Official Technical Resources: NVIDIA (page 13)
Official Technical Resources · NVIDIA: 264 entries in total; this page lists entries 181–195.
Benchmarks
What benchmarks are, how they relate to environments, how to choose one, and how to interpret results.
Evidence use: For understanding GPU computing, training, retrieval, inference serving, reliability or deployment decisions.
Limitations: Directly evidences only the implementation and claims of NVIDIA's technology stack. Recheck versioned interfaces and experimental capabilities before use.
Related concepts: Model Evaluation and Benchmarks · LLM Inference Optimization · Model Deployment Options
Link last checked: 2026-09-25
Diagnose Results
Use BLADE and other analysis skills to diagnose evaluation results — why scores changed, which tasks failed, and what intervention to prioritize.
Evidence use: For understanding GPU computing, training, retrieval, inference serving, reliability or deployment decisions.
Limitations: Directly evidences only the implementation and claims of NVIDIA's technology stack. Recheck versioned interfaces and experimental capabilities before use.
Related concepts: Model Evaluation and Benchmarks · LLM Inference Optimization · Model Deployment Options
Link last checked: 2026-09-25
Compare Harnesses
Ablate the agent harness the same way you ablate the model — swap it, hold everything else fixed, and measure the score delta.
Evidence use: For understanding GPU computing, training, retrieval, inference serving, reliability or deployment decisions.
Limitations: Directly evidences only the implementation and claims of NVIDIA's technology stack. Recheck versioned interfaces and experimental capabilities before use.
Related concepts: Model Evaluation and Benchmarks · LLM Inference Optimization · Model Deployment Options
Link last checked: 2026-09-25
Rollout Health Checks
Verify that evaluation rollouts contain complete, internally consistent trajectory and model-call evidence.
Evidence use: For understanding GPU computing, training, retrieval, inference serving, reliability or deployment decisions.
Limitations: Directly evidences only the implementation and claims of NVIDIA's technology stack. Recheck versioned interfaces and experimental capabilities before use.
Related concepts: Model Evaluation and Benchmarks · LLM Inference Optimization · Model Deployment Options
Link last checked: 2026-09-25
System Design
This section describes how NeMo Gym components interact during startup and execution.
Evidence use: For understanding GPU computing, training, retrieval, inference serving, reliability or deployment decisions.
Limitations: Directly evidences only the implementation and claims of NVIDIA's technology stack. Recheck versioned interfaces and experimental capabilities before use.
Related concepts: Model Evaluation and Benchmarks · LLM Inference Optimization · Model Deployment Options
Link last checked: 2026-09-25
Sandbox API
Use the provider-neutral sandbox module for isolated execution in NeMo Gym agents and environments.
Evidence use: For understanding GPU computing, training, retrieval, inference serving, reliability or deployment decisions.
Limitations: Directly evidences only the implementation and claims of NVIDIA's technology stack. Recheck versioned interfaces and experimental capabilities before use.
Related concepts: Model Evaluation and Benchmarks · LLM Inference Optimization · Model Deployment Options
Link last checked: 2026-09-25
Adding a Sandbox Provider
Implement and register a sandbox runtime backend for NeMo Gym.
Evidence use: For understanding GPU computing, training, retrieval, inference serving, reliability or deployment decisions.
Limitations: Directly evidences only the implementation and claims of NVIDIA's technology stack. Recheck versioned interfaces and experimental capabilities before use.
Related concepts: Model Evaluation and Benchmarks · LLM Inference Optimization · Model Deployment Options
Link last checked: 2026-09-25
OpenSandbox Best Practices
Resource requests vs limits, timeouts, retries, and the failure signatures that look like bugs but are not — lessons from running thousands of concurrent sandboxes.
Evidence use: For understanding GPU computing, training, retrieval, inference serving, reliability or deployment decisions.
Limitations: Directly evidences only the implementation and claims of NVIDIA's technology stack. Recheck versioned interfaces and experimental capabilities before use.
Related concepts: Model Evaluation and Benchmarks · LLM Inference Optimization · Model Deployment Options
Link last checked: 2026-09-25
Distributed Tracing
A NeMo Gym rollout crosses several processes. This page explains how those hops become a single trace, and what to check when they do not.
Evidence use: For understanding GPU computing, training, retrieval, inference serving, reliability or deployment decisions.
Limitations: Directly evidences only the implementation and claims of NVIDIA's technology stack. Recheck versioned interfaces and experimental capabilities before use.
Related concepts: Model Evaluation and Benchmarks · LLM Inference Optimization · Model Deployment Options
Link last checked: 2026-09-25
Collector for Submitted Jobs
gym eval submit runs an OpenTelemetry Collector beside every benchmark job.
Evidence use: For understanding GPU computing, training, retrieval, inference serving, reliability or deployment decisions.
Limitations: Directly evidences only the implementation and claims of NVIDIA's technology stack. Recheck versioned interfaces and experimental capabilities before use.
Related concepts: Model Evaluation and Benchmarks · LLM Inference Optimization · Model Deployment Options
Link last checked: 2026-09-25
Extending
Add spans to your own resources server, agent harness, or model server.
Evidence use: For understanding GPU computing, training, retrieval, inference serving, reliability or deployment decisions.
Limitations: Directly evidences only the implementation and claims of NVIDIA's technology stack. Recheck versioned interfaces and experimental capabilities before use.
Related concepts: Model Evaluation and Benchmarks · LLM Inference Optimization · Model Deployment Options
Link last checked: 2026-09-25
Metrics
NeMo Gym emits a small, deliberate set of metrics. This page lists them, and explains what is deliberately absent and why.
Evidence use: For understanding GPU computing, training, retrieval, inference serving, reliability or deployment decisions.
Limitations: Directly evidences only the implementation and claims of NVIDIA's technology stack. Recheck versioned interfaces and experimental capabilities before use.
Related concepts: Model Evaluation and Benchmarks · LLM Inference Optimization · Model Deployment Options
Link last checked: 2026-09-25
Rollout Evidence
Collect, join, and inspect model-call and agent evidence.
Evidence use: For understanding GPU computing, training, retrieval, inference serving, reliability or deployment decisions.
Limitations: Directly evidences only the implementation and claims of NVIDIA's technology stack. Recheck versioned interfaces and experimental capabilities before use.
Related concepts: Model Evaluation and Benchmarks · LLM Inference Optimization · Model Deployment Options
Link last checked: 2026-09-25
Agentic retrieval (concept) ¶
NeMo Retriever documentation — library, deployment, and operations guides
Evidence use: For understanding GPU computing, training, retrieval, inference serving, reliability or deployment decisions.
Limitations: Directly evidences only the implementation and claims of NVIDIA's technology stack. Recheck versioned interfaces and experimental capabilities before use.
Related concepts: AI Agent · Tool Calling · LLM Inference Optimization
Link last checked: 2026-09-25
Audio and video ingestion ¶
NeMo Retriever documentation — library, deployment, and operations guides
Evidence use: For understanding GPU computing, training, retrieval, inference serving, reliability or deployment decisions.
Limitations: Directly evidences only the implementation and claims of NVIDIA's technology stack. Recheck versioned interfaces and experimental capabilities before use.
Related concepts: Retrieval-Augmented Generation (RAG) · Embedding · LLM Inference Optimization
Link last checked: 2026-09-25