Find matching videos with text or other videos.
Fine-tunes, evaluates, runs, and exports Cosmos-Embed1 models for finding videos by text or video similarity.
When to use it
Use when asked to fine-tune, run inference with, export, or embed and search videos using Cosmos-Embed1.
Give it videos, metadata, and the requested model operation; it runs Cosmos-Embed1 and saves checkpoints, evaluation metrics, search results, or exported models locally.
What you provide
Not yet reviewed.
Requires Docker to pull and run the published Cosmos-Embed container.
Requires NVIDIA Container Toolkit for the GPU-enabled container workflow.
This skill
Hugging Face
Downloads pretrained model components
nvcr.io/nvidia/tao/tao-toolkit:7.3.0-cosmos-embed
Pulls the Cosmos-Embed container