cosmos3-codebase-nav

par nvidia

Naviguer dans le codebase du package Cosmos3 pour trouver où se trouvent les paramètres, configurations, valeurs par défaut, scripts et documentation. Utiliser lorsque l'utilisateur demande « où se trouve X dans… »

npx skills add https://github.com/nvidia/cosmos-framework --skill cosmos3-codebase-nav

Cosmos3 Codebase Navigation

When to use this skill

  • Use this skill when an agent is navigating the Cosmos3 package
  • Use this skill to answer "where is X", "how do I find the config for Y", or any file-location question
  • Use this skill when the user opens or edits cosmos3 files and needs orientation

Path convention

All paths below are relative to this file's location (.agents/skills/cosmos3-codebase-nav/). The repo is laid out as:

  • cosmos_framework/ — main training package (data, model, trainer, callbacks, checkpoint, utils, …).
  • cosmos_framework/configs/base/experiment/ — vfm (generator) experiment SKUs referenced by [train.train_policy].experiment in the recipe TOMLs.
  • cosmos_framework/configs/base/reasoner/experiment/ — vlm (reasoner) experiment SKUs.
  • cosmos_framework/inference/ — inference subpackage (args, model, inference engine, defaults, Ray serving, common helpers).
  • cosmos_framework/scripts/ — top-level entry-point scripts (train, inference, eval, export_model, convert_model_to_dcp, upsample_prompts, caption_from_video, captions_to_sft_jsonl, action_policy_server, …). Invoked as python -m cosmos_framework.scripts.<name>.
  • examples/toml/sft_config/<recipe>.toml + examples/launch_sft_<recipe>.sh — paired SFT recipes (training entry-point input). The shell sources examples/_sft_launcher_common.sh, which forwards into cosmos_framework.scripts.train --sft-toml=....
  • cosmos_framework/configs/toml_config/ — pydantic schemas (sft_config.py) and helpers that validate the recipe TOML at load time.

Quick Reference

Where parameters and defaults live

What you're looking forFile
Sampling params (num_steps, guidance, shift, fps, etc.)../../../cosmos_framework/inference/args.pySamplingArgs, SamplingOverrides
Per-modality default values../../../cosmos_framework/inference/defaults/<mode>/sample_args.json
Setup params (parallelism, checkpoints, model path)../../../cosmos_framework/inference/args.pyOmniSetupArgs, OmniSetupOverrides
Common args base classes../../../cosmos_framework/inference/common/args.pyArgsBase, OverridesBase
Ray serving parallelism presets../../../cosmos_framework/inference/ray/configs/latency.yaml, ../../../cosmos_framework/inference/ray/configs/throughput.yaml
Feature flags../../../cosmos_framework/utils/flags.py
Prompt upsampler system prompt../../../cosmos_framework/inference/defaults/prompt_upsampler.txt
Video captioner system prompt../../../cosmos_framework/inference/defaults/video_captioner.txt
SFT recipe TOMLs (paired with examples/launch_sft_*.sh)../../../examples/toml/sft_config/<recipe>.toml
SFT pydantic schema (validates the recipe TOML)../../../cosmos_framework/configs/toml_config/sft_config.py
Training experiment SKUs (vfm)../../../cosmos_framework/configs/base/experiment/
Training experiment SKUs (vlm / reasoner)../../../cosmos_framework/configs/base/reasoner/experiment/
Example inputs../../../inputs/omni/t2i.json, ../../../inputs/omni/t2v.json, ../../../inputs/omni/i2v.json, …

Available modality modes for defaults: text2image, text2video, image2video, image2image, video2video, forward_dynamics, inverse_dynamics, wam.

Config defaults resolution chain

When a user runs inference, default parameter values are resolved in this order:

cosmos_framework/inference/defaults/<mode>/sample_args.json     # 1. Per-modality JSON defaults (num_steps, guidance, shift, fps, etc.)
        ↓
_load_modality_defaults() in cosmos_framework/inference/args.py # 2. Loaded and cached at import time
        ↓
SamplingArgs / SamplingOverrides                      # 3. Pydantic models with field-level validation
        ↓
OmniSampleOverrides.build_sample()                    # 4. Merges user overrides → final resolved args
        ↓
_RESOLUTION_SHIFT_DEFAULTS[model_size, resolution]    # 5. Model+resolution shift override (if user didn't set shift)
        ↓
CLI flags (--guidance, --shift, etc.)                 # 6. User overrides from command line

The _RESOLUTION_SHIFT_DEFAULTS table in ../../../cosmos_framework/inference/args.py (on OmniSampleOverrides) overrides the default shift based on model size and resolution, unless the user explicitly specified --shift.

ModeDefault fileKey defaults
text2image../../../cosmos_framework/inference/defaults/text2image/sample_args.jsonnum_frames=1, guidance=6.0, shift=10.0
text2video../../../cosmos_framework/inference/defaults/text2video/sample_args.jsonnum_frames=189, guidance=6.0, shift=10.0
image2video../../../cosmos_framework/inference/defaults/image2video/sample_args.jsonnum_frames=189, guidance=6.0, shift=10.0

Action and video2video modes also have defaults under cosmos_framework/inference/defaults/{image2image,video2video,forward_dynamics,inverse_dynamics,policy}/sample_args.json.

Users can also supply a custom defaults file per-request via the defaults_file field in sample arguments (see ../../../docs/inference.md).

Where to make changes

TaskEdit
Change a built-in default value../../../cosmos_framework/inference/defaults/<mode>/sample_args.json
Add a new CLI parameterSamplingArgs + SamplingOverrides in ../../../cosmos_framework/inference/args.py, then add to each sample_args.json
Change parallelism presets../../../cosmos_framework/inference/ray/configs/latency.yaml or throughput.yaml
Add a new script../../../cosmos_framework/scripts/ — follow inference.py as the pattern

Key entry points

Entry pointHow to run
Batch inferencepython -m cosmos_framework.scripts.inference
Trainingpython -m cosmos_framework.scripts.train --sft-toml=examples/toml/sft_config/<recipe>.toml
Online serving (Ray)python -m cosmos_framework.inference.ray.serve
Submit to Ray serverpython -m cosmos_framework.inference.ray.submit
Gradio UIpython -m cosmos_framework.inference.ray.gradio
Prompt upsamplingpython -m cosmos_framework.scripts.upsample_prompts
Model export (HF)python -m cosmos_framework.scripts.export_model
DCP conversionpython -m cosmos_framework.scripts.convert_model_to_dcp
Diffusers conversionpython -m cosmos_framework.scripts.convert_model_to_diffusers
Video captioningpython -m cosmos_framework.scripts.caption_from_video
Captions → SFT JSONLpython -m cosmos_framework.scripts.captions_to_sft_jsonl
Action policy server (LIBERO HTTP)python -m cosmos_framework.scripts.action_policy_server_libero
Action policy server (RoboLab WS)python -m cosmos_framework.scripts.action_policy_server_robolab

Documentation

DocCovers
../../../AGENTS.mdCommands, rules, key file locations (read this first)
../../../README.mdOverview, quickstart, examples
../../../docs/setup.mdInstallation, environment, checkpoints
../../../docs/code_structure.mdRepo layout and per-subpackage tour of cosmos_framework/
../../../docs/inference.mdSample args, default values, custom defaults
../../../docs/training.mdSFT / post-training workflow
../../../docs/faq.mdFAQ, tips, and troubleshooting

Plus de skills de nvidia

compileiq-debug
nvidia
Utilisez quand quelque chose ne va pas : Search() bloque, toutes les évaluations retournent INVALID_SCORE, les scores ne s'améliorent pas, chaque configuration retourne le même nombre, erreurs ptxas…
create-github-pr
nvidia
Créer des pull requests GitHub en utilisant l'interface en ligne de commande gh. Utiliser lorsque l'utilisateur souhaite créer une nouvelle PR, soumettre du code pour révision, ou ouvrir une pull request. Mots-clés de déclenchement -…
nemoclaw-maintainer-cross-issue-sweep
nvidia
Analyse les autres problèmes ouverts pour trouver ceux qu’une PR donnée pourrait également corriger ou casser accidentellement. Génère des opportunités de correctifs adjacents et des risques de contradiction avec fichier:ligne…
fhir-basics
nvidia
Apprend aux agents comment fonctionnent les API FHIR R4, quelles ressources sont disponibles, comment les interroger avec des paramètres de recherche, et comment analyser correctement tous les formats de réponse…
compileiq-validate-result
nvidia
Utiliser APRÈS qu'une recherche soit terminée et AVANT de réclamer un accélérateur ou d'expédier un ACF. Charge le CSV dump_results, extrait les K meilleurs candidats (mono-objectif)…
changelog-audit
nvidia
Auditer le CHANGELOG.md de Warp avant une publication : récupérer les entrées perdues, trier par impact utilisateur, affiner le langage des entrées, ajuster les retours à la ligne et (en mode branche de publication) mettre à jour la comparaison…
maintain-dynamic-plugins
nvidia
Maintenir les chargeurs de plugins dynamiques NeMo Relay, les manifestes, les SDK natifs Rust, le protocole worker gRPC, le SDK worker Python, la documentation, les tests et la couverture du workflow de publication
dgx-diagnose
nvidia
Diagnostiquer les problèmes courants du DGX Station GB300 — plantages CUDA, ciblage incorrect du GPU, bugs de conteneur vLLM/SGLang, problèmes d'état MIG, erreurs NVLink/Fabric Manager,…