored_hidden_states_by_experts() (nemo_automodel.components.moe.megatron.token_dispatcher._DeepepManager method) (nemo_automodel.components.moe.megatron.token_dispatcher._DispatchManager method) get_rope_config() (in module nemo_automodel.components.models.common.utils) get_safetensors_index_path() (in module nemo_automodel.components.checkpoint.checkpointing) get_seq_length() (nemo_automodel.components.models.nemotron_v3.cache.NemotronHybridCache method) get_stage_param_counts() (nemo_automodel.components.distributed.pipelining.autopipeline.AutoPipeline method) get_start_end_idx() (nemo_automodel.components.datasets.llm.megatron.sampler.MegatronPretrainingSampler method) get_submesh() (in module nemo_automodel.components.moe.state_dict_utils) get_sync_ctx() (in module nemo_automodel.components.distributed.utils) get_target() (nemo_automodel.components.datasets.llm.hellaswag.HellaSwag method) get_te_version() (in module nemo_automodel.shared.import_utils) get_text_module() (in module nemo_automodel.components.distributed.pipelining.hf_utils) get_torch_version() (in module nemo_automodel.shared.import_utils) get_total_param_count() (nemo_automodel.components.distributed.pipelining.autopipeline.AutoPipeline method) get_train_context() (in module nemo_automodel.components.distributed.cp_utils) get_transformers_version() (in module nemo_automodel.shared.import_utils) get_wd() (nemo_automodel.components.optim.scheduler.OptimizerParamScheduler method) get_world_size_safe() (in module nemo_automodel.components.distributed.init_utils) glm4_moe_flops() (in module nemo_automodel.components.utils.flops_utils) Glm4MoeAttention (class in nemo_automodel.components.models.glm4_moe.layers) Glm4MoeForCausalLM (class in nemo_automodel.components.models.glm4_moe.model) Glm4MoeLiteForCausalLM (class in nemo_automodel.components.models.glm4_moe_lite.model) Glm4MoeLiteModel (class in nemo_automodel.components.models.glm4_moe_lite.model) Glm4MoeModel (class in nemo_automodel.components.models.glm4_moe.model) Glm4MoeStateDictAdapter (class in nemo_automodel.components.models.glm4_moe.state_dict_adapter) GLUE_MRPC (class in nemo_automodel.components.datasets.llm.seq_cls) GPT2LMHeadModel (class in nemo_automodel.components.models.gpt2) gpt3_flops() (in module nemo_automodel.components.utils.flops_utils) gpt_dataset_config (nemo_automodel.components.datasets.llm.megatron_dataset.MegatronPretraining property) gpt_oss_flops() (in module nemo_automodel.components.utils.flops_utils) gpt_oss_flops_calculator() (in module nemo_automodel.components.utils.flops_utils) GPTDataset (class in nemo_automodel.components.datasets.llm.megatron.gpt_dataset) GPTDatasetConfig (class in nemo_automodel.components.datasets.llm.megatron.gpt_dataset) GptOssAttention (class in nemo_automodel.components.models.gpt_oss.layers) GptOssForCausalLM (class in nemo_automodel.components.models.gpt_oss.model) GptOssModel (class in nemo_automodel.components.models.gpt_oss.model) GPTOSSStateDictAdapter (class in nemo_automodel.components.models.gpt_oss.state_dict_adapter) GPU_INSTALL_STRING (in module nemo_automodel.shared.import_utils) gpu_only_import() (in module nemo_automodel.shared.import_utils) gpu_only_import_from() (in module nemo_automodel.shared.import_utils) grad_reduce_in_fp32 (nemo_automodel.components.distributed.config.MegatronFSDPConfig attribute) GradientCheckpointingLayer (class in nemo_automodel.components.models.mistral3.model) GroupedExperts (class in nemo_automodel.components.moe.experts) GroupedExpertsDeepEP (class in nemo_automodel.components.moe.experts) GroupedExpertsDeepEPLoRA (class in nemo_automodel.components._peft.lora_moe) GroupedExpertsLoRA (class in nemo_automodel.components._peft.lora_moe) GroupedExpertsTE (class in nemo_automodel.components.moe.experts)

H

has_first_stage (nemo_automodel.components.distributed.pipelining.autopipeline.PipelineInfo attribute) has_last_stage (nemo_automodel.components.distributed.pipelining.autopipeline.PipelineInfo attribute) HAS_LIGER_KERNEL (in module nemo_automodel._transformers.kernel_patches) HAVE_DEEP_EP (in module nemo_automodel.components.models.common.utils) HAVE_GMM (in module nemo_automodel.components.models.common.utils) HAVE_MEGATRON_FSDP (in module nemo_automodel.components.distributed.parallelizer)

HAVE_TE (in module nemo_automodel.components.models.common.utils) HAVE_TE_PARALLEL_CE (in module nemo_automodel.components.loss.te_parallel_ce) HEADER_BYTES (in module nemo_automodel.components.datasets.llm.nanogpt_dataset) HEADER_SIZE (in module nemo_automodel.components.datasets.llm.nanogpt_dataset) HellaSwag (class in nemo_automodel.components.datasets.llm.hellaswag) HF_DCP_VERSION (in module nemo_automodel.components.checkpoint._backports.hf_utils) HFCheckpointingMixin (class in nemo_automodel.components.models.common.hf_checkpointing_mixin)

I

ignore_router_for_ac (nemo_automodel.components.moe.config.MoEParallelizerConfig attribute) image_hidden_states (nemo_automodel.components.models.mistral3.model.Ministral3ModelOutputWithPast attribute) import_class_from_path() (in module nemo_automodel.components.distributed.parallelizer) import_classes_from_paths() (in module nemo_automodel.components.distributed.parallelizer) IndexedDataset (class in nemo_automodel.components.datasets.llm.megatron.indexed_dataset) IndexedDatasetBuilder (class in nemo_automodel.components.datasets.llm.megatron.indexed_dataset) IndicesToMultihot (class in nemo_automodel.components.moe.megatron.fused_indices_converter) info (nemo_automodel.components.distributed.pipelining.autopipeline.AutoPipeline property) init_all_rng() (in module nemo_automodel.components.training.rng) init_empty_weights() (in module nemo_automodel.components.utils.model_utils) init_fsdp_with_meta_device (nemo_automodel.components.distributed.config.MegatronFSDPConfig attribute) init_hf_model_buffers() (in module nemo_automodel.components.distributed.pipelining.hf_utils) init_lora_weights() (nemo_automodel.components._peft.lora.LinearLoRA method) (nemo_automodel.components._peft.lora_moe.GroupedExpertsDeepEPLoRA method) (nemo_automodel.components._peft.lora_moe.GroupedExpertsLoRA method) init_path() (nemo_automodel.components.checkpoint._backports._fsspec_filesystem.FileSystem method) (nemo_automodel.components.checkpoint._backports.filesystem.FileSystem method) (nemo_automodel.components.checkpoint._backports.filesystem.FileSystemBase method) init_token_dispatcher() (nemo_automodel.components.moe.experts.GroupedExpertsDeepEP method) (nemo_automodel.components.moe.experts.GroupedExpertsTE method) init_weights() (nemo_automodel.components.models.deepseek_v3.layers.MLA method) (nemo_automodel.components.models.deepseek_v3.model.Block method) (nemo_automodel.components.models.deepseek_v3.model.DeepseekV3Model method) (nemo_automodel.components.models.deepseek_v32.layers.DeepseekV32Indexer method) (nemo_automodel.components.models.deepseek_v32.layers.DeepseekV32MLA method) (nemo_automodel.components.models.glm4_moe.layers.Glm4MoeAttention method) (nemo_automodel.components.models.glm4_moe.model.Block method) (nemo_automodel.components.models.glm4_moe.model.Glm4MoeModel method) (nemo_automodel.components.models.glm4_moe_lite.model.Block method) (nemo_automodel.components.models.glm4_moe_lite.model.Glm4MoeLiteModel method) (nemo_automodel.components.models.gpt_oss.layers.GptOssAttention method) (nemo_automodel.components.models.gpt_oss.model.Block method) (nemo_automodel.components.models.gpt_oss.model.GptOssModel method) (nemo_automodel.components.models.kimi_k25_vl.model.KimiK25VLLanguageModelBackend method) (nemo_automodel.components.models.kimivl.model.KimiVLLanguageModelBackend method) (nemo_automodel.components.models.minimax_m2.layers.MiniMaxM2Attention method) (nemo_automodel.components.models.minimax_m2.model.Block method) (nemo_automodel.components.models.minimax_m2.model.MiniMaxM2Model method) (nemo_automodel.components.models.nemotron_v3.layers.NemotronV3Attention method) (nemo_automodel.components.models.nemotron_v3.layers.NemotronV3Block method) (nemo_automodel.components.models.nemotron_v3.layers.NemotronV3Mamba2Mixer method) (nemo_automodel.components.models.qwen3_5_moe.model.Qwen3_5MoeBlock method) (nemo_automodel.components.models.qwen3_5_moe.model.Qwen3_5MoeTextModelBackend method) (nemo_automodel.components.models.qwen3_moe.layers.Qwen3MoeAttention method) (nemo_automodel.components.models.qwen3_moe.model.Block method) (nemo_automodel.components.models.qwen3_moe.model.Qwen3MoeModel method) (nemo_automodel.components.models.qwen3_next.layers.Qwen3NextAttention method) (nemo_automodel.components.models.qwen3_next.model.Block method) (nemo_automodel.components.models.qwen3_next.model.Qwen3NextModel method) (nemo_automodel.components.models.qwen3_omni_moe.model.Qwen3OmniMoeThinkerTextModel method) (nemo_automodel.components.models.qwen3_vl_moe.model.Qwen3VLMoeTextModelBackend method) (nemo_automodel.components.models.step3p5.layers.Step3p5Attention method) (nemo_automodel.components.models.step3p5.layers.Step3p5MLP method) (nemo_automodel.components.models.step3p5.model.Block method) (nemo_automodel.components.models.step3p5.model.Step3p5Model method) (nemo_automodel.components.moe.experts.GroupedExperts method) (nemo_automodel.components.moe.experts.GroupedExpertsDeepEP method) (nemo_automodel.components.moe.experts.GroupedExpertsTE method) (nemo_automodel.components.moe.layers.FakeBalancedGate method) (nemo_automodel.components.moe.layers.Gate method) (nemo_automodel.components.moe.layers.MLP method) (nemo_automodel.components.moe.layers.MoE method)

initialize() (nemo_automodel.components.datasets.llm.megatron.indexed_dataset.IndexedDataset method) initialize_attn_module_and_func() (in module nemo_automodel.components.attention.utils) initialize_distributed() (in module nemo_automodel.components.distributed.init_utils) initialize_linear_module() (in module nemo_automodel.components.models.common.utils) initialize_model_weights() (nemo_automodel.components.checkpoint.checkpointing.Checkpointer static method) initialize_rms_norm_module() (in module nemo_automodel.components.models.common.utils) initialize_weights() (nemo_automodel.components.models.deepseek_v3.model.DeepseekV3ForCausalLM method) (nemo_automodel.components.models.glm4_moe.model.Glm4MoeForCausalLM method) (nemo_automodel.components.models.glm4_moe_lite.model.Glm4MoeLiteForCausalLM method) (nemo_automodel.components.models.gpt2.GPT2LMHeadModel method) (nemo_automodel.components.models.gpt_oss.model.GptOssForCausalLM method) (nemo_automodel.components.models.kimi_k25_vl.model.KimiK25VLForConditionalGeneration method) (nemo_automodel.components.models.kimivl.model.KimiVLForConditionalGeneration method) (nemo_automodel.components.models.minimax_m2.model.MiniMaxM2ForCausalLM method) (nemo_automodel.components.models.nemotron_v3.model.NemotronHForCausalLM method) (nemo_automodel.components.models.nemotron_v3.model.NemotronV3Model method) (nemo_automodel.components.models.qwen3_5_moe.model.Qwen3_5MoeForConditionalGeneration method) (nemo_automodel.components.models.qwen3_moe.model.Qwen3MoeForCausalLM method) (nemo_automodel.components.models.qwen3_next.model.Qwen3NextForCausalLM method) (nemo_automodel.components.models.qwen3_omni_moe.model.Qwen3OmniMoeThinkerForConditionalGeneration method) (nemo_automodel.components.models.qwen3_vl_moe.model.Qwen3VLMoeForConditionalGeneration method) (nemo_automodel.components.models.step3p5.model.Step3p5ForCausalLM method) INLINE_CORPUS_ID (in module nemo_automodel.components.datasets.llm.retrieval_dataset_inline) inner_kernel() (in module nemo_automodel.components._peft.lora_kernel) instantiate() (nemo_automodel.components.config.loader.ConfigNode method) instantiate_infrastructure() (in module nemo_automodel._transformers.infrastructure) instantiate_path() (nemo_automodel.components.config.loader.ConfigNode method) int16 (nemo_automodel.components.datasets.llm.megatron.indexed_dataset.DType attribute) int32 (nemo_automodel.components.datasets.llm.megatron.indexed_dataset.DType attribute) int64 (nemo_automodel.components.datasets.llm.megatron.indexed_dataset.DType attribute) int8 (nemo_automodel.components.datasets.llm.megatron.indexed_dataset.DType attribute) inter_dim (nemo_automodel.components.moe.config.MoEConfig attribute) inv_freq (nemo_automodel.components.models.llama.rope_utils.LlamaRotaryEmbedding attribute) (nemo_automodel.components.models.mistral3.model.Ministral3RotaryEmbedding attribute) is_async (nemo_automodel.components.checkpoint.checkpointing.CheckpointingConfig attribute) is_causal_lm (nemo_automodel.components._peft.module_matcher.ModuleMatcher attribute) is_ckpt_step (nemo_automodel.components.training.step_scheduler.StepScheduler property) is_compileable (nemo_automodel.components.models.nemotron_v3.cache.NemotronHybridCache attribute) is_composition (nemo_automodel.components.models.nemotron_parse.model.NemotronParseConfig attribute) is_delta_lake_path() (in module nemo_automodel.components.datasets.llm.delta_lake_dataset) is_dion_optimizer() (in module nemo_automodel.components.optim.utils) is_dtensor() (in module nemo_automodel.components.moe.state_dict_utils) IS_FIRST_MICROBATCH (in module nemo_automodel.components.models.common.utils) is_gated_activation() (in module nemo_automodel.components.moe.experts) is_gc_step (nemo_automodel.components.training.step_scheduler.StepScheduler property) is_last_batch (nemo_automodel.components.training.step_scheduler.StepScheduler property) is_last_step (nemo_automodel.components.training.step_scheduler.StepScheduler property) is_main (nemo_automodel.components.distributed.init_utils.DistInfo attribute) is_number_tryexcept() (in module nemo_automodel.components.datasets.llm.megatron_dataset) IS_OPTIM_STEP (in module nemo_automodel.components.models.common.utils) is_peft (nemo_automodel.components.checkpoint.checkpointing.CheckpointingConfig attribute) is_remote_logging_step (nemo_automodel.components.training.step_scheduler.StepScheduler property) is_te_min_version() (in module nemo_automodel.shared.import_utils) is_tensor_unallocated() (in module nemo_automodel.components.models.common.utils) is_tied_word_embeddings() (in module nemo_automodel.components.checkpoint.utils) is_torch_min_version() (in module nemo_automodel.shared.import_utils) is_transformers_conversion_available() (in module nemo_automodel.components.checkpoint.conversion_mapping) is_transformers_min_version() (in module nemo_automodel.shared.import_utils) is_unavailable() (in module nemo_automodel.shared.import_utils) is_val_step (nemo_automodel.components.training.step_scheduler.StepScheduler property) is_zipped_list() (in module nemo_automodel.components.datasets.llm.megatron_dataset) iter_maximal_uniform_dtype_subtrees() (in module nemo_automodel.components.distributed.parallelizer_utils)

J

json2token() (in module nemo_automodel.components.datasets.vlm.utils)

K

KDLoss (class in nemo_automodel.components.loss.kd_loss) keep_fp8_transpose_cache (nemo_automodel.components.distributed.config.MegatronFSDPConfig attribute) keys_to_ignore_at_inference (nemo_automodel.components.models.deepseek_v32.config.DeepseekV32Config attribute) (nemo_automodel.components.models.mistral3.model.Ministral3Config attribute) kimi_k25_vl_collate_fn() (in module nemo_automodel.components.datasets.vlm.collate_fns) kimi_vl_collate_fn() (in module nemo_automodel.components.datasets.vlm.collate_fns) KimiK25VLConfig (class in nemo_automodel.components.models.kimi_k25_vl.model) KimiK25VLForConditionalGeneration (class in nemo_automodel.components.models.kimi_k25_vl.model) KimiK25VLLanguageModelBackend (class in nemo_automodel.components.models.kimi_k25_vl.model)

KimiK25VLModel (class in nemo_automodel.components.models.kimi_k25_vl.model) KimiK25VLMultiModalProjector (class in nemo_automodel.components.models.kimi_k25_vl.model) KimiK25VLStateDictAdapter (class in nemo_automodel.components.models.kimi_k25_vl.state_dict_adapter) KimiVLConfig (class in nemo_automodel.components.models.kimivl.model) KimiVLForConditionalGeneration (class in nemo_automodel.components.models.kimivl.model) KimiVLLanguageModelBackend (class in nemo_automodel.components.models.kimivl.model) KimiVLModel (class in nemo_automodel.components.models.kimivl.model) KimiVLMultiModalProjector (class in nemo_automodel.components.models.kimivl.model) KimiVLStateDictAdapter (class in nemo_automodel.components.models.kimivl.model)

L

layers (nemo_automodel.components.models.kimi_k25_vl.model.KimiK25VLLanguageModelBackend property) (nemo_automodel.components.models.kimi_k25_vl.model.KimiK25VLModel property) (nemo_automodel.components.models.kimivl.model.KimiVLLanguageModelBackend property) (nemo_automodel.components.models.kimivl.model.KimiVLModel property) (nemo_automodel.components.models.qwen3_5_moe.model.Qwen3_5MoeModel property) (nemo_automodel.components.models.qwen3_vl_moe.model.Qwen3VLMoeModel property) layers_per_stage (nemo_automodel.components.distributed.pipelining.config.PipelineConfig attribute) Learnable2DInterpPosEmb (class in nemo_automodel.components.models.kimivl.model) Learnable2DInterpPosEmbDividedFixed (class in nemo_automodel.components.models.kimi_k25_vl.model) LEGACY_MAGIC (in module nemo_automodel.components.datasets.llm.nanogpt_dataset) length (nemo_automodel.components.checkpoint._backports.filesystem._StorageInfo attribute) (nemo_automodel.components.checkpoint._backports.hf_utils._HFStorageInfo attribute) liger_kernel_trf (in module nemo_automodel._transformers.kernel_patches) linear (nemo_automodel.components.models.common.utils.BackendConfig attribute) LinearLoRA (class in nemo_automodel.components._peft.lora) list_stage_modules() (nemo_automodel.components.distributed.pipelining.autopipeline.AutoPipeline method) llama2_flops() (in module nemo_automodel.components.utils.flops_utils) llama3_flops() (in module nemo_automodel.components.utils.flops_utils) LlamaAttention (class in nemo_automodel.components.models.llama.model) LlamaBidirectionalConfig (class in nemo_automodel.components.models.llama_bidirectional.model) LlamaBidirectionalForSequenceClassification (class in nemo_automodel.components.models.llama_bidirectional.model) LlamaBidirectionalModel (class in nemo_automodel.components.models.llama_bidirectional.model) LlamaDecoderLayer (class in nemo_automodel.components.models.llama.model) LlamaForCausalLM (class in nemo_automodel.components.models.llama.model) LlamaMLP (class in nemo_automodel.components.models.llama.model) LlamaModel (class in nemo_automodel.components.models.llama.model) LlamaPreTrainedModel (class in nemo_automodel.components.models.llama.model) LlamaRotaryEmbedding (class in nemo_automodel.components.models.llama.rope_utils) LlamaStateDictAdapter (class in nemo_automodel.components.models.llama.state_dict_adapter) lm_head (nemo_automodel.components.models.kimi_k25_vl.model.KimiK25VLForConditionalGeneration property) (nemo_automodel.components.models.kimivl.model.KimiVLForConditionalGeneration property) lm_head_precision (nemo_automodel.components.moe.config.MoEParallelizerConfig attribute) load_base_model() (nemo_automodel.components.checkpoint.checkpointing.Checkpointer method) load_bin_shard() (in module nemo_automodel.components.datasets.llm.nanogpt_dataset) load_bytes() (nemo_automodel.components.checkpoint._backports.default_planner.DefaultLoadPlanner method) load_corpus() (in module nemo_automodel.components.datasets.llm.retrieval_dataset) load_corpus_metadata() (in module nemo_automodel.components.datasets.llm.retrieval_dataset) load_datasets() (in module nemo_automodel.components.datasets.llm.retrieval_dataset) (in module nemo_automodel.components.datasets.llm.retrieval_dataset_inline) load_model() (nemo_automodel.components.checkpoint.checkpointing.Checkpointer method) load_module_from_file() (in module nemo_automodel.components.config.loader) load_on_dp_ranks() (nemo_automodel.components.checkpoint.checkpointing.Checkpointer method) load_optimizer() (nemo_automodel.components.checkpoint.checkpointing.Checkpointer method) load_state_dict() (nemo_automodel.components.checkpoint.stateful_wrappers.ModelState method) (nemo_automodel.components.checkpoint.stateful_wrappers.OptimizerState method) (nemo_automodel.components.optim.scheduler.OptimizerParamScheduler method) (nemo_automodel.components.training.rng.StatefulRNG method) (nemo_automodel.components.training.step_scheduler.StepScheduler method) load_yaml_config() (in module nemo_automodel.components.config.loader) local_torch_dtype() (in module nemo_automodel._transformers.model_init) log() (nemo_automodel.components.loggers.metric_logger.MetricLogger method) (nemo_automodel.components.training.timers.Timers method) log_artifact() (nemo_automodel.components.loggers.mlflow_utils.MLflowLogger method) log_artifacts() (nemo_automodel.components.loggers.mlflow_utils.MLflowLogger method) log_debug_summary() (nemo_automodel.components.distributed.pipelining.autopipeline.AutoPipeline method) log_metrics() (nemo_automodel.components.loggers.mlflow_utils.MLflowLogger method) log_model() (nemo_automodel.components.loggers.mlflow_utils.MLflowLogger method) log_params() (nemo_automodel.components.loggers.mlflow_utils.MLflowLogger method) logger (in module nemo_automodel._diffusers.auto_diffusion_pipeline) (in module nemo_automodel._transformers.auto_model) (in module nemo_automodel._transformers.auto_tokenizer) (in module nemo_automodel._transformers.biencoder) (in module nemo_automodel._transformers.infrastructure) (in module nemo_automodel._transformers.kernel_patches) (in module nemo_automodel._transformers.model_init) (in module nemo_automodel._transformers.registry) (in module nemo_automodel.components._peft.lora) (in module nemo_automodel.components.checkpoint._backports.consolidate_hf_safetensors) (in module nemo_automodel.components.checkpoint._backports.default_planner) (in module nemo_automodel.components.checkpoint._torch_backports) (in module nemo_automodel.components.datasets.llm.column_mapped_text_instruction_dataset) (in module nemo_automodel.components.datasets.llm.column_mapped_text_instruction_iterable_dataset) (in module nemo_automodel.components.datasets.llm.delta_lake_dataset) (in module nemo_automodel.components.datasets.llm.formatting_utils) (in module nemo_automodel.components.datasets.llm.megatron.builder) (in module nemo_automodel.components.datasets.llm.megatron.gpt_dataset) (in module nemo_automodel.components.datasets.llm.megatron.indexed_dataset) (in module nemo_automodel.components.datasets.llm.megatron.megatron_utils) (in module nemo_automodel.components.datasets.llm.megatron_dataset) (in module nemo_automodel.components.datasets.llm.packed_sequence) (in module nemo_automodel.components.datasets.llm.xlam) (in module nemo_automodel.components.datasets.vlm.collate_fns) (in module nemo_automodel.components.distributed.ddp) (in module nemo_automodel.components.distributed.fsdp2) (in module nemo_automodel.components.distributed.megatron_fsdp) (in module nemo_automodel.components.distributed.parallelizer) (in module nemo_automodel.components.distributed.pipelining.autopipeline) (in module nemo_automodel.components.distributed.pipelining.functional) (in module nemo_automodel.components.distributed.pipelining.hf_utils) (in module nemo_automodel.components.distributed.utils) (in module nemo_automodel.components.loggers.log_utils) (in module nemo_automodel.components.loggers.mlflow_utils) (in module nemo_automodel.components.models.common.combined_projection.state_dict_adapter) (in module nemo_automodel.components.models.common.hf_checkpointing_mixin) (in module nemo_automodel.components.models.common.utils) (in module nemo_automodel.components.models.deepseek_v3.state_dict_adapter) (in module nemo_automodel.components.models.glm4_moe.state_dict_adapter) (in module nemo_automodel.components.models.gpt_oss.model)

LOGGER (in module nemo_automodel.components.models.kimi_k25_vl.model) (in module nemo_automodel.components.models.kimi_k25_vl.state_dict_adapter) (in module nemo_automodel.components.models.kimivl.model) logger (in module nemo_automodel.components.models.llama_bidirectional.export_onnx) (in module nemo_automodel.components.models.mistral3.model) (in module nemo_automodel.components.models.nemotron_v3.state_dict_adapter) (in module nemo_automodel.components.models.qwen3_moe.state_dict_adapter) (in module nemo_automodel.components.models.qwen3_next.state_dict_adapter) (in module nemo_automodel.components.models.qwen3_omni_moe.state_dict_adapter) (in module nemo_automodel.components.models.step3p5.state_dict_adapter) (in module nemo_automodel.components.moe.parallelizer) (in module nemo_automodel.components.optim.scheduler) (in module nemo_automodel.components.optim.utils) (in module nemo_automodel.components.quantization.fp8) (in module nemo_automodel.components.quantization.qat) (in module nemo_automodel.components.training.garbage_collection) (in module nemo_automodel.components.training.step_scheduler) (in module nemo_automodel.components.utils.compile_utils) (in module nemo_automodel.components.utils.model_utils) (in module nemo_automodel.shared.import_utils) lookup_object() (nemo_automodel.components.checkpoint._backports.default_planner.DefaultSavePlanner method) lookup_tensor() (nemo_automodel.components.checkpoint._backports.default_planner.DefaultLoadPlanner method) lora_A_init (nemo_automodel.components._peft.lora.PeftConfig attribute) lora_da_dx_kernel() (in module nemo_automodel.components._peft.lora_kernel) lora_da_dx_update_wrapper() (in module nemo_automodel.components._peft.lora_kernel) lora_db_kernel() (in module nemo_automodel.components._peft.lora_kernel) lora_db_update_wrapper() (in module nemo_automodel.components._peft.lora_kernel) lora_dtype (nemo_automodel.components._peft.lora.PeftConfig attribute) lora_forward_kernel() (in module nemo_automodel.components._peft.lora_kernel) lora_forward_wrapper() (in module nemo_automodel.components._peft.lora_kernel) LoRATritonFunction (class in nemo_automodel.components._peft.lora) loss_flops_calculator() (in module nemo_automodel.components.utils.flops_utils) loss_fn (nemo_automodel.components.distributed.pipelining.config.PipelineConfig attribute) ls() (nemo_automodel.components.checkpoint._backports._fsspec_filesystem.FileSystem method) (nemo_automodel.components.checkpoint._backports.filesystem.FileSystem method)

M

MAGIC (in module nemo_automodel.components.datasets.llm.nanogpt_dataset) main() (in module nemo_automodel.components.models.llama_bidirectional.export_onnx) main_input_name (nemo_automodel.components.models.kimi_k25_vl.model.KimiK25VLForConditionalGeneration attribute) (nemo_automodel.components.models.nemotron_parse.model.NemotronParsePreTrainedModel attribute) (nemo_automodel.components.models.nemotron_v3.model.NemotronHForCausalLM attribute) main_loss_backward_scale (nemo_automodel.components.moe.megatron.moe_utils.MoEAuxLossAutoScaler attribute) make_attention_mask_from_labels() (in module nemo_automodel.components.datasets.utils) make_cord_v2_dataset() (in module nemo_automodel.components.datasets.vlm.datasets) make_cp_batch_and_ctx() (in module nemo_automodel.components.distributed.cp_utils) make_cp_batch_for_te() (in module nemo_automodel.components.distributed.cp_utils) make_cv17_dataset() (in module nemo_automodel.components.datasets.vlm.datasets) make_iterable() (in module nemo_automodel.components.datasets.llm.column_mapped_text_instruction_dataset) make_medpix_dataset() (in module nemo_automodel.components.datasets.vlm.datasets) make_rdr_dataset() (in module nemo_automodel.components.datasets.vlm.datasets) make_retrieval_dataset() (in module nemo_automodel.components.datasets.llm.retrieval_dataset) (in module nemo_automodel.components.datasets.llm.retrieval_dataset_inline) make_squad_dataset() (in module nemo_automodel.components.datasets.llm.squad) make_unimm_chat_dataset() (in module nemo_automodel.components.datasets.vlm.datasets) make_vocab() (in module nemo_automodel.components.datasets.llm.mock) (in module nemo_automodel.components.datasets.llm.mock_packed) make_xlam_dataset() (in module nemo_automodel.components.datasets.llm.xlam) mappings (nemo_automodel.components.checkpoint._backports.default_planner.DefaultLoadPlanner attribute) (nemo_automodel.components.checkpoint._backports.default_planner.DefaultSavePlanner attribute) MaskedCrossEntropy (class in nemo_automodel.components.loss.masked_ce) match() (nemo_automodel.components._peft.module_matcher.ModuleMatcher method) match_all_linear (nemo_automodel.components._peft.lora.PeftConfig attribute) (nemo_automodel.components._peft.module_matcher.ModuleMatcher attribute) maybe_te_autocast() (nemo_automodel.components.models.common.utils.TEFp8Config method) maybe_wait_for_staging() (nemo_automodel.components.checkpoint.checkpointing.Checkpointer method) megatron_fsdp_strategy_parallelize() (in module nemo_automodel.components.distributed.parallelizer) megatron_fsdp_unit_modules (nemo_automodel.components.distributed.config.MegatronFSDPConfig attribute) MegatronFSDPConfig (class in nemo_automodel.components.distributed.config) MegatronFSDPManager (class in nemo_automodel.components.distributed.megatron_fsdp) MegatronPretraining (class in nemo_automodel.components.datasets.llm.megatron_dataset) MegatronPretrainingRandomSampler (class in nemo_automodel.components.datasets.llm.megatron.sampler) MegatronPretrainingSampler (class in nemo_automodel.components.datasets.llm.megatron.sampler) MeshAxisName (class in nemo_automodel.components.distributed.mesh) MeshContext (class in nemo_automodel.components.distributed.mesh) metadata (nemo_automodel.components.checkpoint._backports.consolidate_hf_safetensors._InputFileData attribute) (nemo_automodel.components.datasets.llm.retrieval_dataset.CorpusInfo attribute) metadata_path (nemo_automodel.components.checkpoint._backports.filesystem._FileSystemWriter property) (nemo_automodel.components.checkpoint._backports.hf_storage._HuggingFaceStorageWriter property) metadata_size (nemo_automodel.components.checkpoint._backports.consolidate_hf_safetensors._InputFileData attribute) (nemo_automodel.components.checkpoint._backports.consolidate_hf_safetensors._OutputFileData attribute) MetricLogger (class in nemo_automodel.components.loggers.metric_logger) MetricLoggerDist (class in nemo_automodel.components.loggers.metric_logger) metrics (nemo_automodel.components.loggers.metric_logger.MetricsSample attribute) MetricsSample (class in nemo_automodel.components.loggers.metric_logger) mid_level_dataset_surplus (nemo_automodel.components.datasets.llm.megatron.gpt_dataset.BlendedMegatronDatasetConfig attribute) MiniMaxM2Attention (class in nemo_automodel.components.models.minimax_m2.layers) MiniMaxM2ForCausalLM (class in nemo_automodel.components.models.minimax_m2.model) MiniMaxM2Model (class in nemo_automodel.components.models.minimax_m2.model) MiniMaxM2StateDictAdapter (class in nemo_automodel.components.models.minimax_m2.state_dict_adapter) Ministral3Attention (class in nemo_automodel.components.models.mistral3.model) Ministral3CausalLMOutputWithPast (class in nemo_automodel.components.models.mistral3.model) Ministral3Config (class in nemo_automodel.components.models.mistral3.model) Ministral3DecoderLayer (class in nemo_automodel.components.models.mistral3.model) Ministral3ForCausalLM (class in nemo_automodel.components.models.mistral3.model) Ministral3MLP (class in nemo_automodel.components.models.mistral3.model) Ministral3Model (class in nemo_automodel.components.models.mistral3.model) Ministral3ModelOutputWithPast (class in nemo_automodel.components.models.mistral3.model) Ministral3PreTrainedModel (class in nemo_automodel.components.models.mistral3.model) Ministral3RMSNorm (class in nemo_automodel.components.models.mistral3.model) Ministral3RotaryEmbedding (class in nemo_automodel.components.models.mistral3.model) MISSING_CUT_CROSS_ENTROPY_MSG (in module nemo_automodel.shared.import_utils) MISSING_QWEN_VL_UTILS_MSG (in module nemo_automodel.shared.import_utils) MISSING_TE_PARALLEL_CE_MSG (in module nemo_automodel.components.loss.te_parallel_ce) MISSING_TORCHAO_MSG (in module nemo_automodel.shared.import_utils) MISSING_TRITON_MSG (in module nemo_automodel.shared.import_utils) mixtral_flops() (in module nemo_automodel.components.utils.flops_utils) mkdir() (nemo_automodel.components.checkpoint._backports._fsspec_filesystem.FileSystem method) (nemo_automodel.components.checkpoint._backports.filesystem.FileSystem method) (nemo_automodel.components.checkpoint._backports.filesystem.FileSystemBase method) MLA (class in nemo_automodel.components.models.deepseek_v3.layers) MLflowLogger (class in nemo_automodel.components.loggers.mlflow_utils) MLP (class in nemo_automodel.components.models.gpt2) (class in nemo_automodel.components.moe.layers) mlp (nemo_automodel.components.models.nemotron_v3.layers.NemotronV3Block property) mmap_bin_files (nemo_automodel.components.datasets.llm.megatron.gpt_dataset.BlendedMegatronDatasetConfig attribute) mock (nemo_automodel.components.datasets.llm.megatron.gpt_dataset.BlendedMegatronDatasetConfig attribute) MockIterableDataset (class in nemo_automodel.components.datasets.llm.mock_iterable_dataset) mode (nemo_automodel.components.moe.config.MoEMetricsConfig attribute) (nemo_automodel.components.utils.compile_utils.CompileConfig attribute) model_arch_name_to_cls (nemo_automodel._transformers.registry._ModelRegistry attribute) model_cache_dir (nemo_automodel.components.checkpoint.checkpointing.CheckpointingConfig attribute) model_parts (nemo_automodel.components.distributed.pipelining.autopipeline.PipelineInfo attribute) model_repo_id (nemo_automodel.components.checkpoint.checkpointing.CheckpointingConfig attribute) model_save_format (nemo_automodel.components.checkpoint.checkpointing.CheckpointingConfig attribute) model_state_dict_keys (nemo_automodel.components.checkpoint.checkpointing.CheckpointingConfig attribute) model_type (nemo_automodel.components.models.deepseek_v32.config.DeepseekV32Config attribute) (nemo_automodel.components.models.kimi_k25_vl.model.KimiK25VLConfig attribute) (nemo_automodel.components.models.kimi_k25_vl.model.MoonViT3dConfig attribute) (nemo_automodel.components.models.kimivl.model.KimiVLConfig attribute) (nemo_automodel.components.models.kimivl.model.MoonViTConfig attribute) (nemo_automodel.components.models.llama_bidirectional.model.LlamaBidirectionalConfig attribute) (nemo_automodel.components.models.mistral3.model.Ministral3Config attribute) (nemo_automodel.components.models.nemotron_parse.model.NemotronParseConfig attribute) (nemo_automodel.components.models.nemotron_parse.model.NemotronParseEncoderConfig attribute) (nemo_automodel.components.models.nemotron_parse.model.NemotronParseTextConfig attribute) ModelClass (in module nemo_automodel.components.models.deepseek_v3.model) (in module nemo_automodel.components.models.deepseek_v32.model) (in module nemo_automodel.components.models.glm4_moe.model) (in module nemo_automodel.components.models.glm4_moe_lite.model) (in module nemo_automodel.components.models.gpt_oss.model) (in module nemo_automodel.components.models.kimi_k25_vl.model) (in module nemo_automodel.components.models.kimivl.model) (in module nemo_automodel.components.models.llama.model) (in module nemo_automodel.components.models.llama_bidirectional.model) (in module nemo_automodel.components.models.minimax_m2) (in module nemo_automodel.components.models.minimax_m2.model) (in module nemo_automodel.components.models.mistral3.model) (in module nemo_automodel.components.models.nemotron_parse.model) (in module nemo_automodel.components.models.nemotron_v3.model) (in module nemo_automodel.components.models.qwen2.model) (in module nemo_automodel.components.models.qwen3_moe.model) (in module nemo_automodel.components.models.qwen3_next.model) (in module nemo_automodel.components.models.qwen3_omni_moe.model) (in module nemo_automodel.components.models.qwen3_vl_moe.model) (in module nemo_automodel.components.models.step3p5) (in module nemo_automodel.components.models.step3p5.model) MODELING_PATH (in module nemo_automodel._transformers.registry) modeling_path (nemo_automodel._transformers.registry._ModelRegistry attribute) ModelRegistry (in module nemo_automodel._transformers.registry) MODELS_REQUIRING_TENSOR_MERGING (in module nemo_automodel.components.checkpoint.conversion_mapping) ModelState (class in nemo_automodel.components.checkpoint.stateful_wrappers) module nemo_automodel nemo_automodel._diffusers nemo_automodel._diffusers.auto_diffusion_pipeline nemo_automodel._transformers nemo_automodel._transformers.auto_model nemo_automodel._transformers.auto_tokenizer nemo_automodel._transformers.biencoder nemo_automodel._transformers.infrastructure nemo_automodel._transformers.kernel_patches nemo_automodel._transformers.model_init nemo_automodel._transformers.registry nemo_automodel._transformers.utils nemo_automodel.autonvtx nemo_automodel.components nemo_automodel.components._peft nemo_automodel.components._peft.lora nemo_automodel.components._peft.lora_kernel nemo_automodel.components._peft.lora_moe nemo_automodel.components._peft.module_matcher nemo_automodel.components.attention nemo_automodel.components.attention.flex_attention nemo_automodel.components.attention.utils nemo_automodel.components.checkpoint nemo_automodel.components.checkpoint._backports nemo_automodel.components.checkpoint._backports._fsspec_filesystem nemo_automodel.components.checkpoint._backports._version nemo_automodel.components.checkpoint._backports.consolidate_hf_safetensors nemo_automodel.components.checkpoint._backports.default_planner nemo_automodel.components.checkpoint._backports.filesystem nemo_automodel.components.checkpoint._backports.hf_storage nemo_automodel.components.checkpoint._backports.hf_utils nemo_automodel.components.checkpoint._backports.planner_helpers nemo_automodel.components.checkpoint._torch_backports nemo_automodel.components.checkpoint.addons nemo_automodel.components.checkpoint.checkpointing nemo_automodel.components.checkpoint.conversion_mapping nemo_automodel.components.checkpoint.state_dict_adapter nemo_automodel.components.checkpoint.stateful_wrappers nemo_automodel.components.checkpoint.utils nemo_automodel.components.config nemo_automodel.components.config._arg_parser nemo_automodel.components.config.loader nemo_automodel.components.datasets nemo_automodel.components.datasets.llm nemo_automodel.components.datasets.llm.chat_dataset nemo_automodel.components.datasets.llm.column_mapped_text_instruction_dataset nemo_automodel.components.datasets.llm.column_mapped_text_instruction_iterable_dataset nemo_automodel.components.datasets.llm.delta_lake_dataset nemo_automodel.components.datasets.llm.formatting_utils nemo_automodel.components.datasets.llm.hellaswag nemo_automodel.components.datasets.llm.megatron nemo_automodel.components.datasets.llm.megatron.builder nemo_automodel.components.datasets.llm.megatron.gpt_dataset nemo_automodel.components.datasets.llm.megatron.helpers nemo_automodel.components.datasets.llm.megatron.indexed_dataset nemo_automodel.components.datasets.llm.megatron.megatron_utils nemo_automodel.components.datasets.llm.megatron.sampler nemo_automodel.components.datasets.llm.megatron_dataset nemo_automodel.components.datasets.llm.mock nemo_automodel.components.datasets.llm.mock_iterable_dataset nemo_automodel.components.datasets.llm.mock_packed nemo_automodel.components.datasets.llm.nanogpt_dataset nemo_automodel.components.datasets.llm.packed_sequence nemo_automodel.components.datasets.llm.retrieval_collator nemo_automodel.components.datasets.llm.retrieval_dataset nemo_automodel.components.datasets.llm.retrieval_dataset_inline nemo_automodel.components.datasets.llm.seq_cls nemo_automodel.components.datasets.llm.squad nemo_automodel.components.datasets.llm.xlam nemo_automodel.components.datasets.reservoir_sampler nemo_automodel.components.datasets.utils nemo_automodel.components.datasets.vlm nemo_automodel.components.datasets.vlm.collate_fns nemo_automodel.components.datasets.vlm.datasets nemo_automodel.components.datasets.vlm.utils nemo_automodel.components.distributed nemo_automodel.components.distributed.config nemo_automodel.components.distributed.cp_utils nemo_automodel.components.distributed.ddp nemo_automodel.components.distributed.fsdp2 nemo_automodel.components.distributed.grad_utils nemo_automodel.components.distributed.init_utils nemo_automodel.components.distributed.megatron_fsdp nemo_automodel.components.distributed.mesh nemo_automodel.components.distributed.mesh_utils nemo_automodel.components.distributed.optimized_tp_plans nemo_automodel.components.distributed.parallel_styles nemo_automodel.components.distributed.parallelizer nemo_automodel.components.distributed.parallelizer_utils nemo_automodel.components.distributed.pipelining nemo_automodel.components.distributed.pipelining.autopipeline nemo_automodel.components.distributed.pipelining.config nemo_automodel.components.distributed.pipelining.functional nemo_automodel.components.distributed.pipelining.hf_utils nemo_automodel.components.distributed.tensor_utils nemo_automodel.components.distributed.thd_utils nemo_automodel.components.distributed.utils nemo_automodel.components.launcher nemo_automodel.components.loggers nemo_automodel.components.loggers.log_utils nemo_automodel.components.loggers.metric_logger nemo_automodel.components.loggers.mlflow_utils nemo_automodel.components.loggers.wandb_utils nemo_automodel.components.loss nemo_automodel.components.loss.chunked_ce nemo_automodel.components.loss.kd_loss nemo_automodel.components.loss.linear_ce nemo_automodel.components.loss.masked_ce nemo_automodel.components.loss.te_parallel_ce nemo_automodel.components.models nemo_automodel.components.models.common nemo_automodel.components.models.common.bidirectional nemo_automodel.components.models.common.combined_projection nemo_automodel.components.models.common.combined_projection.combined_mlp nemo_automodel.components.models.common.combined_projection.combined_qkv nemo_automodel.components.models.common.combined_projection.state_dict_adapter nemo_automodel.components.models.common.hf_checkpointing_mixin nemo_automodel.components.models.common.utils nemo_automodel.components.models.deepseek_v3 nemo_automodel.components.models.deepseek_v3.layers nemo_automodel.components.models.deepseek_v3.model nemo_automodel.components.models.deepseek_v3.rope_utils nemo_automodel.components.models.deepseek_v3.state_dict_adapter nemo_automodel.components.models.deepseek_v32 nemo_automodel.components.models.deepseek_v32.config nemo_automodel.components.models.deepseek_v32.layers nemo_automodel.components.models.deepseek_v32.model nemo_automodel.components.models.deepseek_v32.state_dict_adapter nemo_automodel.components.models.glm4_moe nemo_automodel.components.models.glm4_moe.layers nemo_automodel.components.models.glm4_moe.model nemo_automodel.components.models.glm4_moe.state_dict_adapter nemo_automodel.components.models.glm4_moe_lite nemo_automodel.components.models.glm4_moe_lite.model nemo_automodel.components.models.gpt2 nemo_automodel.components.models.gpt_oss nemo_automodel.components.models.gpt_oss.layers nemo_automodel.components.models.gpt_oss.model nemo_automodel.components.models.gpt_oss.rope_utils nemo_automodel.components.models.gpt_oss.state_dict_adapter nemo_automodel.components.models.kimi_k25_vl nemo_automodel.components.models.kimi_k25_vl.model nemo_automodel.components.models.kimi_k25_vl.state_dict_adapter nemo_automodel.components.models.kimivl nemo_automodel.components.models.kimivl.model nemo_automodel.components.models.llama nemo_automodel.components.models.llama.model nemo_automodel.components.models.llama.rope_utils nemo_automodel.components.models.llama.state_dict_adapter nemo_automodel.components.models.llama_bidirectional nemo_automodel.components.models.llama_bidirectional.export_onnx nemo_automodel.components.models.llama_bidirectional.model nemo_automodel.components.models.minimax_m2 nemo_automodel.components.models.minimax_m2.layers nemo_automodel.components.models.minimax_m2.model nemo_automodel.components.models.minimax_m2.state_dict_adapter nemo_automodel.components.models.mistral3 nemo_automodel.components.models.mistral3.model nemo_automodel.components.models.nemotron_parse nemo_automodel.components.models.nemotron_parse.model nemo_automodel.components.models.nemotron_parse.nemotron_parse_loss nemo_automodel.components.models.nemotron_v3 nemo_automodel.components.models.nemotron_v3.cache nemo_automodel.components.models.nemotron_v3.layers nemo_automodel.components.models.nemotron_v3.model nemo_automodel.components.models.nemotron_v3.state_dict_adapter nemo_automodel.components.models.qwen2 nemo_automodel.components.models.qwen2.model nemo_automodel.components.models.qwen2.state_dict_adapter nemo_automodel.components.models.qwen3_5_moe nemo_automodel.components.models.qwen3_5_moe.model nemo_automodel.components.models.qwen3_5_moe.state_dict_adapter nemo_automodel.components.models.qwen3_moe nemo_automodel.components.models.qwen3_moe.layers nemo_automodel.components.models.qwen3_moe.model nemo_automodel.components.models.qwen3_moe.state_dict_adapter nemo_automodel.components.models.qwen3_next nemo_automodel.components.models.qwen3_next.layers nemo_automodel.components.models.qwen3_next.model nemo_automodel.components.models.qwen3_next.state_dict_adapter nemo_automodel.components.models.qwen3_omni_moe nemo_automodel.components.models.qwen3_omni_moe.model nemo_automodel.components.models.qwen3_omni_moe.state_dict_adapter nemo_automodel.components.models.qwen3_vl_moe nemo_automodel.components.models.qwen3_vl_moe.model nemo_automodel.components.models.qwen3_vl_moe.state_dict_adapter nemo_automodel.components.models.step3p5 nemo_automodel.components.models.step3p5.layers nemo_automodel.components.models.step3p5.model nemo_automodel.components.models.step3p5.state_dict_adapter nemo_automodel.components.moe nemo_automodel.components.moe.config nemo_automodel.components.moe.experts nemo_automodel.components.moe.fsdp_mixin nemo_automodel.components.moe.layers nemo_automodel.components.moe.load_balance_metrics nemo_automodel.components.moe.megatron nemo_automodel.components.moe.megatron.fused_a2a nemo_automodel.components.moe.megatron.fused_indices_converter nemo_automodel.components.moe.megatron.moe_utils nemo_automodel.components.moe.megatron.token_dispatcher nemo_automodel.components.moe.parallelizer nemo_automodel.components.moe.state_dict_mixin nemo_automodel.components.moe.state_dict_utils nemo_automodel.components.optim nemo_automodel.components.optim.scheduler nemo_automodel.components.optim.utils nemo_automodel.components.quantization nemo_automodel.components.quantization.fp8 nemo_automodel.components.quantization.qat nemo_automodel.components.quantization.qlora nemo_automodel.components.training nemo_automodel.components.training.garbage_collection nemo_automodel.components.training.model_output_utils nemo_automodel.components.training.rng nemo_automodel.components.training.signal_handler nemo_automodel.components.training.step_scheduler nemo_automodel.components.training.timers nemo_automodel.components.training.utils nemo_automodel.components.utils nemo_automodel.components.utils.compile_utils nemo_automodel.components.utils.flops_utils nemo_automodel.components.utils.model_utils nemo_automodel.components.utils.yaml_utils nemo_automodel.shared nemo_automodel.shared.import_utils nemo_automodel.shared.torch_patches nemo_automodel.shared.utils

module_filter() (in module nemo_automodel.components.loggers.log_utils) module_fqns_per_model_part (nemo_automodel.components.distributed.pipelining.config.PipelineConfig attribute) ModuleMatcher (class in nemo_automodel.components._peft.module_matcher) MoE (class in nemo_automodel.components.moe.layers) moe_config (nemo_automodel.components.distributed.mesh.MeshContext attribute) moe_enable_deepep (nemo_automodel.components.moe.megatron.token_dispatcher.TokenDispatcherConfig attribute) moe_expert_capacity_factor (nemo_automodel.components.moe.megatron.token_dispatcher.TokenDispatcherConfig attribute) moe_inter_dim (nemo_automodel.components.moe.config.MoEConfig attribute) moe_mesh (nemo_automodel.components.distributed.mesh.MeshContext attribute) moe_mlp_flops_calculator() (in module nemo_automodel.components.utils.flops_utils) moe_permute_fusion (nemo_automodel.components.moe.megatron.token_dispatcher.TokenDispatcherConfig attribute) moe_router_dtype (nemo_automodel.components.moe.megatron.token_dispatcher.TokenDispatcherConfig attribute) moe_router_expert_pad_multiple (nemo_automodel.components.moe.megatron.token_dispatcher.TokenDispatcherConfig attribute) moe_router_topk (nemo_automodel.components.moe.megatron.token_dispatcher.TokenDispatcherConfig attribute) MoEAuxLossAutoScaler (class in nemo_automodel.components.moe.megatron.moe_utils) MoEConfig (class in nemo_automodel.components.moe.config) MoEFlexTokenDispatcher (class in nemo_automodel.components.moe.megatron.token_dispatcher) MoEFSDPSyncMixin (class in nemo_automodel.components.moe.fsdp_mixin) MoEMetricsConfig (class in nemo_automodel.components.moe.config) MoEParallelizerConfig (class in nemo_automodel.components.moe.config) MoESplitExpertsStateDictMixin (class in nemo_automodel.components.moe.state_dict_mixin) MoonVision3dPatchEmbed (class in nemo_automodel.components.models.kimi_k25_vl.model) MoonVisionPatchEmbed (class in nemo_automodel.components.models.kimivl.model) MoonViT3dConfig (class in nemo_automodel.components.models.kimi_k25_vl.model) MoonViT3dEncoder (class in nemo_automodel.components.models.kimi_k25_vl.model) MoonViT3dEncoderLayer (class in nemo_automodel.components.models.kimi_k25_vl.model) MoonViT3dMLP (class in nemo_automodel.components.models.kimi_k25_vl.model) MoonViT3dPretrainedModel (class in nemo_automodel.components.models.kimi_k25_vl.model) MoonViTConfig (class in nemo_automodel.components.models.kimivl.model) MoonVitEncoder (class in nemo_automodel.components.models.kimivl.model) MoonVitEncoderLayer (class in nemo_automodel.components.models.kimivl.model) MoonVitMLP (class in nemo_automodel.components.models.kimivl.model) MoonVitPretrainedModel (class in nemo_automodel.components.models.kimivl.model) move_to_device() (in module nemo_automodel.components.training.utils) mp_policy (nemo_automodel.components.distributed.config.FSDP2Config attribute) (nemo_automodel.components.moe.config.MoEParallelizerConfig attribute) MULTIMODAL_SUFFIXES (in module nemo_automodel.components.distributed.pipelining.hf_utils)

N

n_activated_experts (nemo_automodel.components.moe.config.MoEConfig attribute) n_expert_groups (nemo_automodel.components.moe.config.MoEConfig attribute) n_limited_groups (nemo_automodel.components.moe.config.MoEConfig attribute) n_routed_experts (nemo_automodel.components.moe.config.MoEConfig attribute) n_shared_experts (nemo_automodel.components.moe.config.MoEConfig attribute) naming_override (nemo_automodel._transformers.registry._ModelRegistry attribute) NanogptDataset (class in nemo_automodel.components.datasets.llm.nanogpt_dataset) nccl_ub (nemo_automodel.components.distributed.config.MegatronFSDPConfig attribute) nemo_automodel module nemo_automodel._diffusers module nemo_automodel._diffusers.auto_diffusion_pipeline module nemo_automodel._transformers module nemo_automodel._transformers.auto_model module nemo_automodel._transformers.auto_tokenizer module nemo_automodel._transformers.biencoder module nemo_automodel._transformers.infrastructure module nemo_automodel._transformers.kernel_patches module nemo_automodel._transformers.model_init module nemo_automodel._transformers.registry module nemo_automodel._transformers.utils module nemo_automodel.autonvtx module nemo_automodel.components module nemo_automodel.components._peft module nemo_automodel.components._peft.lora module nemo_automodel.components._peft.lora_kernel module nemo_automodel.components._peft.lora_moe module nemo_automodel.components._peft.module_matcher module nemo_automodel.components.attention module nemo_automodel.components.attention.flex_attention module nemo_automodel.components.attention.utils module nemo_automodel.components.checkpoint module nemo_automodel.components.checkpoint._backports module nemo_automodel.components.checkpoint._backports._fsspec_filesystem module nemo_automodel.components.checkpoint._backports._version module nemo_automodel.components.checkpoint._backports.consolidate_hf_safetensors module nemo_automodel.components.checkpoint._backports.default_planner module nemo_automodel.components.checkpoint._backports.filesystem module nemo_automodel.components.checkpoint._backports.hf_storage module nemo_automodel.components.checkpoint._backports.hf_utils module nemo_automodel.components.checkpoint._backports.planner_helpers module nemo_automodel.components.checkpoint._torch_backports module nemo_automodel.components.checkpoint.addons module nemo_automodel.components.checkpoint.checkpointing module nemo_automodel.components.checkpoint.conversion_mapping module nemo_automodel.components.checkpoint.state_dict_adapter module nemo_automodel.components.checkpoint.stateful_wrappers module nemo_automodel.components.checkpoint.utils module nemo_automodel.components.config module nemo_automodel.components.config._arg_parser module nemo_automodel.components.config.loader module nemo_automodel.components.datasets module nemo_automodel.components.datasets.llm module nemo_automodel.components.datasets.llm.chat_dataset module nemo_automodel.components.datasets.llm.column_mapped_text_instruction_dataset module nemo_automodel.components.datasets.llm.column_mapped_text_instruction_iterable_dataset module nemo_automodel.components.datasets.llm.delta_lake_dataset module nemo_automodel.components.datasets.llm.formatting_utils module nemo_automodel.components.datasets.llm.hellaswag module nemo_automodel.components.datasets.llm.megatron module nemo_automodel.components.datasets.llm.megatron.builder module nemo_automodel.components.datasets.llm.megatron.gpt_dataset module nemo_automodel.components.datasets.llm.megatron.helpers module nemo_automodel.components.datasets.llm.megatron.indexed_dataset module nemo_automodel.components.datasets.llm.megatron.megatron_utils module nemo_automodel.components.datasets.llm.megatron.sampler module nemo_automodel.components.datasets.llm.megatron_dataset module nemo_automodel.components.datasets.llm.mock module nemo_automodel.components.datasets.llm.mock_iterable_dataset module nemo_automodel.components.datasets.llm.mock_packed module nemo_automodel.components.datasets.llm.nanogpt_dataset module nemo_automodel.components.datasets.llm.packed_sequence module nemo_automodel.components.datasets.llm.retrieval_collator module nemo_automodel.components.datasets.llm.retrieval_dataset module nemo_automodel.components.datasets.llm.retrieval_dataset_inline module nemo_automodel.components.datasets.llm.seq_cls module nemo_automodel.components.datasets.llm.squad module nemo_automodel.components.datasets.llm.xlam module nemo_automodel.components.datasets.reservoir_sampler module nemo_automodel.components.datasets.utils module nemo_automodel.components.datasets.vlm module nemo_automodel.components.datasets.vlm.collate_fns module nemo_automodel.components.datasets.vlm.datasets module nemo_automodel.components.datasets.vlm.utils module nemo_automodel.components.distributed module nemo_automodel.components.distributed.config module nemo_automodel.components.distributed.cp_utils module nemo_automodel.components.distributed.ddp module nemo_automodel.components.distributed.fsdp2 module nemo_automodel.components.distributed.grad_utils module nemo_automodel.components.distributed.init_utils module nemo_automodel.components.distributed.megatron_fsdp module nemo_automodel.components.distributed.mesh module nemo_automodel.components.distributed.mesh_utils module nemo_automodel.components.distributed.optimized_tp_plans module nemo_automodel.components.distributed.parallel_styles module nemo_automodel.components.distributed.parallelizer module nemo_automodel.components.distributed.parallelizer_utils module nemo_automodel.components.distributed.pipelining module nemo_automodel.components.distributed.pipelining.autopipeline module nemo_automodel.components.distributed.pipelining.config module nemo_automodel.components.distributed.pipelining.functional module nemo_automodel.components.distributed.pipelining.hf_utils module nemo_automodel.components.distributed.tensor_utils module nemo_automodel.components.distributed.thd_utils module nemo_automodel.components.distributed.utils module nemo_automodel.components.launcher module nemo_automodel.components.loggers module nemo_automodel.components.loggers.log_utils module nemo_automodel.components.loggers.metric_logger module nemo_automodel.components.loggers.mlflow_utils module nemo_automodel.components.loggers.wandb_utils module nemo_automodel.components.loss module nemo_automodel.components.loss.chunked_ce module nemo_automodel.components.loss.kd_loss module nemo_automodel.components.loss.linear_ce module nemo_automodel.components.loss.masked_ce module nemo_automodel.components.loss.te_parallel_ce module nemo_automodel.components.models module nemo_automodel.components.models.common module nemo_automodel.components.models.common.bidirectional module nemo_automodel.components.models.common.combined_projection module nemo_automodel.components.models.common.combined_projection.combined_mlp module nemo_automodel.components.models.common.combined_projection.combined_qkv module nemo_automodel.components.models.common.combined_projection.state_dict_adapter module nemo_automodel.components.models.common.hf_checkpointing_mixin module nemo_automodel.components.models.common.utils module nemo_automodel.components.models.deepseek_v3 module nemo_automodel.components.models.deepseek_v3.layers module nemo_automodel.components.models.deepseek_v3.model module nemo_automodel.components.models.deepseek_v3.rope_utils module nemo_automodel.components.models.deepseek_v3.state_dict_adapter module nemo_automodel.components.models.deepseek_v32 module

nemo_automodel.components.models.deepseek_v32.config module nemo_automodel.components.models.deepseek_v32.layers module nemo_automodel.components.models.deepseek_v32.model module nemo_automodel.components.models.deepseek_v32.state_dict_adapter module nemo_automodel.components.models.glm4_moe module nemo_automodel.components.models.glm4_moe.layers module nemo_automodel.components.models.glm4_moe.model module nemo_automodel.components.models.glm4_moe.state_dict_adapter module nemo_automodel.components.models.glm4_moe_lite module nemo_automodel.components.models.glm4_moe_lite.model module nemo_automodel.components.models.gpt2 module nemo_automodel.components.models.gpt_oss module nemo_automodel.components.models.gpt_oss.layers module nemo_automodel.components.models.gpt_oss.model module nemo_automodel.components.models.gpt_oss.rope_utils module nemo_automodel.components.models.gpt_oss.state_dict_adapter module nemo_automodel.components.models.kimi_k25_vl module nemo_automodel.components.models.kimi_k25_vl.model module nemo_automodel.components.models.kimi_k25_vl.state_dict_adapter module nemo_automodel.components.models.kimivl module nemo_automodel.components.models.kimivl.model module nemo_automodel.components.models.llama module nemo_automodel.components.models.llama.model module nemo_automodel.components.models.llama.rope_utils module nemo_automodel.components.models.llama.state_dict_adapter module nemo_automodel.components.models.llama_bidirectional module nemo_automodel.components.models.llama_bidirectional.export_onnx module nemo_automodel.components.models.llama_bidirectional.model module nemo_automodel.components.models.minimax_m2 module nemo_automodel.components.models.minimax_m2.layers module nemo_automodel.components.models.minimax_m2.model module nemo_automodel.components.models.minimax_m2.state_dict_adapter module nemo_automodel.components.models.mistral3 module nemo_automodel.components.models.mistral3.model module nemo_automodel.components.models.nemotron_parse module nemo_automodel.components.models.nemotron_parse.model module nemo_automodel.components.models.nemotron_parse.nemotron_parse_loss module nemo_automodel.components.models.nemotron_v3 module nemo_automodel.components.models.nemotron_v3.cache module nemo_automodel.components.models.nemotron_v3.layers module nemo_automodel.components.models.nemotron_v3.model module nemo_automodel.components.models.nemotron_v3.state_dict_adapter module nemo_automodel.components.models.qwen2 module nemo_automodel.components.models.qwen2.model module nemo_automodel.components.models.qwen2.state_dict_adapter module nemo_automodel.components.models.qwen3_5_moe module nemo_automodel.components.models.qwen3_5_moe.model module nemo_automodel.components.models.qwen3_5_moe.state_dict_adapter module nemo_automodel.components.models.qwen3_moe module nemo_automodel.components.models.qwen3_moe.layers module nemo_automodel.components.models.qwen3_moe.model module nemo_automodel.components.models.qwen3_moe.state_dict_adapter module nemo_automodel.components.models.qwen3_next module nemo_automodel.components.models.qwen3_next.layers module nemo_automodel.components.models.qwen3_next.model module nemo_automodel.components.models.qwen3_next.state_dict_adapter module nemo_automodel.components.models.qwen3_omni_moe module nemo_automodel.components.models.qwen3_omni_moe.model module nemo_automodel.components.models.qwen3_omni_moe.state_dict_adapter module nemo_automodel.components.models.qwen3_vl_moe module nemo_automodel.components.models.qwen3_vl_moe.model module nemo_automodel.components.models.qwen3_vl_moe.state_dict_adapter module nemo_automodel.components.models.step3p5 module nemo_automodel.components.models.step3p5.layers module nemo_automodel.components.models.step3p5.model module nemo_automodel.components.models.step3p5.state_dict_adapter module nemo_automodel.components.moe module nemo_automodel.components.moe.config module nemo_automodel.components.moe.experts module nemo_automodel.components.moe.fsdp_mixin module nemo_automodel.components.moe.layers module nemo_automodel.components.moe.load_balance_metrics module nemo_automodel.components.moe.megatron module nemo_automodel.components.moe.megatron.fused_a2a module nemo_automodel.components.moe.megatron.fused_indices_converter module nemo_automodel.components.moe.megatron.moe_utils module nemo_automodel.components.moe.megatron.token_dispatcher module nemo_automodel.components.moe.parallelizer module nemo_automodel.components.moe.state_dict_mixin module nemo_automodel.components.moe.state_dict_utils module nemo_automodel.components.optim module nemo_automodel.components.optim.scheduler module nemo_automodel.components.optim.utils module nemo_automodel.components.quantization module nemo_automodel.components.quantization.fp8 module nemo_automodel.components.quantization.qat module nemo_automodel.components.quantization.qlora module nemo_automodel.components.training module nemo_automodel.components.training.garbage_collection module nemo_automodel.components.training.model_output_utils module nemo_automodel.components.training.rng module nemo_automodel.components.training.signal_handler module nemo_automodel.components.training.step_scheduler module nemo_automodel.components.training.timers module nemo_automodel.components.training.utils module nemo_automodel.components.utils module nemo_automodel.components.utils.compile_utils module nemo_automodel.components.utils.flops_utils module nemo_automodel.components.utils.model_utils module nemo_automodel.components.utils.yaml_utils module nemo_automodel.shared module nemo_automodel.shared.import_utils module nemo_automodel.shared.torch_patches module nemo_automodel.shared.utils module NeMoAutoDiffusionPipeline (class in nemo_automodel._diffusers.auto_diffusion_pipeline) NeMoAutoModelBiencoder (class in nemo_automodel._transformers.auto_model) NeMoAutoModelForCausalLM (class in nemo_automodel._transformers.auto_model) NeMoAutoModelForImageTextToText (class in nemo_automodel._transformers.auto_model) NeMoAutoModelForMultimodalLM (class in nemo_automodel._transformers.auto_model) NeMoAutoModelForSequenceClassification (class in nemo_automodel._transformers.auto_model) NeMoAutoModelForTextToWaveform (class in nemo_automodel._transformers.auto_model) NeMoAutoTokenizer (class in nemo_automodel._transformers.auto_tokenizer) nemotron_flops() (in module nemo_automodel.components.utils.flops_utils) nemotron_parse_collate_fn() (in module nemo_automodel.components.datasets.vlm.collate_fns) nemotronh_flops() (in module nemo_automodel.components.utils.flops_utils) NemotronHForCausalLM (class in nemo_automodel.components.models.nemotron_v3.model) NemotronHParallelizationStrategy (class in nemo_automodel.components.distributed.parallelizer) NemotronHybridCache (class in nemo_automodel.components.models.nemotron_v3.cache) NemotronParseConfig (class in nemo_automodel.components.models.nemotron_parse.model) NemotronParseDecoder (class in nemo_automodel.components.models.nemotron_parse.model) NemotronParseEncoderConfig (class in nemo_automodel.components.models.nemotron_parse.model) NemotronParseForConditionalGeneration (class in nemo_automodel.components.models.nemotron_parse.model) NemotronParseLoss (class in nemo_automodel.components.models.nemotron_parse.nemotron_parse_loss) NemotronParsePreTrainedModel (class in nemo_automodel.components.models.nemotron_parse.model) NemotronParseTextConfig (class in nemo_automodel.components.models.nemotron_parse.model) NemotronV3Attention (class in nemo_automodel.components.models.nemotron_v3.layers) NemotronV3Block (class in nemo_automodel.components.models.nemotron_v3.layers) NemotronV3Mamba2Mixer (class in nemo_automodel.components.models.nemotron_v3.layers) NemotronV3MambaRMSNormGated (class in nemo_automodel.components.models.nemotron_v3.layers) NemotronV3Model (class in nemo_automodel.components.models.nemotron_v3.model) NemotronV3StateDictAdapter (class in nemo_automodel.components.models.nemotron_v3.state_dict_adapter) neva_projection_flops() (in module nemo_automodel.components.utils.flops_utils) new_is_triton_greater_or_equal() (in module nemo_automodel.components.loss.linear_ce) new_is_triton_greater_or_equal_3_2_0() (in module nemo_automodel.components.loss.linear_ce) no_hf_meta_device() (in module nemo_automodel._transformers.model_init) NON_QUANTIZED_KEY_PATTERNS (in module nemo_automodel.components.models.deepseek_v3.state_dict_adapter) (in module nemo_automodel.components.models.minimax_m2.state_dict_adapter) norm (nemo_automodel.components.models.kimi_k25_vl.model.KimiK25VLLanguageModelBackend property) (nemo_automodel.components.models.kimi_k25_vl.model.KimiK25VLModel property) (nemo_automodel.components.models.kimivl.model.KimiVLLanguageModelBackend property) (nemo_automodel.components.models.kimivl.model.KimiVLModel property) (nemo_automodel.components.models.qwen3_5_moe.model.Qwen3_5MoeModel property) (nemo_automodel.components.models.qwen3_vl_moe.model.Qwen3VLMoeModel property) norm_topk_prob (nemo_automodel.components.moe.config.MoEConfig attribute) normalize() (in module nemo_automodel.components.datasets.llm.megatron.gpt_dataset) np_rng_state (nemo_automodel.components.training.rng.RNGState attribute) null_decorator (in module nemo_automodel.components.moe.megatron.fused_indices_converter) null_decorator() (in module nemo_automodel.shared.import_utils) num_dataset_builder_threads (nemo_automodel.components.datasets.llm.megatron.gpt_dataset.BlendedMegatronDatasetConfig attribute) num_moe_experts (nemo_automodel.components.moe.megatron.token_dispatcher.TokenDispatcherConfig attribute) numel_low_level_dataset() (nemo_automodel.components.datasets.llm.megatron.gpt_dataset.GPTDataset static method)

O

offload_policy (nemo_automodel.components.distributed.config.FSDP2Config attribute) offset (nemo_automodel.components.checkpoint._backports.filesystem._StorageInfo attribute) (nemo_automodel.components.checkpoint._backports.hf_utils._HFStorageInfo attribute) offset_in_file (nemo_automodel.components.checkpoint._backports.consolidate_hf_safetensors._FqnData attribute) optimal_dtype() (nemo_automodel.components.datasets.llm.megatron.indexed_dataset.DType class method) OptimizerParamScheduler (class in nemo_automodel.components.optim.scheduler)

OptimizerState (class in nemo_automodel.components.checkpoint.stateful_wrappers) options (nemo_automodel.components.utils.compile_utils.CompileConfig attribute) original_model_root_dir (nemo_automodel.components.checkpoint.checkpointing.CheckpointingConfig attribute) original_state_dict (nemo_automodel.components.checkpoint._backports.default_planner.DefaultLoadPlanner attribute) overlap_grad_reduce (nemo_automodel.components.distributed.config.MegatronFSDPConfig attribute) overlap_param_gather (nemo_automodel.components.distributed.config.MegatronFSDPConfig attribute)

P

pack_dataset() (in module nemo_automodel.components.datasets.llm.packed_sequence) PACK_TYPE (in module nemo_automodel.components.datasets.llm.packed_sequence) packed_block_causal_mask() (in module nemo_automodel.components.datasets.llm.packed_sequence) packed_sequence_thd_collater() (in module nemo_automodel.components.datasets.utils) pad_within_micro() (in module nemo_automodel.components.datasets.utils) parallel_cross_entropy (in module nemo_automodel.components.loss.te_parallel_ce) PARALLELIZATION_STRATEGIES (in module nemo_automodel.components.distributed.parallelizer) ParallelizationStrategy (class in nemo_automodel.components.distributed.parallelizer) parallelize() (nemo_automodel.components.distributed.ddp.DDPManager method) (nemo_automodel.components.distributed.fsdp2.FSDP2Manager method) (nemo_automodel.components.distributed.megatron_fsdp.MegatronFSDPManager method) (nemo_automodel.components.distributed.parallelizer.DefaultParallelizationStrategy method) (nemo_automodel.components.distributed.parallelizer.NemotronHParallelizationStrategy method) (nemo_automodel.components.distributed.parallelizer.ParallelizationStrategy method) (nemo_automodel.components.distributed.parallelizer.WanParallelizationStrategy method) parallelize_axis_kwargs() (nemo_automodel.components.distributed.mesh.MeshContext method) parallelize_for_pp() (in module nemo_automodel._transformers.infrastructure) PARALLELIZE_FUNCTIONS (in module nemo_automodel.components.distributed.optimized_tp_plans) parallelize_model() (in module nemo_automodel.components.moe.parallelizer) ParallelizeFnProtocol (class in nemo_automodel.components.distributed.pipelining.functional) parse_and_normalize_split() (in module nemo_automodel.components.datasets.llm.megatron.gpt_dataset) parse_args_and_load_config() (in module nemo_automodel.components.config._arg_parser) parse_cli_argv() (in module nemo_automodel.components.config._arg_parser) parse_moe_layers_enum() (in module nemo_automodel.components.models.step3p5.model) parts (nemo_automodel.components.distributed.pipelining.autopipeline.AutoPipeline property) passage_instruction (nemo_automodel.components.datasets.llm.retrieval_dataset.CorpusInfo property) patch() (in module nemo_automodel.autonvtx) patch_causal_lm_model (nemo_automodel.components.distributed.pipelining.config.PipelineConfig attribute) patch_hf_model_for_pp() (in module nemo_automodel.components.distributed.pipelining.hf_utils) patch_inner_model (nemo_automodel.components.distributed.pipelining.config.PipelineConfig attribute) patch_linear_module() (in module nemo_automodel.components._peft.lora) patch_merger() (in module nemo_automodel.components.models.kimivl.model) patch_moe_module() (in module nemo_automodel.components._peft.lora) patch_prepare_fa2_from_position_ids() (in module nemo_automodel.components.utils.compile_utils) patch_stage_backward_maybe_with_nosync (nemo_automodel.components.distributed.pipelining.config.PipelineConfig attribute) patched_backward_maybe_with_nosync() (in module nemo_automodel.components.moe.fsdp_mixin) path (nemo_automodel.components.datasets.llm.retrieval_dataset.CorpusInfo property) path_to_cache (nemo_automodel.components.datasets.llm.megatron.gpt_dataset.BlendedMegatronDatasetConfig attribute) PeftAddon (class in nemo_automodel.components.checkpoint.addons) PeftConfig (class in nemo_automodel.components._peft.lora) permute() (in module nemo_automodel.components.moe.megatron.moe_utils) phi4_mm_collate_fn() (in module nemo_automodel.components.datasets.vlm.collate_fns) pipeline_axis_kwargs() (nemo_automodel.components.distributed.mesh.MeshContext method) pipeline_config (nemo_automodel.components.distributed.mesh.MeshContext attribute)

pipeline_model() (in module nemo_automodel.components.distributed.pipelining.functional) PipelineConfig (class in nemo_automodel.components.distributed.pipelining.config) PipelineInfo (class in nemo_automodel.components.distributed.pipelining.autopipeline) pool() (in module nemo_automodel._transformers.biencoder) position_ids_to_freqs_cis() (in module nemo_automodel.components.models.gpt_oss.rope_utils) post_save() (nemo_automodel.components.checkpoint.addons.CheckpointAddon method) (nemo_automodel.components.checkpoint.addons.ConsolidatedHFAddon method) (nemo_automodel.components.checkpoint.addons.PeftAddon method) postprocess_output_for_attn() (in module nemo_automodel.components.attention.utils) PP (nemo_automodel.components.distributed.mesh.MeshAxisName attribute) pp_batch_size (nemo_automodel.components.distributed.pipelining.config.PipelineConfig attribute) pp_enabled (nemo_automodel.components.distributed.mesh.MeshContext property) pp_microbatch_size (nemo_automodel.components.distributed.pipelining.config.PipelineConfig attribute) pp_schedule (nemo_automodel.components.distributed.pipelining.config.PipelineConfig attribute) pp_schedule_csv (nemo_automodel.components.distributed.pipelining.config.PipelineConfig attribute) pp_seq_len (nemo_automodel.components.distributed.pipelining.config.PipelineConfig attribute) pp_size (nemo_automodel.components.distributed.mesh.MeshContext property) pre_save() (nemo_automodel.components.checkpoint.addons.CheckpointAddon method) (nemo_automodel.components.checkpoint.addons.ConsolidatedHFAddon method) (nemo_automodel.components.checkpoint.addons.PeftAddon method) precompute_float8_dynamic_scale_for_fsdp (nemo_automodel.components.quantization.fp8.FP8Config attribute) precompute_freqs_cis() (in module nemo_automodel.components.models.deepseek_v3.rope_utils) prefix (nemo_automodel.components.checkpoint._backports.filesystem._StoragePrefix attribute) prepare_after_first_microbatch() (in module nemo_automodel.components.training.utils) prepare_decoder_input_ids_from_labels() (nemo_automodel.components.models.nemotron_parse.model.NemotronParseForConditionalGeneration method) prepare_for_final_backward() (in module nemo_automodel.components.training.utils) (nemo_automodel.components.moe.fsdp_mixin.MoEFSDPSyncMixin method) prepare_for_grad_accumulation() (in module nemo_automodel.components.training.utils) (nemo_automodel.components.moe.fsdp_mixin.MoEFSDPSyncMixin method) prepare_global_plan() (nemo_automodel.components.checkpoint._backports.filesystem._FileSystemWriter method) (nemo_automodel.components.checkpoint._backports.filesystem.FileSystemReader method) (nemo_automodel.components.checkpoint._backports.hf_storage._HuggingFaceStorageWriter method) prepare_inputs_for_generation() (nemo_automodel.components.models.nemotron_v3.model.NemotronHForCausalLM method) prepare_local_plan() (nemo_automodel.components.checkpoint._backports.filesystem._FileSystemWriter method) (nemo_automodel.components.checkpoint._backports.filesystem.FileSystemReader method) prepare_qat_model() (in module nemo_automodel.components.quantization.qat) preprocess_args_and_kwargs_for_attn() (in module nemo_automodel.components.attention.utils) preserve_fp32_weights (nemo_automodel.components.distributed.config.MegatronFSDPConfig attribute) pretty_print_stages() (nemo_automodel.components.distributed.pipelining.autopipeline.AutoPipeline method) print_trainable_parameters() (in module nemo_automodel.components.utils.model_utils) process() (nemo_automodel.components.datasets.utils.SFTSingleTurnPreprocessor method) process_group (nemo_automodel.components.checkpoint.checkpointing._AsyncSaveContext attribute) process_input_for_thd() (in module nemo_automodel.components.distributed.thd_utils) process_text_batch() (in module nemo_automodel.components.datasets.vlm.utils)

Q

QATConfig (class in nemo_automodel.components.quantization.qat) quantize_to_int4() (in module nemo_automodel.components.models.kimi_k25_vl.state_dict_adapter) quantizer_type (nemo_automodel.components.quantization.qat.QATConfig attribute) query_instruction (nemo_automodel.components.datasets.llm.retrieval_dataset.CorpusInfo property) Question (nemo_automodel.components.datasets.llm.column_mapped_text_instruction_dataset.ColumnTypes attribute) quick_geglu() (in module nemo_automodel.components.moe.megatron.moe_utils) quick_geglu_back() (in module nemo_automodel.components.moe.megatron.moe_utils) quick_geglu_deepep() (in module nemo_automodel.components.moe.experts) quick_geglu_with_lora() (in module nemo_automodel.components._peft.lora_moe) quick_gelu() (in module nemo_automodel.components.moe.megatron.moe_utils) qwen2_5_collate_fn() (in module nemo_automodel.components.datasets.vlm.collate_fns) Qwen2Attention (class in nemo_automodel.components.models.qwen2.model) Qwen2DecoderLayer (class in nemo_automodel.components.models.qwen2.model) Qwen2ForCausalLM (class in nemo_automodel.components.models.qwen2.model) Qwen2Model (class in nemo_automodel.components.models.qwen2.model) Qwen2PreTrainedModel (class in nemo_automodel.components.models.qwen2.model) Qwen2RotaryEmbedding (in module nemo_automodel.components.models.llama.rope_utils) Qwen2StateDictAdapter (class in nemo_automodel.components.models.qwen2.state_dict_adapter) Qwen3_5MoeBlock (class in nemo_automodel.components.models.qwen3_5_moe.model) Qwen3_5MoeForConditionalGeneration (class in nemo_automodel.components.models.qwen3_5_moe.model) Qwen3_5MoeModel (class in nemo_automodel.components.models.qwen3_5_moe.model)

Qwen3_5MoeStateDictAdapter (class in nemo_automodel.components.models.qwen3_5_moe.state_dict_adapter) Qwen3_5MoeTextModelBackend (class in nemo_automodel.components.models.qwen3_5_moe.model) qwen3_flops() (in module nemo_automodel.components.utils.flops_utils) qwen3_omni_collate_fn() (in module nemo_automodel.components.datasets.vlm.collate_fns) Qwen3MoeAttention (class in nemo_automodel.components.models.qwen3_moe.layers) Qwen3MoeForCausalLM (class in nemo_automodel.components.models.qwen3_moe.model) Qwen3MoeModel (class in nemo_automodel.components.models.qwen3_moe.model) Qwen3MoeStateDictAdapter (class in nemo_automodel.components.models.qwen3_moe.state_dict_adapter) Qwen3NextAttention (class in nemo_automodel.components.models.qwen3_next.layers) Qwen3NextForCausalLM (class in nemo_automodel.components.models.qwen3_next.model) Qwen3NextModel (class in nemo_automodel.components.models.qwen3_next.model) Qwen3NextRMSNorm (class in nemo_automodel.components.models.qwen3_next.layers) Qwen3NextStateDictAdapter (class in nemo_automodel.components.models.qwen3_next.state_dict_adapter) Qwen3OmniMoeStateDictAdapter (class in nemo_automodel.components.models.qwen3_omni_moe.state_dict_adapter) Qwen3OmniMoeThinkerForConditionalGeneration (class in nemo_automodel.components.models.qwen3_omni_moe.model) Qwen3OmniMoeThinkerTextModel (class in nemo_automodel.components.models.qwen3_omni_moe.model) Qwen3VLMoeBlock (class in nemo_automodel.components.models.qwen3_vl_moe.model) Qwen3VLMoeForConditionalGeneration (class in nemo_automodel.components.models.qwen3_vl_moe.model) Qwen3VLMoeModel (class in nemo_automodel.components.models.qwen3_vl_moe.model) Qwen3VLMoeStateDictAdapter (class in nemo_automodel.components.models.qwen3_vl_moe.state_dict_adapter) Qwen3VLMoeTextModelBackend (class in nemo_automodel.components.models.qwen3_vl_moe.model)

R

RadioWithNeck (class in nemo_automodel.components.models.nemotron_parse.model) random_rng_state (nemo_automodel.components.training.rng.RNGState attribute) random_seed (nemo_automodel.components.datasets.llm.megatron.gpt_dataset.BlendedMegatronDatasetConfig attribute) rank (nemo_automodel.components.distributed.init_utils.DistInfo attribute) RankFilter (class in nemo_automodel.components.loggers.log_utils) raw_config (nemo_automodel.components.config.loader.ConfigNode property) read() (nemo_automodel.components.datasets.llm.megatron.indexed_dataset._BinReader method) (nemo_automodel.components.datasets.llm.megatron.indexed_dataset._FileBinReader method) (nemo_automodel.components.datasets.llm.megatron.indexed_dataset._MMapBinReader method) read_data() (nemo_automodel.components.checkpoint._backports.filesystem.FileSystemReader method) (nemo_automodel.components.checkpoint._backports.hf_storage._HuggingFaceStorageReader method) read_metadata() (nemo_automodel.components.checkpoint._backports.filesystem.FileSystemReader method) (nemo_automodel.components.checkpoint._backports.hf_storage._HuggingFaceStorageReader method) recipe (nemo_automodel.components.models.common.utils.TEFp8Config attribute) recipe_name (nemo_automodel.components.quantization.fp8.FP8Config attribute) reduce_loss() (in module nemo_automodel.components.distributed.utils) register() (nemo_automodel._transformers.auto_tokenizer.NeMoAutoTokenizer class method) register_modeling_path() (nemo_automodel._transformers.registry._ModelRegistry method) register_parallel_strategy() (in module nemo_automodel.components.distributed.parallelizer) relative_path (nemo_automodel.components.checkpoint._backports.filesystem._StorageInfo attribute) (nemo_automodel.components.checkpoint._backports.hf_utils._HFStorageInfo attribute) release() (nemo_automodel.components.training.signal_handler.DistributedSignalHandler method) relu2_deepep() (in module nemo_automodel.components.moe.experts) rename() (nemo_automodel.components.checkpoint._backports._fsspec_filesystem.FileSystem method) (nemo_automodel.components.checkpoint._backports.filesystem.FileSystem method) (nemo_automodel.components.checkpoint._backports.filesystem.FileSystemBase method) reorder_cache() (nemo_automodel.components.models.nemotron_v3.cache.NemotronHybridCache method) repeat_kv() (in module nemo_automodel.components.models.mistral3.model) requires_tensor_merging() (in module nemo_automodel.components.checkpoint.conversion_mapping) ReservoirSampler (class in nemo_automodel.components.datasets.reservoir_sampler) reset() (nemo_automodel.components.checkpoint._backports.filesystem._FileSystemWriter method) (nemo_automodel.components.checkpoint._backports.filesystem.FileSystemReader method) (nemo_automodel.components.training.timers.DummyTimer method) (nemo_automodel.components.training.timers.Timer method) (nemo_automodel.components.training.timers.TimerBase method)

reset_attention_mask (nemo_automodel.components.datasets.llm.megatron.gpt_dataset.GPTDatasetConfig attribute) reset_parameters() (nemo_automodel.components.models.qwen3_next.layers.Qwen3NextRMSNorm method) (nemo_automodel.components.models.step3p5.layers.Step3p5RMSNorm method) reset_position_ids (nemo_automodel.components.datasets.llm.megatron.gpt_dataset.GPTDatasetConfig attribute) reshard_after_forward (nemo_automodel.components.moe.config.MoEParallelizerConfig attribute) resolve_data() (nemo_automodel.components.checkpoint._backports.default_planner.DefaultSavePlanner method) resolve_tensor() (nemo_automodel.components.checkpoint._backports.default_planner.DefaultLoadPlanner method) resolve_trust_remote_code() (in module nemo_automodel.components.utils.model_utils) resolve_yaml_env_vars() (in module nemo_automodel.components.config.loader) RetrievalBiencoderCollator (class in nemo_automodel.components.datasets.llm.retrieval_collator) rm_file() (nemo_automodel.components.checkpoint._backports._fsspec_filesystem.FileSystem method) (nemo_automodel.components.checkpoint._backports.filesystem.FileSystem method) (nemo_automodel.components.checkpoint._backports.filesystem.FileSystemBase method) rms_norm (nemo_automodel.components.models.common.utils.BackendConfig attribute) RNGState (class in nemo_automodel.components.training.rng) Rope2DPosEmb (class in nemo_automodel.components.models.kimivl.model) Rope2DPosEmbRepeated (class in nemo_automodel.components.models.kimi_k25_vl.model) rope_fusion (nemo_automodel.components.models.common.utils.BackendConfig attribute) RotaryEmbedding (class in nemo_automodel.components.models.gpt_oss.rope_utils) (in module nemo_automodel.components.models.llama.rope_utils) RotaryEmbedParallel (class in nemo_automodel.components.distributed.optimized_tp_plans) rotate_half() (in module nemo_automodel.components.models.llama.rope_utils) (in module nemo_automodel.components.models.mistral3.model) round_virtual_stages_to_pp_multiple (nemo_automodel.components.distributed.pipelining.config.PipelineConfig attribute) route_scale (nemo_automodel.components.moe.config.MoEConfig attribute) router_bias (nemo_automodel.components.moe.config.MoEConfig attribute) RowwiseParallelLora (class in nemo_automodel.components.distributed.parallel_styles) run() (nemo_automodel.components.training.garbage_collection.GarbageCollection method)

S

SAFE_BASE_DIR (in module nemo_automodel.components.config.loader) safe_import() (in module nemo_automodel.shared.import_utils) safe_import_from() (in module nemo_automodel.shared.import_utils) safe_import_te() (in module nemo_automodel.shared.import_utils) safe_yaml_representers() (in module nemo_automodel.components.utils.yaml_utils) SAFETENSORS (nemo_automodel.components.checkpoint._backports.filesystem.SerializationFormat attribute) save_config() (in module nemo_automodel.components.checkpoint.checkpointing) save_consolidated (nemo_automodel.components.checkpoint.checkpointing.CheckpointingConfig attribute) save_model() (nemo_automodel.components.checkpoint.checkpointing.Checkpointer method) save_on_dp_ranks() (nemo_automodel.components.checkpoint.checkpointing.Checkpointer method) save_optimizer() (nemo_automodel.components.checkpoint.checkpointing.Checkpointer method) save_pretrained() (nemo_automodel._transformers.biencoder.BiencoderModel method) (nemo_automodel.components.models.common.hf_checkpointing_mixin.HFCheckpointingMixin method) SAVED_OFFSETS_KEY (in module nemo_automodel.components.checkpoint._backports.hf_utils) scale_grads_and_clip_grad_norm() (in module nemo_automodel.components.training.utils) scale_grads_by_divisor() (in module nemo_automodel.components.distributed.pipelining.functional) scale_grads_in_schedule (nemo_automodel.components.distributed.pipelining.config.PipelineConfig attribute) schedule (nemo_automodel.components.distributed.pipelining.autopipeline.PipelineInfo attribute) ScopedModuleOffloading (class in nemo_automodel.components.training.utils) ScopedRNG (class in nemo_automodel.components.training.rng) score_func (nemo_automodel.components.moe.config.MoEConfig attribute) SENSITIVE_KEY_SUBSTRINGS (in module nemo_automodel.components.config.loader) sequence_length (nemo_automodel.components.datasets.llm.megatron.gpt_dataset.BlendedMegatronDatasetConfig attribute) sequence_lengths (nemo_automodel.components.datasets.llm.megatron.indexed_dataset.IndexedDataset property) sequence_parallel (nemo_automodel.components.distributed.config.FSDP2Config attribute) (nemo_automodel.components.distributed.config.MegatronFSDPConfig attribute) SequenceParallelAllGatherActivation (class in nemo_automodel.components.distributed.optimized_tp_plans) SequenceParallelLora (class in nemo_automodel.components.distributed.parallel_styles) SerializationFormat (class in nemo_automodel.components.checkpoint._backports.filesystem) set_barrier_group() (nemo_automodel.components.training.timers.Timer method) set_by_dotted() (nemo_automodel.components.config.loader.ConfigNode method) set_decoder() (nemo_automodel.components.models.llama.model.LlamaForCausalLM method) set_enable_user_modules() (in module nemo_automodel.components.config.loader) set_epoch() (nemo_automodel.components.datasets.llm.column_mapped_text_instruction_iterable_dataset.ColumnMappedTextInstructionIterableDataset method) (nemo_automodel.components.datasets.llm.delta_lake_dataset._LimitedDeltaLakeDataset method) (nemo_automodel.components.datasets.llm.delta_lake_dataset.DeltaLakeDataset method) (nemo_automodel.components.training.step_scheduler.StepScheduler method) set_input_embeddings() (nemo_automodel.components.models.deepseek_v3.model.DeepseekV3ForCausalLM method) (nemo_automodel.components.models.deepseek_v32.model.DeepseekV32ForCausalLM method) (nemo_automodel.components.models.glm4_moe.model.Glm4MoeForCausalLM method) (nemo_automodel.components.models.glm4_moe_lite.model.Glm4MoeLiteForCausalLM method) (nemo_automodel.components.models.gpt_oss.model.GptOssForCausalLM method) (nemo_automodel.components.models.kimi_k25_vl.model.KimiK25VLForConditionalGeneration method) (nemo_automodel.components.models.kimi_k25_vl.model.KimiK25VLLanguageModelBackend method) (nemo_automodel.components.models.kimivl.model.KimiVLForConditionalGeneration method) (nemo_automodel.components.models.kimivl.model.KimiVLLanguageModelBackend method) (nemo_automodel.components.models.llama.model.LlamaForCausalLM method) (nemo_automodel.components.models.minimax_m2.model.MiniMaxM2ForCausalLM method) (nemo_automodel.components.models.mistral3.model.Ministral3ForCausalLM method) (nemo_automodel.components.models.nemotron_parse.model.NemotronParseDecoder method) (nemo_automodel.components.models.nemotron_parse.model.NemotronParseForConditionalGeneration method) (nemo_automodel.components.models.nemotron_v3.model.NemotronHForCausalLM method) (nemo_automodel.components.models.qwen2.model.Qwen2ForCausalLM method) (nemo_automodel.components.models.qwen3_5_moe.model.Qwen3_5MoeTextModelBackend method) (nemo_automodel.components.models.qwen3_moe.model.Qwen3MoeForCausalLM method) (nemo_automodel.components.models.qwen3_next.model.Qwen3NextForCausalLM method) (nemo_automodel.components.models.qwen3_omni_moe.model.Qwen3OmniMoeThinkerForConditionalGeneration method) (nemo_automodel.components.models.qwen3_vl_moe.model.Qwen3VLMoeForConditionalGeneration method) (nemo_automodel.components.models.qwen3_vl_moe.model.Qwen3VLMoeTextModelBackend method) (nemo_automodel.components.models.step3p5.model.Step3p5ForCausalLM method) set_is_first_microbatch() (in module nemo_automodel.components.models.common.utils) set_is_optim_step() (in module nemo_automodel.components.models.common.utils) set_moe_mesh() (nemo_automodel.components.moe.experts.GroupedExpertsTE method) set_output_embeddings() (nemo_automodel.components.models.deepseek_v3.model.DeepseekV3ForCausalLM method) (nemo_automodel.components.models.deepseek_v32.model.DeepseekV32ForCausalLM method) (nemo_automodel.components.models.glm4_moe.model.Glm4MoeForCausalLM method) (nemo_automodel.components.models.glm4_moe_lite.model.Glm4MoeLiteForCausalLM method) (nemo_automodel.components.models.gpt_oss.model.GptOssForCausalLM method) (nemo_automodel.components.models.kimi_k25_vl.model.KimiK25VLForConditionalGeneration method) (nemo_automodel.components.models.kimivl.model.KimiVLForConditionalGeneration method) (nemo_automodel.components.models.llama.model.LlamaForCausalLM method) (nemo_automodel.components.models.minimax_m2.model.MiniMaxM2ForCausalLM method) (nemo_automodel.components.models.mistral3.model.Ministral3ForCausalLM method) (nemo_automodel.components.models.nemotron_parse.model.NemotronParseForConditionalGeneration method) (nemo_automodel.components.models.nemotron_v3.model.NemotronHForCausalLM method) (nemo_automodel.components.models.qwen2.model.Qwen2ForCausalLM method) (nemo_automodel.components.models.qwen3_moe.model.Qwen3MoeForCausalLM method) (nemo_automodel.components.models.qwen3_next.model.Qwen3NextForCausalLM method) (nemo_automodel.components.models.qwen3_omni_moe.model.Qwen3OmniMoeThinkerForConditionalGeneration method) (nemo_automodel.components.models.qwen3_vl_moe.model.Qwen3VLMoeForConditionalGeneration method) (nemo_automodel.components.models.step3p5.model.Step3p5ForCausalLM method) set_up_planner() (nemo_automodel.components.checkpoint._backports.default_planner._EmptyStateDictLoadPlanner method) (nemo_automodel.components.checkpoint._backports.default_planner.DefaultLoadPlanner method) (nemo_automodel.components.checkpoint._backports.default_planner.DefaultSavePlanner method) set_up_storage_reader() (nemo_automodel.components.checkpoint._backports.filesystem.FileSystemReader method) set_up_storage_writer() (nemo_automodel.components.checkpoint._backports.filesystem._FileSystemWriter method) setup_context() (nemo_automodel.components._peft.lora.LoRATritonFunction static method)

setup_logging() (in module nemo_automodel.components.loggers.log_utils) setup_metadata() (nemo_automodel.components.moe.megatron.token_dispatcher._DeepepManager method) (nemo_automodel.components.moe.megatron.token_dispatcher._DispatchManager method) setup_qkv_projection() (nemo_automodel.components.models.common.combined_projection.combined_qkv.CombinedQKVAttentionMixin method) SFTSingleTurnPreprocessor (class in nemo_automodel.components.datasets.utils) shape (nemo_automodel.components.checkpoint._backports.hf_utils._HFStorageInfo attribute) shape_in_file (nemo_automodel.components.checkpoint._backports.consolidate_hf_safetensors._FqnData attribute) SHAPE_KEY (in module nemo_automodel.components.checkpoint._backports.hf_utils) shard() (nemo_automodel.components.datasets.llm.column_mapped_text_instruction_iterable_dataset.ColumnMappedTextInstructionIterableDataset method) (nemo_automodel.components.datasets.llm.delta_lake_dataset._LimitedDeltaLakeDataset method) (nemo_automodel.components.datasets.llm.delta_lake_dataset.DeltaLakeDataset method) (nemo_automodel.components.datasets.llm.delta_lake_dataset.DeltaLakeIterator method) SHARDED_FILE_NAME (in module nemo_automodel.components.checkpoint._backports.hf_utils) shared_comm_manager (nemo_automodel.components.moe.megatron.token_dispatcher.MoEFlexTokenDispatcher attribute) shared_expert_activation (nemo_automodel.components.moe.config.MoEConfig attribute) shared_expert_gate (nemo_automodel.components.moe.config.MoEConfig attribute) shared_expert_inter_dim (nemo_automodel.components.moe.config.MoEConfig attribute) SHARING_DEEPEP_MANAGER (in module nemo_automodel.components.moe.megatron.token_dispatcher) should_load_expert_for_rank() (in module nemo_automodel.components.moe.state_dict_utils) should_quantize_key() (in module nemo_automodel.components.models.deepseek_v3.state_dict_adapter) (in module nemo_automodel.components.models.minimax_m2.state_dict_adapter) shuffle() (nemo_automodel.components.datasets.llm.column_mapped_text_instruction_iterable_dataset.ColumnMappedTextInstructionIterableDataset method) (nemo_automodel.components.datasets.llm.delta_lake_dataset._LimitedDeltaLakeDataset method) (nemo_automodel.components.datasets.llm.delta_lake_dataset.DeltaLakeDataset method) signals_received() (nemo_automodel.components.training.signal_handler.DistributedSignalHandler method) sigterm_received (nemo_automodel.components.training.step_scheduler.StepScheduler property) single_rank_consolidation (nemo_automodel.components.checkpoint.checkpointing.CheckpointingConfig attribute) size() (nemo_automodel.components.datasets.llm.megatron.indexed_dataset.DType class method) skip_random_init() (in module nemo_automodel.components.utils.model_utils) skip_task_head_prefixes_for_base_model (nemo_automodel.components.checkpoint.checkpointing.CheckpointingConfig attribute) sliding_window_overwrite() (in module nemo_automodel._transformers.utils) softmax_before_topk (nemo_automodel.components.moe.config.MoEConfig attribute) Split (class in nemo_automodel.components.datasets.llm.megatron.gpt_dataset) split (nemo_automodel.components.datasets.llm.megatron.gpt_dataset.BlendedMegatronDatasetConfig attribute) split_batch_into_thd_chunks() (in module nemo_automodel.components.distributed.thd_utils) split_experts_weights_dtensor_aware() (in module nemo_automodel.components.moe.state_dict_utils) split_matrix (nemo_automodel.components.datasets.llm.megatron.gpt_dataset.BlendedMegatronDatasetConfig attribute) split_model_into_stages() (in module nemo_automodel.components.distributed.pipelining.functional) squeeze_input_for_thd() (in module nemo_automodel.components.utils.model_utils) stack_and_move_tensor_metrics_to_cpu() (in module nemo_automodel.components.loggers.metric_logger) stage() (nemo_automodel.components.checkpoint._backports.filesystem.FileSystemWriter method) stage_ids_this_rank() (in module nemo_automodel.components.distributed.pipelining.functional) stager (nemo_automodel.components.checkpoint.checkpointing._AsyncSaveContext attribute) stages (nemo_automodel.components.distributed.pipelining.autopipeline.PipelineInfo attribute) staging_active (nemo_automodel.components.checkpoint.checkpointing._AsyncSaveContext attribute) staging_dir (nemo_automodel.components.checkpoint.checkpointing.CheckpointingConfig attribute) start() (nemo_automodel.components.training.timers.DummyTimer method) (nemo_automodel.components.training.timers.Timer method) (nemo_automodel.components.training.timers.TimerBase method) start_loading() (nemo_automodel.components.checkpoint._backports.filesystem._OverlappingCpuLoader method) (nemo_automodel.components.checkpoint._backports.filesystem._SerialCpuLoader method) (nemo_automodel.components.checkpoint._backports.filesystem._TensorLoader method) state_dict() (nemo_automodel.components.checkpoint.stateful_wrappers.ModelState method) (nemo_automodel.components.checkpoint.stateful_wrappers.OptimizerState method) (nemo_automodel.components.moe.experts.GroupedExpertsTE method) (nemo_automodel.components.optim.scheduler.OptimizerParamScheduler method) (nemo_automodel.components.training.rng.StatefulRNG method) (nemo_automodel.components.training.step_scheduler.StepScheduler method) StateDictAdapter (class in nemo_automodel.components.checkpoint.state_dict_adapter) StatefulRNG (class in nemo_automodel.components.training.rng) step (nemo_automodel.components.loggers.metric_logger.MetricsSample attribute) step() (nemo_automodel.components.optim.scheduler.OptimizerParamScheduler method) Step3p5Attention (class in nemo_automodel.components.models.step3p5.layers) Step3p5ForCausalLM (class in nemo_automodel.components.models.step3p5.model) Step3p5MLP (class in nemo_automodel.components.models.step3p5.layers) Step3p5Model (class in nemo_automodel.components.models.step3p5.model) Step3p5RMSNorm (class in nemo_automodel.components.models.step3p5.layers) Step3p5RotaryEmbedding (class in nemo_automodel.components.models.step3p5.layers) Step3p5StateDictAdapter (class in nemo_automodel.components.models.step3p5.state_dict_adapter) StepScheduler (class in nemo_automodel.components.training.step_scheduler) stop() (nemo_automodel.components.training.timers.DummyTimer method) (nemo_automodel.components.training.timers.Timer method) (nemo_automodel.components.training.timers.TimerBase method) storage_meta() (nemo_automodel.components.checkpoint._backports.filesystem._FileSystemWriter method) strategy_config (nemo_automodel.components.distributed.mesh.MeshContext attribute) STRATEGY_MAP (in module nemo_automodel.components.distributed.mesh) SUFFIX (in module nemo_automodel.components.checkpoint._backports.hf_utils) SUPPORTED_BACKBONES (in module nemo_automodel._transformers.biencoder) supported_models (nemo_automodel._transformers.registry._ModelRegistry property) supports_gradient_checkpointing (nemo_automodel.components.models.kimi_k25_vl.model.KimiK25VLForConditionalGeneration attribute) (nemo_automodel.components.models.llama.model.LlamaPreTrainedModel attribute) (nemo_automodel.components.models.mistral3.model.Ministral3PreTrainedModel attribute) (nemo_automodel.components.models.nemotron_parse.model.NemotronParsePreTrainedModel attribute) (nemo_automodel.components.models.qwen2.model.Qwen2PreTrainedModel attribute) suppress_wandb_log_messages() (in module nemo_automodel.components.loggers.wandb_utils) swiglu() (in module nemo_automodel.components.moe.megatron.moe_utils) swiglu_back() (in module nemo_automodel.components.moe.megatron.moe_utils) swiglu_with_lora() (in module nemo_automodel.components._peft.lora_moe)

T

take() (nemo_automodel.components.datasets.llm.delta_lake_dataset._LimitedDeltaLakeDataset method) (nemo_automodel.components.datasets.llm.delta_lake_dataset.DeltaLakeDataset method) target_modules (nemo_automodel.components._peft.lora.PeftConfig attribute) (nemo_automodel.components._peft.module_matcher.ModuleMatcher attribute) task_type (nemo_automodel.components.datasets.llm.retrieval_dataset.CorpusInfo property) te_fp8 (nemo_automodel.components.models.common.utils.BackendConfig attribute) TEFp8Config (class in nemo_automodel.components.models.common.utils) TEParallelCrossEntropy (class in nemo_automodel.components.loss.te_parallel_ce) test (nemo_automodel.components.datasets.llm.megatron.gpt_dataset.Split attribute) TEXT_MODULE_ATTRS (in module nemo_automodel.components.distributed.pipelining.hf_utils) TextQADataset (class in nemo_automodel.components.datasets.llm.retrieval_dataset) Timer (class in nemo_automodel.components.training.timers) TimerBase (class in nemo_automodel.components.training.timers) Timers (class in nemo_automodel.components.training.timers) timestamp (nemo_automodel.components.loggers.metric_logger.MetricsSample attribute) to_cpu() (in module nemo_automodel.components.distributed.tensor_utils) to_dict() (nemo_automodel.components._peft.lora.PeftConfig method) (nemo_automodel.components.config.loader.ConfigNode method) (nemo_automodel.components.distributed.config.DDPConfig method) (nemo_automodel.components.distributed.config.FSDP2Config method) (nemo_automodel.components.distributed.config.MegatronFSDPConfig method) (nemo_automodel.components.distributed.pipelining.config.PipelineConfig method) (nemo_automodel.components.loggers.metric_logger.MetricsSample method) (nemo_automodel.components.models.kimi_k25_vl.model.KimiK25VLConfig method) (nemo_automodel.components.models.kimivl.model.KimiVLConfig method) (nemo_automodel.components.models.nemotron_parse.model.NemotronParseConfig method) (nemo_automodel.components.moe.config.MoEParallelizerConfig method) (nemo_automodel.components.quantization.fp8.FP8Config method) (nemo_automodel.components.quantization.qat.QATConfig method) (nemo_automodel.components.utils.compile_utils.CompileConfig method) to_empty_parameters_only() (in module nemo_automodel.components.checkpoint.checkpointing) to_hf() (nemo_automodel.components.checkpoint.state_dict_adapter.StateDictAdapter method) (nemo_automodel.components.models.common.bidirectional.BiencoderStateDictAdapter method) (nemo_automodel.components.models.common.combined_projection.state_dict_adapter.CombinedProjectionStateDictAdapter method) (nemo_automodel.components.models.deepseek_v3.state_dict_adapter.DeepSeekV3StateDictAdapter method) (nemo_automodel.components.models.glm4_moe.state_dict_adapter.Glm4MoeStateDictAdapter method) (nemo_automodel.components.models.gpt_oss.state_dict_adapter.GPTOSSStateDictAdapter method) (nemo_automodel.components.models.kimi_k25_vl.state_dict_adapter.KimiK25VLStateDictAdapter method) (nemo_automodel.components.models.kimivl.model.KimiVLStateDictAdapter method) (nemo_automodel.components.models.minimax_m2.state_dict_adapter.MiniMaxM2StateDictAdapter method) (nemo_automodel.components.models.nemotron_v3.state_dict_adapter.NemotronV3StateDictAdapter method) (nemo_automodel.components.models.qwen3_5_moe.state_dict_adapter.Qwen3_5MoeStateDictAdapter method) (nemo_automodel.components.models.qwen3_moe.state_dict_adapter.Qwen3MoeStateDictAdapter method) (nemo_automodel.components.models.qwen3_next.state_dict_adapter.Qwen3NextStateDictAdapter method) (nemo_automodel.components.models.qwen3_omni_moe.state_dict_adapter.Qwen3OmniMoeStateDictAdapter method) (nemo_automodel.components.models.qwen3_vl_moe.state_dict_adapter.Qwen3VLMoeStateDictAdapter method) (nemo_automodel.components.models.step3p5.state_dict_adapter.Step3p5StateDictAdapter method)

to_local_if_dtensor() (in module nemo_automodel.components.distributed.tensor_utils) to_yaml_dict() (nemo_automodel.components.config.loader.ConfigNode method) token_permutation() (nemo_automodel.components.moe.megatron.token_dispatcher.MoEFlexTokenDispatcher method) token_permutation2() (nemo_automodel.components.moe.megatron.token_dispatcher.MoEFlexTokenDispatcher method) token_unpermutation() (nemo_automodel.components.moe.megatron.token_dispatcher.MoEFlexTokenDispatcher method) TokenDispatcherConfig (class in nemo_automodel.components.moe.megatron.token_dispatcher) tokenizer (nemo_automodel.components.datasets.llm.megatron.gpt_dataset.BlendedMegatronDatasetConfig attribute) top_k_experts (nemo_automodel.components.moe.config.MoEMetricsConfig attribute) torch_rng_state (nemo_automodel.components.training.rng.RNGState attribute) TORCH_SAVE (nemo_automodel.components.checkpoint._backports.filesystem.SerializationFormat attribute) TP (nemo_automodel.components.distributed.mesh.MeshAxisName attribute) tp_plan (nemo_automodel.components.distributed.config.FSDP2Config attribute) (nemo_automodel.components.distributed.config.MegatronFSDPConfig attribute) tp_size (nemo_automodel.components.distributed.mesh.MeshContext property) tpool_patch_merger() (in module nemo_automodel.components.models.kimi_k25_vl.model) train (nemo_automodel.components.datasets.llm.megatron.gpt_dataset.Split attribute) train_gate (nemo_automodel.components.moe.config.MoEConfig attribute) transform_descriptors (nemo_automodel.components.checkpoint._backports.filesystem._StorageInfo attribute) transform_load_stream() (nemo_automodel.components.checkpoint._backports.filesystem._StorageReaderTransforms method) transform_object() (nemo_automodel.components.checkpoint._backports.default_planner.DefaultSavePlanner method) transform_save_stream() (nemo_automodel.components.checkpoint._backports.filesystem._StorageWriterTransforms method) transform_tensor() (nemo_automodel.components.checkpoint._backports.default_planner.DefaultLoadPlanner method) transformer_flops() (in module nemo_automodel.components.utils.flops_utils) TransformerBlock (class in nemo_automodel.components.models.gpt2) translate_to_lora() (in module nemo_automodel.components.distributed.parallel_styles) translate_to_torch_parallel_style() (in module nemo_automodel.components.distributed.parallelizer) translate_value() (in module nemo_automodel.components.config.loader) TritonLinearLoRA (class in nemo_automodel.components._peft.lora) try_load_blend_from_json() (in module nemo_automodel.components.datasets.llm.megatron_dataset)

U

uint16 (nemo_automodel.components.datasets.llm.megatron.indexed_dataset.DType attribute) uint8 (nemo_automodel.components.datasets.llm.megatron.indexed_dataset.DType attribute) UnavailableError UnavailableMeta (class in nemo_automodel.shared.import_utils) UniformSubtreeItem (in module nemo_automodel.components.distributed.parallelizer_utils) unpermute() (in module nemo_automodel.components.moe.megatron.moe_utils) unshard_fsdp2_model() (in module nemo_automodel.components.distributed.parallelizer)

update() (nemo_automodel.components.models.nemotron_v3.cache.NemotronHybridCache method) update_bias() (nemo_automodel.components.moe.layers.FakeBalancedGate method) (nemo_automodel.components.moe.layers.Gate method) update_conv_state() (nemo_automodel.components.models.nemotron_v3.cache.NemotronHybridCache method) update_moe_gate_bias() (nemo_automodel.components.models.deepseek_v3.model.DeepseekV3ForCausalLM method) (nemo_automodel.components.models.deepseek_v3.model.DeepseekV3Model method) use_dora (nemo_automodel.components._peft.lora.PeftConfig attribute) use_triton (nemo_automodel.components._peft.lora.PeftConfig attribute)

V

valid (nemo_automodel.components.datasets.llm.megatron.gpt_dataset.Split attribute) validate_checkpoint_id() (nemo_automodel.components.checkpoint._backports._fsspec_filesystem.FileSystem class method) (nemo_automodel.components.checkpoint._backports._fsspec_filesystem.FsspecReader class method) (nemo_automodel.components.checkpoint._backports._fsspec_filesystem.FsspecWriter class method) (nemo_automodel.components.checkpoint._backports.filesystem._FileSystemWriter class method) (nemo_automodel.components.checkpoint._backports.filesystem.FileSystem class method) (nemo_automodel.components.checkpoint._backports.filesystem.FileSystemBase class method) (nemo_automodel.components.checkpoint._backports.filesystem.FileSystemReader class method) validate_dataset_asset_accessibility() (in module nemo_automodel.components.datasets.llm.megatron_dataset) validate_dtensor_expert_sharding() (in module nemo_automodel.components.moe.state_dict_utils) validate_hf_model_for_pipeline_support() (in module nemo_automodel.components.distributed.pipelining.hf_utils) validate_tp_mesh() (in module nemo_automodel.components.distributed.parallelizer) validate_tp_mesh_for_nemotron_nas() (in module nemo_automodel.components.distributed.parallelizer)

values() (nemo_automodel.components.checkpoint._backports.filesystem._OverlappingCpuLoader method) (nemo_automodel.components.checkpoint._backports.filesystem._SerialCpuLoader method) (nemo_automodel.components.checkpoint._backports.filesystem._TensorLoader method) verify_fp8_conversion() (in module nemo_automodel.components.quantization.fp8) verify_onnx() (in module nemo_automodel.components.models.llama_bidirectional.export_onnx) verify_qlora_quantization() (in module nemo_automodel.components.quantization.qlora) VERSION (in module nemo_automodel.components.datasets.llm.nanogpt_dataset) vision_attention_flash() (in module nemo_automodel.components.models.kimi_k25_vl.model) (in module nemo_automodel.components.models.kimivl.model) vision_attention_sdpa() (in module nemo_automodel.components.models.kimi_k25_vl.model) (in module nemo_automodel.components.models.kimivl.model) visualize_current_schedule() (nemo_automodel.components.distributed.pipelining.autopipeline.AutoPipeline method) VocabParallelEmbedding (class in nemo_automodel.components.distributed.optimized_tp_plans)

W

WanParallelizationStrategy (class in nemo_automodel.components.distributed.parallelizer) warning_filter() (in module nemo_automodel.components.loggers.log_utils) weighted_bias_quick_geglu() (in module nemo_automodel.components.moe.megatron.moe_utils) weighted_bias_quick_geglu_back() (in module nemo_automodel.components.moe.megatron.moe_utils) weighted_bias_quick_geglu_impl() (in module nemo_automodel.components.moe.megatron.moe_utils) weighted_bias_swiglu_impl() (in module nemo_automodel.components.moe.megatron.moe_utils) weighted_quick_geglu() (in module nemo_automodel.components.moe.megatron.moe_utils) weighted_quick_geglu_back() (in module nemo_automodel.components.moe.megatron.moe_utils) weighted_swiglu() (in module nemo_automodel.components.moe.megatron.moe_utils) weighted_swiglu_back() (in module nemo_automodel.components.moe.megatron.moe_utils) WeightedBiasQuickGeGLUFunction (class in nemo_automodel.components.moe.megatron.moe_utils)

WeightedQuickGeGLUFunction (class in nemo_automodel.components.moe.megatron.moe_utils) WeightedSwiGLUFunction (class in nemo_automodel.components.moe.megatron.moe_utils) wildcard_match() (in module nemo_automodel.components._peft.module_matcher) with_barrier() (nemo_automodel.components.training.timers.TimerBase method) world_size (nemo_automodel.components.distributed.init_utils.DistInfo attribute) wrap_outer_model (nemo_automodel.components.moe.config.MoEParallelizerConfig attribute) write() (nemo_automodel.components.datasets.llm.megatron.indexed_dataset._IndexWriter method) (nemo_automodel.components.training.timers.Timers method) write_data() (nemo_automodel.components.checkpoint._backports.filesystem._FileSystemWriter method) (nemo_automodel.components.checkpoint._backports.hf_storage._HuggingFaceStorageWriter method) write_to_wandb() (nemo_automodel.components.training.timers.Timers method)

Y

yarn_get_mscale() (in module nemo_automodel.components.models.deepseek_v3.rope_utils)

Z

zero_dp_strategy (nemo_automodel.components.distributed.config.MegatronFSDPConfig attribute)