File size: 2,538 Bytes
f81612c
c459035
 
 
 
 
 
 
 
 
 
f81612c
c459035
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
---
license: mit
tags:
  - pytorch
  - safetensors
  - 3d
  - rigging
  - animation
  - skeleton
  - skinning
base_model: VAST-AI/UniRig
---

# UniRig Models (Safetensors)

This repository contains the UniRig model weights converted to safetensors format for safer and faster loading.

## Attribution

These models are converted from the original [VAST-AI/UniRig](https://huggingface.co/VAST-AI/UniRig) repository by [VAST-AI-Research](https://github.com/VAST-AI-Research).

**Original Work:**
- **Paper:** [UniRig: A Unified Framework for 3D Character Rigging](https://arxiv.org/abs/2405.02986)
- **GitHub:** [VAST-AI-Research/UniRig](https://github.com/VAST-AI-Research/UniRig)
- **Authors:** VAST AI Research Team

All credit for the model architecture and training goes to the original authors.

## Models

| Model | Description | Size |
|-------|-------------|------|
| `skeleton.safetensors` | Skeleton prediction model (articulation-xl_quantization_256) | ~1.4 GB |
| `skin.safetensors` | Skinning weights model (articulation-xl) | ~4.1 GB |

## Why Safetensors?

- **Safer**: No arbitrary code execution risk (no pickle)
- **Faster**: Memory-mapped loading for quick startup
- **Portable**: Standard format supported across frameworks

## Usage

### Loading with safetensors

```python
from safetensors.torch import load_file

# Load skeleton model
skeleton_weights = load_file("skeleton.safetensors")

# Load skin model
skin_weights = load_file("skin.safetensors")
```

### Using with UniRig

These weights are compatible with the original UniRig codebase. Replace the `.ckpt` loading code:

```python
# Original (ckpt)
# checkpoint = torch.load("model.ckpt")
# state_dict = checkpoint['state_dict']

# With safetensors
from safetensors.torch import load_file
state_dict = load_file("model.safetensors")
```

## Conversion Details

Converted using the following process:

```python
import torch
from safetensors.torch import save_file

checkpoint = torch.load("model.ckpt", map_location='cpu', weights_only=False)
state_dict = checkpoint['state_dict']
save_file(state_dict, "model.safetensors")
```

The conversion preserves all tensor values, shapes, and dtypes exactly.

## License

This work is licensed under the MIT License, following the original UniRig license.

## Citation

If you use these models, please cite the original work:

```bibtex
@article{unirig2024,
  title={UniRig: A Unified Framework for 3D Character Rigging},
  author={VAST AI Research},
  journal={arXiv preprint arXiv:2405.02986},
  year={2024}
}
```