skills.rest
yo-steven
yo-stevenCOMMUNITY
Explore all skills by this author →
GitHub

nemo-gym-reward-profiling

Join materialized inputs with rollout results for Nemo Gym reward profiling.

Updated May 23, 2026
Search Tags:#token usage#nemo-gym#reward profiling#jsonl artifacts#rollout analysis#repeated rollouts#task identity
One-click install
npx skills add https://github.com/yo-steven/skills-exploration-20260522 --skill nemo-gym-reward-profiling
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: nemo-gym-reward-profiling
Source: https://github.com/yo-steven/skills-exploration-20260522/tree/main/skills/NeMo-Gym/nemo-gym-reward-profiling
Command: npx skills add https://github.com/yo-steven/skills-exploration-20260522 --skill nemo-gym-reward-profiling

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

Reward profiling in Nemo Gym can be hard to run correctly and interpret when repeats, materialized inputs, and per-rollout results must be joined by task and rollout identity.

Core Features & Use Cases

  • Orchestrates the standard profiling flow: use ng_run to start servers, ng_collect_rollouts to produce rollouts.jsonl and materialized inputs, and ng_reward_profile to generate reward profiling outputs.
  • Explains how identities and artifacts connect: clarifies how _ng_task_index and _ng_rollout_index link *_materialized_inputs.jsonl, rollouts.jsonl, and *_reward_profiling.jsonl.
  • Supports practical profiling scenarios: repeated rollouts for per-task averages and variance, strict vs partial profiling (including allow_partial_rollouts), and inspecting rollout_infos for reward and token metrics.

Quick Start

Run ng_run to start the Nemo Gym services, then ng_collect_rollouts to write rollouts.jsonl and *_materialized_inputs.jsonl, and finally ng_reward_profile to produce *_reward_profiling.jsonl from the materialized inputs and rollout artifacts.

Frequently Asked Questions about nemo-gym-reward-profiling

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I profile Nemo Gym rewards across repeated rollouts?▼

Nemo Gym reward profiling requires running ng_run, ng_collect_rollouts, and ng_reward_profile sequentially to generate JSONL artifacts and per-task summaries from repeated rollouts.

How do materialized inputs and rollout results connect during reward profiling?▼

Materialized inputs and rollout results connect via _ng_task_index and _ng_rollout_index identity keys, linking *_materialized_inputs.jsonl, rollouts.jsonl, and *_reward_profiling.jsonl artifacts.

What does allow_partial_rollouts do in Nemo Gym reward profiling?▼

The allow_partial_rollouts setting controls strict versus partial profiling, allowing reward profiling to proceed even when incomplete rollout data is present across repeated task executions.

How can I analyze per-task reward averages and variance from rollouts?▼

Per-task reward averages and variance are analyzed by inspecting rollout_infos and completion metrics within the reward profiling outputs, keyed by _ng_task_index across repeated rollouts.

Why are my Nemo Gym reward profiling artifacts not aligning with task inputs?▼

Reward profiling artifacts fail to align when task and rollout identity keys like _ng_task_index and _ng_rollout_index are not consistently matched between *_materialized_inputs.jsonl and rollouts.jsonl.

Related Skills

Explore More Data & Analytics Skills→
yo-stevenyo-steven

add-benchmark

Integrate new NeMo-Gym benchmarks with JSONL data and verify() rewards.

Community
Advanced
NeuroSkill-comNeuroSkill-com
11

neuroskill-sessions

Summarize per-session EXG metrics with trends and JSON or HTTP outputs.

Official
Advanced
NVIDIANVIDIA
2.8k

add-benchmark

Scaffold NeMo-Gym benchmarks with resources servers, data pipelines, and YAML configs.

Official
Advanced
yo-stevenyo-steven

nemo-gym-debugging

Classify Nemo Gym rollout failures and guide evidence-first debugging.

Community
Advanced