Agent skill · software engineering · nvidia
tao-finetune-cosmos-embed
Cosmos-Embed1 video-text embedding for text-to-video retrieval, video-to-video search, semantic deduplication, and fine-tuning. Use when the user asks to "fine-tune Cosmos-Embed1", "run cosmos-embed inference", "export Cosmos-Embed1", "embed videos", or "search videos with text".
Why this skill is useful
Adds executable scripts for training, evaluating, and exporting video-text embeddings that extend the AI's capabilities in multimodal retrieval.
What it needs
Requires docker, nvidia-container-toolkit installed locally. Requires huggingface account access. About 7k tokens when loaded. Last updated 2026-08-07. 2,816 stars on the source repository.
What this skill does
Cosmos-Embed Standalone install? If this session was not initialized by the TAO skill bank plugin, run the tao-setup skill first (host preflight, credentials, cross-skill discovery). Cosmos-Embed1 is a joint video-text embedder for text-to-video retrieval, video-to-video search, zero-shot/kNN classification, and semantic deduplication. The packaged CLI is cosmos-embed1 and supports train, evaluate, inference, and export. Container image and per-action commands are in references/skillinfo.yaml. Compact starting specs are in references/spectemplate.yaml. Train Action Policy AutoML is not packaged for this model skill because there are no Cosmos-Embed schemas under schemas/. Always use the direct model skill actions for train, evaluate, inference, and export, even when a higher-level request includes automlpolicy: on. Do not route Cosmos-Embed through workflow or AutoML skills until model-specific train schemas and templates are added. Non-train actions such as evaluate, inference, export, and deploy flows stay in this model skill. The per-run automlpolicy override does not change model metadata. Quick Start Use the published Cosmos-Embed container pinned below (also declared in references/skillinfo.yaml). Do not build from the private Cosmos-Embed1 source tree for normal skill use; build from source only when developing the container itself. Expected local workspace layout: Use these Docker options for all actions unless the local Docker/platform skill gives a stricter environment-specific command: The pinned 7.1.0 image downloads the BERT Q-Former component under /usr/local/lib/python3.12/dist-packages/pretrainedcheckpoints, even when Hugging Face cache variables point elsewhere. The dedicated bind mount above must be writable by the container UID selected by the platform skill; create it as the launching user before submission. Keep this mount when applying the platform's non-root UID/GID and writable HOME settings. …
How to use it
Reference it in AdaL, Claude Code, Cursor or any coding agent — nothing to install:
@skills nvidia/tao-finetune-cosmos-embed