Individual skill
tao-finetune-cosmos-reason
Shared Cosmos3 frontend that explicitly routes Cosmos Framework and Cosmos-RL, validates runtime model/video-dataset/SLURM inputs, consumes an SQSH or packaged backend image, optionally plans explicit clean source builds, prepares checkpoints, validates the first update in-process, and returns token-weighted losses and task-aware accuracy.
skills/tao-finetune-cosmos-reason/SKILL.md
Source commit 25340619336e89bcc40365c28bfe9e790430468e
What was checked
- Matches the observed repository HEAD
- Yes
- Scanner rules are current
- No
- Repository scan complete
- No
- Skill scan complete
- No
- Observed behavior
- Not checked yet
- Last checked
- Nov 2, 2026
These are observations at the recorded commit, not a guarantee of safety. Review the source before running it.
Install this skill
Choose an agent in your Stack or copy a command below. Commands fetch the pinned commit and copy only this skill directory. Existing destinations are refused. Copying a command does not install anything.
An install command is unavailable: current rules, complete repository and skill scans, an eligible behavior level and a permitted license are required.
Repository context
- Repository stars (not skill ratings)
- 3,892
- Repository last updated
- Nov 6, 2026
- License
- Apache-2.0
Observed agent integration: No specific integration evidence. Standard SKILL.md installation does not imply verified compatibility.
- agent-skills Repository topic
- agentic-ai Repository topic
- ai-agents Repository topic
- claude-code Repository topic
- codex Repository topic
- coding-agents Repository topic
- cuda Repository topic
- developer-tools Repository topic
- llm Repository topic
- omniverse Repository topic
- model Declared skill tag
- cosmos Declared skill tag
- multimodal Declared skill tag
- training Declared skill tag
- operations Derived category
Files and excerpt
File metadata and permitted, redacted excerpts come from the recorded scan. Source links are pinned to that commit.
Browse files
BENCHMARK.mddoc · 9,849 bytesSKILL.mdcore · 19,862 bytesSKILL.mdcore · 19,862 bytesSKILL.mdcore · 19,862 bytesSKILL.mdcore · 19,862 bytesSKILL.mdcore · 19,862 bytesSKILL.mdcore · 19,862 bytesSKILL.mdcore · 19,862 bytesSKILL.mdcore · 19,862 bytesSKILL.mdcore · 19,862 bytesSKILL.mdcore · 19,862 bytesSKILL.mdcore · 19,862 bytesSKILL.mdcore · 19,862 bytesSKILL.mdcore · 19,862 bytesSKILL.mdcore · 19,862 bytesSKILL.mdcore · 19,862 bytesSKILL.mdcore · 19,862 bytesSKILL.mdcore · 19,862 bytesconfig/skillspector-baseline.yamlasset · 2,087 bytesevals/evals.jsonasset · 8,548 bytesreferences/cosmos-actions-parameters.mddoc · 19,275 bytesreferences/cosmos-automl-deft.mddoc · 5,586 bytesreferences/cosmos-backend-operations.mddoc · 8,083 bytesreferences/cosmos-data-specs.mddoc · 3,617 bytesreferences/cosmos-framework-backend.yamlasset · 7,159 bytesreferences/cosmos-reason-automl.mddoc · 14,534 bytesreferences/cosmos-reason-evaluate.mddoc · 11,903 bytesreferences/cosmos-reason-launch.mddoc · 13,094 bytesreferences/cosmos-reason-parameters.mddoc · 17,243 bytesreferences/cosmos-reason-single-gpu-video.mddoc · 4,603 bytesreferences/cosmos-reproducibility-gates.mddoc · 10,644 bytesreferences/cosmos-rl-backend.yamlasset · 7,920 bytesreferences/demo_datasets.yamlasset · 503 bytesreferences/detailed-guide.mddoc · 645 bytesreferences/skill_info.yamlasset · 19,443 bytesreferences/spec_template_evaluate.yamlasset · 1,548 bytesreferences/spec_template_inference.yamlasset · 240 bytesreferences/spec_template_quantize.yamlasset · 345 bytesreferences/spec_template_train.yamlasset · 1,824 bytesschemas/cosmos-job-metadata.schema.jsonasset · 1,675 bytesschemas/evaluate.schema.jsonasset · 13,048 bytesschemas/inference.schema.jsonasset · 1,721 bytesschemas/manifest.jsonasset · 3,202 bytesschemas/quantize.schema.jsonasset · 2,479 bytesschemas/train.schema.jsonasset · 36,188 bytesschemas/train_request.schema.jsonasset · 10,083 bytesscripts/cosmos_common.pyruns · 42,241 bytesscripts/cosmos_rl_checkpoint_action.pyruns · 10,149 bytesscripts/cosmos_workflow.pyruns · 274,681 bytesscripts/evaluate_video_conversation_single_gpu.pyruns · 7,683 bytesscripts/evaluation_workflow.pyruns · 63,339 bytesscripts/extract_cosmos_metrics.pyruns · 12,147 bytesscripts/framework_checkpoint_action.pyruns · 23,910 bytesscripts/framework_evaluation_image_preflight.pyruns · 6,085 bytesscripts/gate_docker_train_evaluate.pyruns · 20,635 bytesscripts/prepare_cosmos3_vlm_checkpoint.pyruns · 14,413 bytesscripts/train_video_conversation_single_gpu.pyruns · 6,314 bytesskill-card.mddoc · 4,579 bytesskill.oms.sigasset · 14,889 bytestests/test_gate_docker_train_evaluate.pyruns · 6,958 bytestests/test_prepare_cosmos3_vlm_checkpoint.pyruns · 2,849 bytestests/test_runtime_preflight_contract.pyruns · 37,246 bytes
Read excerpt
# Cosmos3 TAO training
Keep one shared model-facing frontend. Backend image fields and contract paths
live under `backend_contracts` in `references/skill_info.yaml`; image literals
are stamped from `versions.yaml`, while referenced backend YAMLs define native
runtime schemas. Never translate between them.
## Mandatory runtime intake
Before planning training, collect all of the following. Do not infer a path
from history, another user, a prior job, an image, or a developer checkout.
- `base_model_path_or_uri`. For a Hugging Face model ID or URL, accept an
optional friendly `base_model_revision` such as a branch or tag. If omitted,