> For clean Markdown of any page, append .md to the page URL. > For a complete documentation index, see https://docs.nvidia.com/nemo/gym/main/llms.txt. # Main ## Docs - [NeMo Gym](https://docs.nvidia.com/nemo/gym/main/about.md): NeMo Gym is a library for evaluating and improving models and agents using environments. - [Concepts](https://docs.nvidia.com/nemo/gym/main/about/concepts.md): Understand what environments are, how they power evaluation, agent optimization, and training, and how the same components serve all three. - [Environments](https://docs.nvidia.com/nemo/gym/main/about/concepts/environments.md): Environments decompose into dataset, agent harness, verifier, and state — unifying evaluation, agent optimization, and training. - [Evaluation](https://docs.nvidia.com/nemo/gym/main/about/concepts/evaluation.md): What evaluation is, how to evaluate agents and models, and what makes a good eval. - [Training](https://docs.nvidia.com/nemo/gym/main/about/concepts/training.md): Post-training techniques and how environments are used for training. - [Key Terminology](https://docs.nvidia.com/nemo/gym/main/about/concepts/key-terminology.md): Essential vocabulary for agent evaluation, policy-model training, RL workflows, and NeMo Gym. - [Architecture](https://docs.nvidia.com/nemo/gym/main/about/architecture.md): Understand how NeMo Gym implements environments as composable server components. - [Ecosystem](https://docs.nvidia.com/nemo/gym/main/about/ecosystem.md): How NeMo Gym fits within the NVIDIA NeMo platform and the broader agentic ecosystem. - [Release Notes](https://docs.nvidia.com/nemo/gym/main/about/release-notes.md): - [Prerequisites](https://docs.nvidia.com/nemo/gym/main/get-started/prerequisites.md): What you need to install and run NeMo Gym. - [Installation](https://docs.nvidia.com/nemo/gym/main/get-started/installation.md): Install NeMo Gym and verify your setup. - [Quickstart](https://docs.nvidia.com/nemo/gym/main/get-started/quickstart.md): Run your first agent evaluation with NeMo Gym. - [Prepare Tasks](https://docs.nvidia.com/nemo/gym/main/data.md): Define the tasks NeMo Gym runs, as JSONL rows served by a Resources Server or as Harbor task folders. - [JSONL Datasets](https://docs.nvidia.com/nemo/gym/main/data/jsonl.md): Define tasks as JSONL rows served by a Resources Server: the row format, dataset configuration, validation and CLI commands. - [Prepare and Validate](https://docs.nvidia.com/nemo/gym/main/data/jsonl/prepare-validate.md): - [Download from Hugging Face](https://docs.nvidia.com/nemo/gym/main/data/jsonl/download-huggingface.md): - [Prompt Config](https://docs.nvidia.com/nemo/gym/main/data/jsonl/prompt-config.md): - [Harbor Task Folders](https://docs.nvidia.com/nemo/gym/main/data/task-folders.md): Write, validate and run tasks as task folders in the Harbor format, on the harbor Environment Server. - [Author Task Folders](https://docs.nvidia.com/nemo/gym/main/data/task-folders/harbor-tasks.md): Write a task folder, check it with the oracle, run an agent on it, and pull or publish Harbor datasets. - [Task Folder Format](https://docs.nvidia.com/nemo/gym/main/data/task-folders/task-folder-format.md): The files of a task folder and a dataset of task folders, the task.toml and dataset.toml fields Gym reads, Dockerfile modes, Harbor compatibility and what the loader refuses. - [Task Folder Verifiers](https://docs.nvidia.com/nemo/gym/main/data/task-folders/task-folder-verifiers.md): The contract a task folder's tests/ must meet: reward files, Python verifiers, reference answers, sidecar inputs and shared or separate verification. - [Multi-Step Task Folders](https://docs.nvidia.com/nemo/gym/main/data/task-folders/multi-step-tasks.md): Write a task folder with [[steps]] that runs as several turns in one sandbox, each with its own instruction and verifier, rolled up into one reward. - [Multi-Container Tasks](https://docs.nvidia.com/nemo/gym/main/data/task-folders/compose-tasks.md): Author a task folder whose environment runs sidecar containers beside the agent's sandbox, described with Docker Compose. - [Task Folder Patterns](https://docs.nvidia.com/nemo/gym/main/data/task-folders/task-folder-patterns.md): Generate many tasks from one template, type each task's fields, give a task its own MCP tool, and share verifier code across tasks. - [Configure Agents](https://docs.nvidia.com/nemo/gym/main/agent-server.md): - [Integrate Existing Agents](https://docs.nvidia.com/nemo/gym/main/agent-server/integrate-existing-agents.md): Choose an agent that Gym already integrates, pair it with a benchmark, and see what each one supports. - [Drive a Remote Agent](https://docs.nvidia.com/nemo/gym/main/agent-server/remote-agent.md): Evaluate an agent service you host yourself — a composition of OpenAI Responses-compliant agents, with Gym driving the loop - [Agent Skills](https://docs.nvidia.com/nemo/gym/main/agent-server/agent-skills.md): Evaluate agent skills as a run-level variable, decoupled from the dataset - [Add or Update a Harness](https://docs.nvidia.com/nemo/gym/main/agent-server/add-or-update-a-harness.md): Integrate harnesses with explicit host and sandbox execution paths, reusable lifecycle helpers, and validated upgrades. - [Run NOOA SWE-bench Pro with restricted networking](https://docs.nvidia.com/nemo/gym/main/agent-server/nooa-swe-networking.md): Prepare NOOA offline and restrict only the SWE-bench Pro task sandbox on OpenSandbox. - [NOOA history replay and training metadata](https://docs.nvidia.com/nemo/gym/main/agent-server/nooa-history-replay.md): Understand exact Gym replay, portable history, and the measured cost of durable training metadata. - [Design How Agents Interact with Environments](https://docs.nvidia.com/nemo/gym/main/agent-interactions.md): Choose the Environment Server that orders each episode: how a task is set up, how the agent is driven, and how the attempt is scored. - [Harbor Environment Server](https://docs.nvidia.com/nemo/gym/main/agent-interactions/harbor.md): Run task folders with the harbor Environment Server, which seeds a sandbox from each folder, runs the agent in it and scores the attempt with the task's own verifier. - [Configure Models](https://docs.nvidia.com/nemo/gym/main/model-server.md): - [OpenAI](https://docs.nvidia.com/nemo/gym/main/model-server/openai.md): Connect to OpenAI or another Responses endpoint without translating between API formats - [Azure OpenAI](https://docs.nvidia.com/nemo/gym/main/model-server/azure-openai.md): Use models hosted on Azure OpenAI deployments - [Inference Providers](https://docs.nvidia.com/nemo/gym/main/model-server/inference-providers.md): Use hosted inference providers like Fireworks, Together.ai, OpenRouter, and more for eval workloads - [Model-Call Capture](https://docs.nvidia.com/nemo/gym/main/model-server/model-call-capture.md): Optional per-rollout model-call evidence for evaluation and debugging - [vLLM](https://docs.nvidia.com/nemo/gym/main/model-server/vllm.md): Wrapper for an existing, external vLLM server - [Local vLLM](https://docs.nvidia.com/nemo/gym/main/model-server/local-vllm.md): Gym-managed vLLM server deployment - [Local vLLM Proxy](https://docs.nvidia.com/nemo/gym/main/model-server/local-vllm-proxy.md): Expose one Local vLLM deployment as multiple model servers - [Anthropic Messages](https://docs.nvidia.com/nemo/gym/main/model-server/anthropic-messages.md): Every Gym model server exposes POST /v1/messages so Anthropic-native harnesses like Claude Code can run against any backend - [Switchyard](https://docs.nvidia.com/nemo/gym/main/model-server/switchyard.md): Run any Gym benchmark against a Switchyard model router and compare routing strategies against a fixed baseline - [Build Verifiers](https://docs.nvidia.com/nemo/gym/main/build-verifiers.md): Build the verifier that scores rollouts, in a Resources Server or a task folder, and the patterns for scoring. - [Verification Patterns](https://docs.nvidia.com/nemo/gym/main/build-verifiers/verification-patterns.md): Patterns for evaluating agent behavior and computing scores. - [Equivalence Match](https://docs.nvidia.com/nemo/gym/main/build-verifiers/verification-patterns/equivalence-match.md): Extract an agent's answer and compare it deterministically with a reference value, structure, or rule. - [Execution and State Match](https://docs.nvidia.com/nemo/gym/main/build-verifiers/verification-patterns/execution-and-state-match.md): Run generated code, queries, actions, or proofs and score the observed output or resulting state. - [LLM-as-Judge](https://docs.nvidia.com/nemo/gym/main/build-verifiers/verification-patterns/llm-as-judge.md): Prompt an LLM to evaluate the agent's output against rubrics, instructions, or reference answers. - [Multi-Reward Verification](https://docs.nvidia.com/nemo/gym/main/build-verifiers/multi-reward-verification.md): Score a rollout on multiple reward components for richer evaluation and multi-objective RL - [Evaluate](https://docs.nvidia.com/nemo/gym/main/evaluation.md): Run repeatable model and agent evaluations with NeMo Gym environments. - [Benchmarks](https://docs.nvidia.com/nemo/gym/main/evaluation/benchmarks.md): What benchmarks are, how they relate to environments, how to choose one, and how to interpret results. - [Compare Harnesses](https://docs.nvidia.com/nemo/gym/main/evaluation/harness.md): Ablate the agent harness the same way you ablate the model — swap it, hold everything else fixed, and measure the score delta. - [Browse Environments](https://docs.nvidia.com/nemo/gym/main/evaluation/environment-list.md): Browse built-in benchmark and training environments. - [Aggregate Metrics](https://docs.nvidia.com/nemo/gym/main/evaluation/aggregate-metrics.md): Compute summary statistics, customize per-environment metrics, and merge shard outputs from gym eval aggregate. - [Compare Results](https://docs.nvidia.com/nemo/gym/main/evaluation/compare-results.md): Compare evaluation runs and aggregate normalized benchmark scores with unpaired Welch confidence intervals. - [Rollout Health Checks](https://docs.nvidia.com/nemo/gym/main/evaluation/rollout-health.md): Verify that evaluation rollouts contain complete, internally consistent trajectory and model-call evidence. - [Diagnose Results](https://docs.nvidia.com/nemo/gym/main/evaluation/diagnose-results.md): Use BLADE and other analysis skills to diagnose evaluation results — why scores changed, which tasks failed, and what intervention to prioritize. - [Resume an Evaluation](https://docs.nvidia.com/nemo/gym/main/evaluation/resume.md): Reuse saved rollout results and check input and configuration identity before resuming. - [Training Tutorials](https://docs.nvidia.com/nemo/gym/main/tutorials/training-tutorials.md): - [NeMo RL](https://docs.nvidia.com/nemo/gym/main/tutorials/training-tutorials/nemo-rl-grpo.md): - [About Workplace Assistant](https://docs.nvidia.com/nemo/gym/main/tutorials/training-tutorials/nemo-rl-grpo/about-workplace-assistant.md): - [Gym Configuration](https://docs.nvidia.com/nemo/gym/main/tutorials/training-tutorials/nemo-rl-grpo/gym-configuration.md): - [Multi-Node Training](https://docs.nvidia.com/nemo/gym/main/tutorials/training-tutorials/nemo-rl-grpo/multi-node-training.md): - [NeMo RL Configuration](https://docs.nvidia.com/nemo/gym/main/tutorials/training-tutorials/nemo-rl-grpo/nemo-rl-configuration.md): - [Setup](https://docs.nvidia.com/nemo/gym/main/tutorials/training-tutorials/nemo-rl-grpo/setup.md): - [Single Node Training](https://docs.nvidia.com/nemo/gym/main/tutorials/training-tutorials/nemo-rl-grpo/single-node-training.md): - [Unsloth](https://docs.nvidia.com/nemo/gym/main/tutorials/training-tutorials/unsloth.md): - [Multi-Environment Training](https://docs.nvidia.com/nemo/gym/main/tutorials/training-tutorials/multi-environment-training.md): - [Training with VeRL](https://docs.nvidia.com/nemo/gym/main/tutorials/training-tutorials/verl.md): Run RL training on NeMo Gym environments using the nemo_gym recipe in verl. - [Offline Training (SFT/DPO)](https://docs.nvidia.com/nemo/gym/main/tutorials/training-tutorials/offline-training-w-rollouts.md): - [Training on External Agent Harnesses](https://docs.nvidia.com/nemo/gym/main/tutorials/training-tutorials/external-agent-harnesses.md): Capture exact token ids and log probabilities when an external agent harness drives its own model calls. - [Evaluation Tutorials](https://docs.nvidia.com/nemo/gym/main/tutorials/evaluation-tutorials.md): Run benchmark-specific evaluation workflows with NeMo Gym. - [Evaluate EvalPlus](https://docs.nvidia.com/nemo/gym/main/tutorials/evaluation-tutorials/evalplus.md): Run the EvalPlus coding benchmark and inspect rollout and aggregate metric outputs. - [BLADE Analysis Skill](https://docs.nvidia.com/nemo/gym/main/tutorials/evaluation-tutorials/blade-analysis-skill.md): Build and validate a BLADE-style benchmark analysis package from NeMo Gym rollout artifacts. - [Evaluate SWE-bench Pro with Hermes](https://docs.nvidia.com/nemo/gym/main/tutorials/evaluation-tutorials/hermes-swe-bench-pro.md): Prepare SWE-bench Pro tasks and run Hermes through the Environment Server. - [NOOA sessions and SWE-bench Pro](https://docs.nvidia.com/nemo/gym/main/tutorials/evaluation-tutorials/nooa-swe-bench-pro.md): Run NOOA through single_agent_turn and a Resources-owned sandbox. - [Evaluate Terminal-Bench 2.1 on the Harbor Path](https://docs.nvidia.com/nemo/gym/main/tutorials/evaluation-tutorials/terminal-bench-2-1-harbor.md): Run the 89 Terminal-Bench 2.1 task folders as published through the harbor Environment Server, then read the results. - [Run Multi-Turn User Simulation with NeMo User Sim](https://docs.nvidia.com/nemo/gym/main/tutorials/evaluation-tutorials/nemo-user-sim-rollout-collection.md): Generate NeMo User Sim validation tasks, evaluate an assistant, and inspect one rollout. - [Inspect NeMo User Sim Rollouts](https://docs.nvidia.com/nemo/gym/main/tutorials/evaluation-tutorials/nemo-user-sim-rollout-collection/rollout-inspection.md): Inspect conversations, tools, scores, invocations, and captured participant model calls. - [Troubleshoot NeMo User Sim Rollouts](https://docs.nvidia.com/nemo/gym/main/tutorials/evaluation-tutorials/nemo-user-sim-rollout-collection/rollout-troubleshooting.md): Diagnose collection, simulation, and grading failures in NeMo User Sim evaluations. - [NeMo User Sim Integration Reference](https://docs.nvidia.com/nemo/gym/main/tutorials/evaluation-tutorials/nemo-user-sim-rollout-collection/integration-reference.md): Architecture, ownership, request contracts, and saved fields for the NeMo User Sim integration. - [Evaluate Visual Coding with OpenCode](https://docs.nvidia.com/nemo/gym/main/tutorials/evaluation-tutorials/visual-coding-opencode.md): Run BabyVision with OpenCode in a sandbox session, with a Python image toolkit and a GPT judge. - [GDPval with NOOA](https://docs.nvidia.com/nemo/gym/main/tutorials/evaluation-tutorials/nooa-gdpval.md): Run NOOA with GDPval's existing sandbox sessions, task tools, and grading. - [Reverify Rollouts](https://docs.nvidia.com/nemo/gym/main/tutorials/evaluation-tutorials/reverify-rollouts.md): Recompute rewards from existing rollouts without re-running model inference. - [Terminal-Bench 2.1 with NOOA](https://docs.nvidia.com/nemo/gym/main/tutorials/evaluation-tutorials/nooa-terminal-bench-2-1.md): Run NOOA with the canonical Terminal-Bench 2.1 preparation and verifier. - [Build Environments](https://docs.nvidia.com/nemo/gym/main/environment-tutorials.md): - [Resources Server APIs](https://docs.nvidia.com/nemo/gym/main/environment-tutorials/resources-server-apis.md): Understand the supported Resources Server protocol APIs and choose the appropriate one to build your Environment - [Single-Step Environment](https://docs.nvidia.com/nemo/gym/main/environment-tutorials/single-step-environment.md): - [Multi-Step Environment](https://docs.nvidia.com/nemo/gym/main/environment-tutorials/multi-step-environment.md): - [Stateful Environment](https://docs.nvidia.com/nemo/gym/main/environment-tutorials/stateful-environment.md): - [MCP Resources Server](https://docs.nvidia.com/nemo/gym/main/environment-tutorials/mcp-resources-server.md): Serve a Resources Server's tools over the Model Context Protocol (MCP) with one config flag, and verify their use — with a runnable Claude Code example - [Real-World Environment](https://docs.nvidia.com/nemo/gym/main/environment-tutorials/real-world-environment.md): - [Generating Training Data](https://docs.nvidia.com/nemo/gym/main/environment-tutorials/real-world-environment/generating-training-data.md): - [Resources Server Implementation](https://docs.nvidia.com/nemo/gym/main/environment-tutorials/real-world-environment/resources-server-implementation.md): - [Integrate external libraries](https://docs.nvidia.com/nemo/gym/main/environment-tutorials/integrate-external-environments.md): - [Model Recipes](https://docs.nvidia.com/nemo/gym/main/model-recipes.md): - [Nemotron 3 Nano](https://docs.nvidia.com/nemo/gym/main/model-recipes/nemotron-3-nano.md): - [Nemotron 3 Super](https://docs.nvidia.com/nemo/gym/main/model-recipes/nemotron-3-super.md): - [Infrastructure](https://docs.nvidia.com/nemo/gym/main/infrastructure.md): - [Deployment Topology](https://docs.nvidia.com/nemo/gym/main/infrastructure/deployment-topology.md): - [Sandbox API](https://docs.nvidia.com/nemo/gym/main/infrastructure/sandbox.md): Use the provider-neutral sandbox module for isolated execution in NeMo Gym agents and environments. - [OpenSandbox Provider](https://docs.nvidia.com/nemo/gym/main/infrastructure/sandbox/opensandbox.md): Configure NeMo Gym sandboxes backed by an OpenSandbox server. - [Apptainer Provider](https://docs.nvidia.com/nemo/gym/main/infrastructure/sandbox/apptainer.md): Configure NeMo Gym sandboxes backed by local Apptainer instances. - [E2B Provider](https://docs.nvidia.com/nemo/gym/main/infrastructure/sandbox/e2b.md): Configure NeMo Gym sandboxes backed by E2B or an E2B-compatible gateway. - [OpenSandbox Best Practices](https://docs.nvidia.com/nemo/gym/main/infrastructure/sandbox/opensandbox-best-practices.md): Resource requests vs limits, timeouts, retries, and the failure signatures that look like bugs but are not — lessons from running thousands of concurrent sandboxes. - [Debugging OpenSandbox Sandboxes](https://docs.nvidia.com/nemo/gym/main/infrastructure/sandbox/opensandbox-debugging.md): How to see what CPU and memory a sandbox is getting and using, and how to detect OOM kills from the client. - [Docker Provider](https://docs.nvidia.com/nemo/gym/main/infrastructure/sandbox/docker.md): Configure NeMo Gym sandboxes backed by the local Docker daemon. - [ECS Fargate](https://docs.nvidia.com/nemo/gym/main/infrastructure/sandbox/ecs-fargate.md): Run sandboxes as AWS ECS Fargate tasks reached over an SSH sidecar. - [Adding a Sandbox Provider](https://docs.nvidia.com/nemo/gym/main/infrastructure/sandbox/adding-a-provider.md): Implement and register a sandbox runtime backend for NeMo Gym. - [Daytona Access](https://docs.nvidia.com/nemo/gym/main/infrastructure/sandbox/daytona-access.md): Obtain hosted Daytona credentials and check permissions and quota before running Gym workloads. - [Compose Services](https://docs.nvidia.com/nemo/gym/main/infrastructure/sandbox/compose.md): Run Docker Compose services through the asynchronous sandbox API. - [Engineering Notes](https://docs.nvidia.com/nemo/gym/main/infrastructure/engineering-notes.md): - [aiohttp vs httpx](https://docs.nvidia.com/nemo/gym/main/infrastructure/engineering-notes/aiohttp-vs-httpx.md): - [Responses API](https://docs.nvidia.com/nemo/gym/main/infrastructure/engineering-notes/responses-api-evolution.md): - [SWE RL Case Study](https://docs.nvidia.com/nemo/gym/main/infrastructure/engineering-notes/swe-rl-case-study.md): - [System Design](https://docs.nvidia.com/nemo/gym/main/infrastructure/engineering-notes/system-design.md): How gym env start loads configuration and launches servers, how rollout collection reaches agents through Environment Servers, and how Gym shuts down. - [Tavily Search Access and Setup](https://docs.nvidia.com/nemo/gym/main/infrastructure/tavily-search.md): Configure search credentials, an exclusion policy, and evaluation inputs for Tavily search. - [NVCF Long Requests (h2-ping-sidecar)](https://docs.nvidia.com/nemo/gym/main/infrastructure/nvcf-h2-ping-sidecar.md): Keep requests to NVCF endpoints alive past 340 seconds by letting Gym start an HTTP/2 PING sidecar. - [Orchestration](https://docs.nvidia.com/nemo/gym/main/orchestration.md): Deploy model services and run benchmarks on a cluster with gym eval submit. - [Quickstart](https://docs.nvidia.com/nemo/gym/main/orchestration/quickstart.md): Submit a benchmark with one vLLM service to a Slurm cluster. - [Configuration Reference](https://docs.nvidia.com/nemo/gym/main/orchestration/configuration.md): Every field of the gym eval submit config, with defaults and validation rules. - [Slurm Executor](https://docs.nvidia.com/nemo/gym/main/orchestration/slurm-executor.md): How gym eval submit turns a config into Slurm jobs, and what each job and run directory contains. - [Job Records](https://docs.nvidia.com/nemo/gym/main/orchestration/job-records.md): The manifest, resolved config, and local index that gym eval submit leaves behind. - [How-To Guides](https://docs.nvidia.com/nemo/gym/main/orchestration/how-to.md): Task-oriented guides for gym eval submit: multi-node vLLM, prefill/decode, node pools, custom harnesses, and resumable runs. - [Multi-Node vLLM](https://docs.nvidia.com/nemo/gym/main/orchestration/how-to/multi-node-vllm.md): Scale a vLLM service across GPUs and nodes with gym eval submit. - [Prefill/Decode Disaggregation](https://docs.nvidia.com/nemo/gym/main/orchestration/how-to/prefill-decode-disaggregation.md): Serve a model as separate prefill and decode vLLM tiers behind a vllm-router. - [Node Pools and Ray](https://docs.nvidia.com/nemo/gym/main/orchestration/how-to/node-pools-and-ray.md): Give services nodes of their own and span a Ray cluster across node pools. - [Custom Harness](https://docs.nvidia.com/nemo/gym/main/orchestration/how-to/custom-harness.md): Run your own script instead of gym eval run for a benchmark with its own harness. - [Resumable Runs](https://docs.nvidia.com/nemo/gym/main/orchestration/how-to/resumable-runs.md): Continue a benchmark automatically after a time limit, preemption, or node failure. - [Troubleshooting](https://docs.nvidia.com/nemo/gym/main/orchestration/troubleshooting.md): Errors you may hit with gym eval submit, what they mean, and common questions. - [Observability](https://docs.nvidia.com/nemo/gym/main/observability.md): Collect rollout evidence and monitor Gym services. - [Rollout Evidence](https://docs.nvidia.com/nemo/gym/main/observability/rollout-evidence.md): Collect, join, and inspect model-call and agent evidence. - [Harness Conformance](https://docs.nvidia.com/nemo/gym/main/observability/harness-conformance.md): Compare measured harness conformance and inspect retained trajectory evidence. - [OpenTelemetry](https://docs.nvidia.com/nemo/gym/main/observability/opentelemetry.md): Trace and measure Gym services with nemo-lens. - [Configuration](https://docs.nvidia.com/nemo/gym/main/observability/configuration.md): - [Distributed Tracing](https://docs.nvidia.com/nemo/gym/main/observability/distributed-tracing.md): - [Span Groups](https://docs.nvidia.com/nemo/gym/main/observability/span-groups.md): - [Metrics](https://docs.nvidia.com/nemo/gym/main/observability/metrics.md): - [Extending](https://docs.nvidia.com/nemo/gym/main/observability/extending.md): - [Collector for Submitted Jobs](https://docs.nvidia.com/nemo/gym/main/observability/eval-submit-collector.md): gym eval submit runs an OpenTelemetry Collector beside every benchmark job. - [Failure observations](https://docs.nvidia.com/nemo/gym/main/observability/rollout-failures.md): Failure evidence, explicit causes, reports and shared outcome policies. - [Reference](https://docs.nvidia.com/nemo/gym/main/reference.md): - [Configuration](https://docs.nvidia.com/nemo/gym/main/reference/configuration.md): - [RL Framework Compatibility](https://docs.nvidia.com/nemo/gym/main/reference/rl-framework-compatibility.md): NeMo Gym compatibility with NeMo RL, Unsloth, and VeRL - [CLI Commands](https://docs.nvidia.com/nemo/gym/main/reference/cli-commands.md): Reference for the unified gym CLI: command groups, flags, and migration from the legacy ng_* commands. - [FAQ](https://docs.nvidia.com/nemo/gym/main/reference/faq.md): - [ATIF Export](https://docs.nvidia.com/nemo/gym/main/reference/atif-export.md): Supported trajectories, preserved evidence, and limitations of strict ATIF export. - [Task Folder Files and Storage](https://docs.nvidia.com/nemo/gym/main/reference/task-folder-storage.md): Where task-folder runs write their files, which digests identify a task and a dataset, what a rerun reuses and what you can delete. - [Troubleshooting](https://docs.nvidia.com/nemo/gym/main/troubleshooting.md): - [Configuration Errors](https://docs.nvidia.com/nemo/gym/main/troubleshooting/configuration.md): - [Task Folders](https://docs.nvidia.com/nemo/gym/main/troubleshooting/task-folders.md): Symptoms, causes and fixes for task-folder runs of gym eval run and gym dataset validate. - [Overview](https://docs.nvidia.com/nemo/gym/main/contribute.md): - [Development Setup](https://docs.nvidia.com/nemo/gym/main/contribute/development-setup.md): - [Environments](https://docs.nvidia.com/nemo/gym/main/contribute/environments.md): - [Add a benchmark](https://docs.nvidia.com/nemo/gym/main/contribute/environments/adding-a-benchmark.md): Prepare, declare, and validate a reproducible benchmark using reusable Gym components and a manifest. - [New Environment](https://docs.nvidia.com/nemo/gym/main/contribute/environments/new-environment.md): Compose task and harness components, declare a workload manifest, and validate a contribution. - [Integrate RL Frameworks](https://docs.nvidia.com/nemo/gym/main/contribute/rl-framework-integration.md): - [Generation Backend](https://docs.nvidia.com/nemo/gym/main/contribute/rl-framework-integration/generation-backend-and-openai-compatible-http-server.md): - [Integration Footprint](https://docs.nvidia.com/nemo/gym/main/contribute/rl-framework-integration/gym-integration-footprint-and-form-factor.md): - [On-Policy Corrections](https://docs.nvidia.com/nemo/gym/main/contribute/rl-framework-integration/openai-compatible-http-server-on-policy-correction.md): - [Success Criteria](https://docs.nvidia.com/nemo/gym/main/contribute/rl-framework-integration/gym-rl-framework-integration-success-criteria.md): - [Agent Skills](https://docs.nvidia.com/nemo/gym/main/contribute/agent-skills.md): Vetted workflows for AI coding assistants bundled with NeMo Gym. - [Add an Exporter](https://docs.nvidia.com/nemo/gym/main/contribute/exporters.md): Send a finished evaluation run to a new destination with gym eval export. - [API Reference](https://docs.nvidia.com/nemo/gym/main/api/reference/api-reference.md): Auto-generated Python API reference for the nemo_gym library, and how the repository is laid out. - [nemo_gym](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym.md) - [nemo_gym.agents](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/agents.md) - [nemo_gym.agents.agent_activation_types](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/agents/agent_activation_types.md) - [nemo_gym.agents.agent_outcome](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/agents/agent_outcome.md) - [nemo_gym.agents.agent_tools](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/agents/agent_tools.md) - [nemo_gym.agents.base_responses_api_agent](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/agents/base_responses_api_agent.md) - [nemo_gym.agents.skills](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/agents/skills.md) - [nemo_gym.agents.testing](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/agents/testing.md) - [nemo_gym.agents.testing.agent_session_conformance](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/agents/testing/agent_session_conformance.md) - [nemo_gym.agents.testing.failure_assertions](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/agents/testing/failure_assertions.md) - [nemo_gym.agents.utils](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/agents/utils.md) - [nemo_gym.agents.utils.ordered_operations](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/agents/utils/ordered_operations.md) - [nemo_gym.agents.utils.sandbox_network](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/agents/utils/sandbox_network.md) - [nemo_gym.agents.utils.sandbox_observation](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/agents/utils/sandbox_observation.md) - [nemo_gym.agents.utils.sandbox_python](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/agents/utils/sandbox_python.md) - [nemo_gym.agents.utils.sandbox_session](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/agents/utils/sandbox_session.md) - [nemo_gym.agents.utils.supervisor_client](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/agents/utils/supervisor_client.md) - [nemo_gym.agents.utils.task_timeout](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/agents/utils/task_timeout.md) - [nemo_gym.agents.utils.verify_call](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/agents/utils/verify_call.md) - [nemo_gym.artifacts](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/artifacts.md) - [nemo_gym.artifacts.exporter_registry](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/artifacts/exporter_registry.md) - [nemo_gym.artifacts.exporters](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/artifacts/exporters.md) - [nemo_gym.artifacts.exporters.atif](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/artifacts/exporters/atif.md) - [nemo_gym.artifacts.exporters.base](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/artifacts/exporters/base.md) - [nemo_gym.artifacts.exporters.evalrdl](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/artifacts/exporters/evalrdl.md) - [nemo_gym.artifacts.exporters.isolated](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/artifacts/exporters/isolated.md) - [nemo_gym.artifacts.exporters.metrics](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/artifacts/exporters/metrics.md) - [nemo_gym.artifacts.exporters.mlflow](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/artifacts/exporters/mlflow.md) - [nemo_gym.artifacts.exporters.mlflow_config](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/artifacts/exporters/mlflow_config.md) - [nemo_gym.artifacts.exporters.wandb](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/artifacts/exporters/wandb.md) - [nemo_gym.artifacts.exporters.wandb_config](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/artifacts/exporters/wandb_config.md) - [nemo_gym.artifacts.resolved_config](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/artifacts/resolved_config.md) - [nemo_gym.artifacts.rollout_rows](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/artifacts/rollout_rows.md) - [nemo_gym.artifacts.run_artifacts](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/artifacts/run_artifacts.md) - [nemo_gym.artifacts.run_files](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/artifacts/run_files.md) - [nemo_gym.cli](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/cli.md) - [nemo_gym.cli.agents](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/cli/agents.md) - [nemo_gym.cli.dataset](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/cli/dataset.md) - [nemo_gym.cli.dev](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/cli/dev.md) - [nemo_gym.cli.env](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/cli/env.md) - [nemo_gym.cli.eval](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/cli/eval.md) - [nemo_gym.cli.general](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/cli/general.md) - [nemo_gym.cli.legacy](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/cli/legacy.md) - [nemo_gym.cli.main](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/cli/main.md) - [nemo_gym.cli.models](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/cli/models.md) - [nemo_gym.cli.resources_servers](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/cli/resources_servers.md) - [nemo_gym.cli.utils](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/cli/utils.md) - [nemo_gym.client](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/client.md) - [nemo_gym.client.http_client](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/client/http_client.md) - [nemo_gym.client.openai_client](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/client/openai_client.md) - [nemo_gym.client.server_client](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/client/server_client.md) - [nemo_gym.config](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/config.md) - [nemo_gym.config.config_types](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/config/config_types.md) - [nemo_gym.config.global_config](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/config/global_config.md) - [nemo_gym.config.secret_utils](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/config/secret_utils.md) - [nemo_gym.config.server_metadata](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/config/server_metadata.md) - [nemo_gym.core](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/core.md) - [nemo_gym.core.correlation](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/core/correlation.md) - [nemo_gym.core.episode_context](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/core/episode_context.md) - [nemo_gym.core.episode_types](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/core/episode_types.md) - [nemo_gym.core.failure_kinds](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/core/failure_kinds.md) - [nemo_gym.core.failure_records](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/core/failure_records.md) - [nemo_gym.core.failure_transport](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/core/failure_transport.md) - [nemo_gym.core.media](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/core/media.md) - [nemo_gym.core.model_call_sampling](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/core/model_call_sampling.md) - [nemo_gym.core.operation_timing](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/core/operation_timing.md) - [nemo_gym.core.outcome_policy](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/core/outcome_policy.md) - [nemo_gym.core.responses_api_types](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/core/responses_api_types.md) - [nemo_gym.core.tool_output_images](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/core/tool_output_images.md) - [nemo_gym.core.typed_exceptions](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/core/typed_exceptions.md) - [nemo_gym.data](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data.md) - [nemo_gym.data.dataset_metrics](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/dataset_metrics.md) - [nemo_gym.data.dataset_orchestrator](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/dataset_orchestrator.md) - [nemo_gym.data.dataset_preparation](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/dataset_preparation.md) - [nemo_gym.data.gitlab_utils](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/gitlab_utils.md) - [nemo_gym.data.hf_utils](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/hf_utils.md) - [nemo_gym.data.prompt](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/prompt.md) - [nemo_gym.data.task_data](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/task_data.md) - [nemo_gym.data.task_materialization](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/task_materialization.md) - [nemo_gym.data.task_selection](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/task_selection.md) - [nemo_gym.data.tasks](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/tasks.md) - [nemo_gym.data.tasks.harbor](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/tasks/harbor.md) - [nemo_gym.data.tasks.harbor.build_compose](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/tasks/harbor/build_compose.md) - [nemo_gym.data.tasks.harbor.build_options](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/tasks/harbor/build_options.md) - [nemo_gym.data.tasks.harbor.build_plan](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/tasks/harbor/build_plan.md) - [nemo_gym.data.tasks.harbor.build_records](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/tasks/harbor/build_records.md) - [nemo_gym.data.tasks.harbor.build_tools](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/tasks/harbor/build_tools.md) - [nemo_gym.data.tasks.harbor.compose_overlay](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/tasks/harbor/compose_overlay.md) - [nemo_gym.data.tasks.harbor.compose_variables](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/tasks/harbor/compose_variables.md) - [nemo_gym.data.tasks.harbor.dataset_config](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/tasks/harbor/dataset_config.md) - [nemo_gym.data.tasks.harbor.digest](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/tasks/harbor/digest.md) - [nemo_gym.data.tasks.harbor.dockerfile](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/tasks/harbor/dockerfile.md) - [nemo_gym.data.tasks.harbor.entries](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/tasks/harbor/entries.md) - [nemo_gym.data.tasks.harbor.errors](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/tasks/harbor/errors.md) - [nemo_gym.data.tasks.harbor.harbor_compat](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/tasks/harbor/harbor_compat.md) - [nemo_gym.data.tasks.harbor.hub](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/tasks/harbor/hub.md) - [nemo_gym.data.tasks.harbor.image_build](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/tasks/harbor/image_build.md) - [nemo_gym.data.tasks.harbor.image_configs](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/tasks/harbor/image_configs.md) - [nemo_gym.data.tasks.harbor.manifest](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/tasks/harbor/manifest.md) - [nemo_gym.data.tasks.harbor.materialize](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/tasks/harbor/materialize.md) - [nemo_gym.data.tasks.harbor.models](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/tasks/harbor/models.md) - [nemo_gym.data.tasks.harbor.package_store](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/tasks/harbor/package_store.md) - [nemo_gym.data.tasks.harbor.parameterized](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/tasks/harbor/parameterized.md) - [nemo_gym.data.tasks.harbor.prepare](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/tasks/harbor/prepare.md) - [nemo_gym.data.tasks.harbor.prepare_builds](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/tasks/harbor/prepare_builds.md) - [nemo_gym.data.tasks.harbor.publish](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/tasks/harbor/publish.md) - [nemo_gym.data.tasks.harbor.python_verifier](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/tasks/harbor/python_verifier.md) - [nemo_gym.data.tasks.harbor.records](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/tasks/harbor/records.md) - [nemo_gym.data.tasks.harbor.registry](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/tasks/harbor/registry.md) - [nemo_gym.data.tasks.harbor.run_summary](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/tasks/harbor/run_summary.md) - [nemo_gym.data.tasks.harbor.scaffold](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/tasks/harbor/scaffold.md) - [nemo_gym.data.tasks.harbor.task](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/tasks/harbor/task.md) - [nemo_gym.data.tasks.harbor.task_data](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/tasks/harbor/task_data.md) - [nemo_gym.data.tasks.harbor.validation_report](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/tasks/harbor/validation_report.md) - [nemo_gym.data.tasks.harbor.verifier_env](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/tasks/harbor/verifier_env.md) - [nemo_gym.data.vision_input](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/data/vision_input.md) - [nemo_gym.environment](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/environment.md) - [nemo_gym.environment.agent_registry](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/environment/agent_registry.md) - [nemo_gym.environment.assets](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/environment/assets.md) - [nemo_gym.environment.benchmark_preparation](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/environment/benchmark_preparation.md) - [nemo_gym.environment.benchmarks](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/environment/benchmarks.md) - [nemo_gym.environment.comparison](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/environment/comparison.md) - [nemo_gym.environment.deliverables](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/environment/deliverables.md) - [nemo_gym.environment.discovery](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/environment/discovery.md) - [nemo_gym.environment.in_process](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/environment/in_process.md) - [nemo_gym.environment.manifest](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/environment/manifest.md) - [nemo_gym.environment.model_registry](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/environment/model_registry.md) - [nemo_gym.environment.onboarding](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/environment/onboarding.md) - [nemo_gym.environment.publication](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/environment/publication.md) - [nemo_gym.environment.registry](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/environment/registry.md) - [nemo_gym.environment.resources_server_registry](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/environment/resources_server_registry.md) - [nemo_gym.environment.scaffold](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/environment/scaffold.md) - [nemo_gym.environment.validation](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/environment/validation.md) - [nemo_gym.eval](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/eval.md) - [nemo_gym.eval.comparison](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/eval/comparison.md) - [nemo_gym.eval.comparison.diff](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/eval/comparison/diff.md) - [nemo_gym.eval.comparison.loading](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/eval/comparison/loading.md) - [nemo_gym.eval.comparison.pooling](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/eval/comparison/pooling.md) - [nemo_gym.eval.comparison.report](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/eval/comparison/report.md) - [nemo_gym.eval.comparison.runner](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/eval/comparison/runner.md) - [nemo_gym.eval.comparison.schema](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/eval/comparison/schema.md) - [nemo_gym.eval.comparison.statistics](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/eval/comparison/statistics.md) - [nemo_gym.eval.comparison.statistics.unpaired_welch_fixed_suite](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/eval/comparison/statistics/unpaired_welch_fixed_suite.md) - [nemo_gym.eval.comparison.suite](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/eval/comparison/suite.md) - [nemo_gym.eval.comparison.suite.coverage](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/eval/comparison/suite/coverage.md) - [nemo_gym.eval.comparison.suite.failures](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/eval/comparison/suite/failures.md) - [nemo_gym.eval.comparison.suite.loading](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/eval/comparison/suite/loading.md) - [nemo_gym.eval.comparison.suite.pooling](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/eval/comparison/suite/pooling.md) - [nemo_gym.eval.comparison.suite.processing](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/eval/comparison/suite/processing.md) - [nemo_gym.eval.comparison.suite.report](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/eval/comparison/suite/report.md) - [nemo_gym.eval.comparison.suite.runner](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/eval/comparison/suite/runner.md) - [nemo_gym.eval.comparison.suite.schema](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/eval/comparison/suite/schema.md) - [nemo_gym.eval.eval_plan](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/eval/eval_plan.md) - [nemo_gym.eval.rollout_compare](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/eval/rollout_compare.md) - [nemo_gym.eval.rollout_compare.command](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/eval/rollout_compare/command.md) - [nemo_gym.eval.rollout_compare.compare](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/eval/rollout_compare/compare.md) - [nemo_gym.eval.rollout_compare.join](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/eval/rollout_compare/join.md) - [nemo_gym.eval.rollout_compare.report](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/eval/rollout_compare/report.md) - [nemo_gym.eval.rollout_compare.rows](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/eval/rollout_compare/rows.md) - [nemo_gym.eval.rollout_compare.schema](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/eval/rollout_compare/schema.md) - [nemo_gym.eval.rollout_reverification](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/eval/rollout_reverification.md) - [nemo_gym.metrics](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/metrics.md) - [nemo_gym.metrics.metrics_config](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/metrics/metrics_config.md) - [nemo_gym.metrics.reward_profile](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/metrics/reward_profile.md) - [nemo_gym.models](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models.md) - [nemo_gym.models.anthropic_converter](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models/anthropic_converter.md) - [nemo_gym.models.anthropic_utils](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models/anthropic_utils.md) - [nemo_gym.models.base_responses_api_model](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models/base_responses_api_model.md) - [nemo_gym.models.chat_completions_converter](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models/chat_completions_converter.md) - [nemo_gym.models.chat_streaming](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models/chat_streaming.md) - [nemo_gym.models.model_call_capture](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models/model_call_capture.md) - [nemo_gym.models.responses_streaming](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models/responses_streaming.md) - [nemo_gym.models.token_capture](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models/token_capture.md) - [nemo_gym.models.token_capture.adapters](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models/token_capture/adapters.md) - [nemo_gym.models.token_capture.adapters.megatron](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models/token_capture/adapters/megatron.md) - [nemo_gym.models.token_capture.adapters.vllm](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models/token_capture/adapters/vllm.md) - [nemo_gym.models.token_capture.builder](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models/token_capture/builder.md) - [nemo_gym.models.token_capture.completion](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models/token_capture/completion.md) - [nemo_gym.models.token_capture.config](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models/token_capture/config.md) - [nemo_gym.models.token_capture.conformance](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models/token_capture/conformance.md) - [nemo_gym.models.token_capture.consumer](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models/token_capture/consumer.md) - [nemo_gym.models.token_capture.control_routes](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models/token_capture/control_routes.md) - [nemo_gym.models.token_capture.delivery](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models/token_capture/delivery.md) - [nemo_gym.models.token_capture.external_capture](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models/token_capture/external_capture.md) - [nemo_gym.models.token_capture.fingerprint](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models/token_capture/fingerprint.md) - [nemo_gym.models.token_capture.harness_notes](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models/token_capture/harness_notes.md) - [nemo_gym.models.token_capture.harness_notes.base](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models/token_capture/harness_notes/base.md) - [nemo_gym.models.token_capture.harness_notes.opencode](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models/token_capture/harness_notes/opencode.md) - [nemo_gym.models.token_capture.lineage](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models/token_capture/lineage.md) - [nemo_gym.models.token_capture.protocols](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models/token_capture/protocols.md) - [nemo_gym.models.token_capture.records](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models/token_capture/records.md) - [nemo_gym.models.token_capture.replay](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models/token_capture/replay.md) - [nemo_gym.models.token_capture.sink](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models/token_capture/sink.md) - [nemo_gym.models.token_capture.staging](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models/token_capture/staging.md) - [nemo_gym.models.token_capture.staging.capture](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models/token_capture/staging/capture.md) - [nemo_gym.models.token_capture.staging.conformance](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models/token_capture/staging/conformance.md) - [nemo_gym.models.token_capture.staging.digest](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models/token_capture/staging/digest.md) - [nemo_gym.models.token_capture.staging.protocols](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models/token_capture/staging/protocols.md) - [nemo_gym.models.token_capture.staging.rebuild](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models/token_capture/staging/rebuild.md) - [nemo_gym.models.token_capture.staging.records](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models/token_capture/staging/records.md) - [nemo_gym.models.token_capture.staging.routes](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models/token_capture/staging/routes.md) - [nemo_gym.models.token_capture.staging.terminal](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models/token_capture/staging/terminal.md) - [nemo_gym.models.token_capture.store](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models/token_capture/store.md) - [nemo_gym.models.token_capture.terminal](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/models/token_capture/terminal.md) - [nemo_gym.orchestration](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/orchestration.md) - [nemo_gym.orchestration.api](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/orchestration/api.md) - [nemo_gym.orchestration.collect_evidence](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/orchestration/collect_evidence.md) - [nemo_gym.orchestration.executors](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/orchestration/executors.md) - [nemo_gym.orchestration.executors.base](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/orchestration/executors/base.md) - [nemo_gym.orchestration.executors.connection](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/orchestration/executors/connection.md) - [nemo_gym.orchestration.executors.git_ref](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/orchestration/executors/git_ref.md) - [nemo_gym.orchestration.executors.judge_profile](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/orchestration/executors/judge_profile.md) - [nemo_gym.orchestration.executors.kubernetes](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/orchestration/executors/kubernetes.md) - [nemo_gym.orchestration.executors.kubernetes_script](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/orchestration/executors/kubernetes_script.md) - [nemo_gym.orchestration.executors.otel](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/orchestration/executors/otel.md) - [nemo_gym.orchestration.executors.resume_script](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/orchestration/executors/resume_script.md) - [nemo_gym.orchestration.executors.script_templates](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/orchestration/executors/script_templates.md) - [nemo_gym.orchestration.executors.slurm](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/orchestration/executors/slurm.md) - [nemo_gym.orchestration.executors.slurm_script](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/orchestration/executors/slurm_script.md) - [nemo_gym.orchestration.executors.utils](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/orchestration/executors/utils.md) - [nemo_gym.orchestration.export_step](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/orchestration/export_step.md) - [nemo_gym.orchestration.exports](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/orchestration/exports.md) - [nemo_gym.orchestration.jobs](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/orchestration/jobs.md) - [nemo_gym.orchestration.ray_serve_gateway](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/orchestration/ray_serve_gateway.md) - [nemo_gym.orchestration.recovery](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/orchestration/recovery.md) - [nemo_gym.orchestration.submit](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/orchestration/submit.md) - [nemo_gym.package_info](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/package_info.md) - [nemo_gym.rollouts](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/rollouts.md) - [nemo_gym.rollouts.batch_status](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/rollouts/batch_status.md) - [nemo_gym.rollouts.harness_capabilities](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/rollouts/harness_capabilities.md) - [nemo_gym.rollouts.harness_capabilities.behavior](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/rollouts/harness_capabilities/behavior.md) - [nemo_gym.rollouts.harness_capabilities.checker](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/rollouts/harness_capabilities/checker.md) - [nemo_gym.rollouts.harness_capabilities.checks](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/rollouts/harness_capabilities/checks.md) - [nemo_gym.rollouts.harness_capabilities.cli](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/rollouts/harness_capabilities/cli.md) - [nemo_gym.rollouts.harness_capabilities.compaction](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/rollouts/harness_capabilities/compaction.md) - [nemo_gym.rollouts.harness_capabilities.contracts](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/rollouts/harness_capabilities/contracts.md) - [nemo_gym.rollouts.harness_capabilities.health](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/rollouts/harness_capabilities/health.md) - [nemo_gym.rollouts.harness_capabilities.results](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/rollouts/harness_capabilities/results.md) - [nemo_gym.rollouts.health](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/rollouts/health.md) - [nemo_gym.rollouts.health.checks](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/rollouts/health/checks.md) - [nemo_gym.rollouts.health.timeouts](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/rollouts/health/timeouts.md) - [nemo_gym.rollouts.health.types](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/rollouts/health/types.md) - [nemo_gym.rollouts.model_calls](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/rollouts/model_calls.md) - [nemo_gym.rollouts.omitted_rollouts](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/rollouts/omitted_rollouts.md) - [nemo_gym.rollouts.rollout_collection](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/rollouts/rollout_collection.md) - [nemo_gym.rollouts.rollout_health](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/rollouts/rollout_health.md) - [nemo_gym.rollouts.rollout_outcomes](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/rollouts/rollout_outcomes.md) - [nemo_gym.rollouts.rollout_perf](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/rollouts/rollout_perf.md) - [nemo_gym.rollouts.rollout_records](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/rollouts/rollout_records.md) - [nemo_gym.rollouts.rollout_recovery](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/rollouts/rollout_recovery.md) - [nemo_gym.rollouts.rollout_store](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/rollouts/rollout_store.md) - [nemo_gym.rollouts.trajectory_builder](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/rollouts/trajectory_builder.md) - [nemo_gym.sandbox](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox.md) - [nemo_gym.sandbox.access](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/access.md) - [nemo_gym.sandbox.adapters](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/adapters.md) - [nemo_gym.sandbox.adapters.docker_compose](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/adapters/docker_compose.md) - [nemo_gym.sandbox.api](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/api.md) - [nemo_gym.sandbox.attribution](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/attribution.md) - [nemo_gym.sandbox.compose_config](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/compose_config.md) - [nemo_gym.sandbox.compose_fields](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/compose_fields.md) - [nemo_gym.sandbox.compose_volumes](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/compose_volumes.md) - [nemo_gym.sandbox.config](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/config.md) - [nemo_gym.sandbox.failures](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/failures.md) - [nemo_gym.sandbox.observability](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/observability.md) - [nemo_gym.sandbox.providers](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/providers.md) - [nemo_gym.sandbox.providers.apptainer](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/providers/apptainer.md) - [nemo_gym.sandbox.providers.apptainer.provider](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/providers/apptainer/provider.md) - [nemo_gym.sandbox.providers.base](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/providers/base.md) - [nemo_gym.sandbox.providers.daytona](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/providers/daytona.md) - [nemo_gym.sandbox.providers.daytona.provider](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/providers/daytona/provider.md) - [nemo_gym.sandbox.providers.docker](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/providers/docker.md) - [nemo_gym.sandbox.providers.docker.provider](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/providers/docker/provider.md) - [nemo_gym.sandbox.providers.docker.pty](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/providers/docker/pty.md) - [nemo_gym.sandbox.providers.e2b](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/providers/e2b.md) - [nemo_gym.sandbox.providers.e2b.build](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/providers/e2b/build.md) - [nemo_gym.sandbox.providers.e2b.provider](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/providers/e2b/provider.md) - [nemo_gym.sandbox.providers.ecs_fargate](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/providers/ecs_fargate.md) - [nemo_gym.sandbox.providers.ecs_fargate.engine](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/providers/ecs_fargate/engine.md) - [nemo_gym.sandbox.providers.ecs_fargate.provider](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/providers/ecs_fargate/provider.md) - [nemo_gym.sandbox.providers.enroot](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/providers/enroot.md) - [nemo_gym.sandbox.providers.enroot.provider](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/providers/enroot/provider.md) - [nemo_gym.sandbox.providers.local](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/providers/local.md) - [nemo_gym.sandbox.providers.local.provider](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/providers/local/provider.md) - [nemo_gym.sandbox.providers.opensandbox](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/providers/opensandbox.md) - [nemo_gym.sandbox.providers.opensandbox.cleanup_sandboxes](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/providers/opensandbox/cleanup_sandboxes.md) - [nemo_gym.sandbox.providers.opensandbox.failures](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/providers/opensandbox/failures.md) - [nemo_gym.sandbox.providers.opensandbox.provider](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/providers/opensandbox/provider.md) - [nemo_gym.sandbox.providers.opensandbox.pty](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/providers/opensandbox/pty.md) - [nemo_gym.sandbox.providers.opensandbox.resource_telemetry](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/providers/opensandbox/resource_telemetry.md) - [nemo_gym.sandbox.providers.opensandbox.snapshots](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/providers/opensandbox/snapshots.md) - [nemo_gym.sandbox.providers.openshell](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/providers/openshell.md) - [nemo_gym.sandbox.providers.openshell.provider](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/providers/openshell/provider.md) - [nemo_gym.sandbox.providers.registry](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/providers/registry.md) - [nemo_gym.sandbox.providers.utils](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/providers/utils.md) - [nemo_gym.sandbox.resource_telemetry](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/resource_telemetry.md) - [nemo_gym.sandbox.shell](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/shell.md) - [nemo_gym.sandbox.transfer](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/transfer.md) - [nemo_gym.sandbox.utils](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/sandbox/utils.md) - [nemo_gym.server_runtime](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/server_runtime.md) - [nemo_gym.server_runtime.base_environment_server](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/server_runtime/base_environment_server.md) - [nemo_gym.server_runtime.base_resources_server](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/server_runtime/base_resources_server.md) - [nemo_gym.server_runtime.base_server](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/server_runtime/base_server.md) - [nemo_gym.server_runtime.entrypoint_flags](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/server_runtime/entrypoint_flags.md) - [nemo_gym.server_runtime.mcp_auto_exposure](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/server_runtime/mcp_auto_exposure.md) - [nemo_gym.server_runtime.ray_support](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/server_runtime/ray_support.md) - [nemo_gym.server_runtime.registry_server](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/server_runtime/registry_server.md) - [nemo_gym.server_runtime.request_handling](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/server_runtime/request_handling.md) - [nemo_gym.server_runtime.request_handling.disconnect](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/server_runtime/request_handling/disconnect.md) - [nemo_gym.server_runtime.request_handling.errors](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/server_runtime/request_handling/errors.md) - [nemo_gym.server_runtime.request_handling.exception_handling](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/server_runtime/request_handling/exception_handling.md) - [nemo_gym.server_runtime.request_handling.model_sampling](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/server_runtime/request_handling/model_sampling.md) - [nemo_gym.server_runtime.request_handling.request_context](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/server_runtime/request_handling/request_context.md) - [nemo_gym.server_runtime.request_handling.rollout_context](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/server_runtime/request_handling/rollout_context.md) - [nemo_gym.server_runtime.request_handling.session_id](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/server_runtime/request_handling/session_id.md) - [nemo_gym.server_runtime.request_handling.session_routing](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/server_runtime/request_handling/session_routing.md) - [nemo_gym.server_runtime.request_handling.trace_context](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/server_runtime/request_handling/trace_context.md) - [nemo_gym.server_runtime.runtime_dir](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/server_runtime/runtime_dir.md) - [nemo_gym.server_runtime.sandbox_lending](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/server_runtime/sandbox_lending.md) - [nemo_gym.server_runtime.server_status](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/server_runtime/server_status.md) - [nemo_gym.server_runtime.single_agent_turn_types](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/server_runtime/single_agent_turn_types.md) - [nemo_gym.server_runtime.supervisor](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/server_runtime/supervisor.md) - [nemo_gym.server_runtime.supervisor.h2_ping_sidecar](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/server_runtime/supervisor/h2_ping_sidecar.md) - [nemo_gym.server_runtime.supervisor.h2_ping_sidecar.config](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/server_runtime/supervisor/h2_ping_sidecar/config.md) - [nemo_gym.server_runtime.supervisor.h2_ping_sidecar.launcher](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/server_runtime/supervisor/h2_ping_sidecar/launcher.md) - [nemo_gym.server_runtime.supervisor.server_supervisor](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/server_runtime/supervisor/server_supervisor.md) - [nemo_gym.server_runtime.supervisor.server_venv](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/server_runtime/supervisor/server_venv.md) - [nemo_gym.server_runtime.testing](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/server_runtime/testing.md) - [nemo_gym.server_runtime.testing.session_conformance](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/server_runtime/testing/session_conformance.md) - [nemo_gym.server_runtime.tool_access](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/server_runtime/tool_access.md) - [nemo_gym.server_runtime.uvicorn_reuse_port](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/server_runtime/uvicorn_reuse_port.md) - [nemo_gym.server_runtime.verifier_observability](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/server_runtime/verifier_observability.md) - [nemo_gym.telemetry](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/telemetry.md) - [nemo_gym.telemetry.concurrency](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/telemetry/concurrency.md) - [nemo_gym.telemetry.config](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/telemetry/config.md) - [nemo_gym.telemetry.connection_pool](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/telemetry/connection_pool.md) - [nemo_gym.telemetry.contrib](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/telemetry/contrib.md) - [nemo_gym.telemetry.endpoints](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/telemetry/endpoints.md) - [nemo_gym.telemetry.gym_metrics](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/telemetry/gym_metrics.md) - [nemo_gym.telemetry.memory](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/telemetry/memory.md) - [nemo_gym.telemetry.metrics](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/telemetry/metrics.md) - [nemo_gym.telemetry.observer](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/telemetry/observer.md) - [nemo_gym.telemetry.profiling](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/telemetry/profiling.md) - [nemo_gym.telemetry.resources_operations](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/telemetry/resources_operations.md) - [nemo_gym.telemetry.setup](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/telemetry/setup.md) - [nemo_gym.telemetry.span_groups](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/telemetry/span_groups.md) - [nemo_gym.telemetry.spans](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/telemetry/spans.md) - [nemo_gym.telemetry.startup](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/telemetry/startup.md) - [nemo_gym.trajectory](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/trajectory.md) - [nemo_gym.trajectory.atif_export](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/trajectory/atif_export.md) - [nemo_gym.trajectory.atif_json](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/trajectory/atif_json.md) - [nemo_gym.trajectory.atif_reverification](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/trajectory/atif_reverification.md) - [nemo_gym.trajectory.atif_v1_7](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/trajectory/atif_v1_7.md) - [nemo_gym.trajectory.model_call_record](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/trajectory/model_call_record.md) - [nemo_gym.trajectory.tool_timing](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/trajectory/tool_timing.md) - [nemo_gym.trajectory.trajectory_format](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/trajectory/trajectory_format.md) - [nemo_gym.trajectory.trajectory_record](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/trajectory/trajectory_record.md) - [nemo_gym.utils](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/utils.md) - [nemo_gym.utils.compat](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/utils/compat.md) - [nemo_gym.utils.decorators](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/utils/decorators.md) - [nemo_gym.utils.exception_inspection](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/utils/exception_inspection.md) - [nemo_gym.utils.gym_commit](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/utils/gym_commit.md) - [nemo_gym.utils.install_lock](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/utils/install_lock.md) - [nemo_gym.utils.network_policy](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/utils/network_policy.md) - [nemo_gym.utils.path_utils](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/utils/path_utils.md) - [nemo_gym.utils.server_process](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/utils/server_process.md) - [nemo_gym.utils.text](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/utils/text.md) - [nemo_gym.utils.tqdm](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/utils/tqdm.md) - [nemo_gym.utils.venv_setup](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/utils/venv_setup.md) - [nemo_gym.verification](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/verification.md) - [nemo_gym.verification.judge](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/verification/judge.md) - [nemo_gym.verification.sandbox_verification](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/verification/sandbox_verification.md) - [nemo_gym.verification.verifier_fixture](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/verification/verifier_fixture.md) - [nemo_gym.verifiers](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/verifiers.md) - [nemo_gym.verifiers.files](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/verifiers/files.md) - [nemo_gym.verifiers.runner](https://docs.nvidia.com/nemo/gym/main/nemo-gym/nemo_gym/verifiers/runner.md)