Instructions to use jlweights/SeedVR2-3B-MLX-Serve with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use jlweights/SeedVR2-3B-MLX-Serve with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] huggingface-cli download --local-dir SeedVR2-3B-MLX-Serve jlweights/SeedVR2-3B-MLX-Serve
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Atomic Chat
| { | |
| "model_type": "seedvr2", | |
| "architectures": [ | |
| "NaDiT" | |
| ], | |
| "vid_dim": 2560, | |
| "num_layers": 32, | |
| "mm_layers": 10, | |
| "heads": 20, | |
| "head_dim": 128, | |
| "vid_in_channels": 33, | |
| "vid_out_channels": 16, | |
| "patch_size": [ | |
| 1, | |
| 2, | |
| 2 | |
| ], | |
| "rope_dim": 128, | |
| "window": [ | |
| 4, | |
| 3, | |
| 3 | |
| ], | |
| "vae": { | |
| "latent_channels": 16, | |
| "spatial_downsample_factor": 8, | |
| "temporal_downsample_factor": 4, | |
| "scaling_factor": 0.9152 | |
| }, | |
| "_converted_by": "tests/convert_seedvr2_weights.py" | |
| } |