Instructions to use nvidia/Cosmos3-Edge with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Cosmos
How to use nvidia/Cosmos3-Edge with Cosmos:
# No code snippets available yet for this library. # To use this model, check the repository files and the library's documentation. # Want to help? PRs adding snippets are welcome at: # https://github.com/huggingface/huggingface.js
- Notebooks
- Google Colab
- Kaggle
Update I2V, FD, and ID examples for the converted checkpoint
#61
by weishao-nv - opened
Summary
- Update
example_i2v_prompt.jsonto use the 5-second timing intervals0-2,2-4, and4-5. - Regenerate the 20-step I2V examples using the converted checkpoint with both vLLM-Omni and Diffusers.
- Regenerate the forward-dynamics and inverse-dynamics outputs with both runtimes.
- Regenerate the vLLM-Omni inverse-dynamics trajectory images from their corresponding output JSON files.
Validation
- Both I2V videos contain 121 frames at 832x480 and 24 FPS.
- Both stitched forward-dynamics videos contain 32 generated frames at 256x256 and 20 FPS.
- All inverse-dynamics JSON outputs contain 60x9 actions.
weishao-nv changed pull request status to merged