Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,92 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: other
|
| 3 |
+
license_name: ltx-video-license
|
| 4 |
+
license_link: https://huggingface.co/Lightricks/LTX-2/blob/main/LICENSE.md
|
| 5 |
+
library_name: mlx
|
| 6 |
+
tags:
|
| 7 |
+
- video-generation
|
| 8 |
+
- mlx
|
| 9 |
+
- apple-silicon
|
| 10 |
+
- ltx-2
|
| 11 |
+
- dev
|
| 12 |
+
- quantized
|
| 13 |
+
---
|
| 14 |
+
|
| 15 |
+
# LTX-2 19B Dev (4-bit) - MLX
|
| 16 |
+
|
| 17 |
+
This is a 4-bit quantized version of the [LTX-2 19B Dev](https://huggingface.co/Lightricks/LTX-2) model, optimized for Apple Silicon using MLX.
|
| 18 |
+
|
| 19 |
+
## Model Description
|
| 20 |
+
|
| 21 |
+
LTX-2 is a state-of-the-art video generation model from Lightricks. This version has been quantized to 4-bit precision for efficient inference on Apple Silicon devices with MLX.
|
| 22 |
+
|
| 23 |
+
### Key Features
|
| 24 |
+
|
| 25 |
+
- **Pipeline**: Dev (full control with CFG scale)
|
| 26 |
+
- **Quantization**: 4-bit precision
|
| 27 |
+
- **Framework**: MLX (Apple Silicon optimized)
|
| 28 |
+
- **Memory**: ~19GB VRAM required
|
| 29 |
+
|
| 30 |
+
## Usage
|
| 31 |
+
|
| 32 |
+
### Installation
|
| 33 |
+
|
| 34 |
+
```bash
|
| 35 |
+
pip install git+https://github.com/CharafChnioune/mlx-video.git
|
| 36 |
+
```
|
| 37 |
+
|
| 38 |
+
### Command Line
|
| 39 |
+
|
| 40 |
+
```bash
|
| 41 |
+
# Basic generation
|
| 42 |
+
mlx-video --prompt "A beautiful sunset over the ocean" \
|
| 43 |
+
--model-repo AITRADER/ltx2-dev-4bit-mlx \
|
| 44 |
+
--pipeline dev \
|
| 45 |
+
--height 512 --width 512 \
|
| 46 |
+
--num-frames 33
|
| 47 |
+
|
| 48 |
+
# Dev pipeline with CFG
|
| 49 |
+
mlx-video --prompt 'A cat playing with yarn' \\
|
| 50 |
+
--model-repo AITRADER/ltx2-dev-4bit-mlx \\
|
| 51 |
+
--pipeline dev \\
|
| 52 |
+
--steps 40 --cfg-scale 4.0
|
| 53 |
+
```
|
| 54 |
+
|
| 55 |
+
### Python API
|
| 56 |
+
|
| 57 |
+
```python
|
| 58 |
+
from mlx_video import generate_video
|
| 59 |
+
|
| 60 |
+
video = generate_video(
|
| 61 |
+
prompt="A beautiful sunset over the ocean",
|
| 62 |
+
model_repo="AITRADER/ltx2-dev-4bit-mlx",
|
| 63 |
+
pipeline="dev",
|
| 64 |
+
height=512,
|
| 65 |
+
width=512,
|
| 66 |
+
num_frames=33,
|
| 67 |
+
)
|
| 68 |
+
```
|
| 69 |
+
|
| 70 |
+
## Model Files
|
| 71 |
+
|
| 72 |
+
- `ltx-2-19b-dev-mlx.safetensors` - Main model weights (4-bit quantized)
|
| 73 |
+
- `quantization.json` - Quantization configuration
|
| 74 |
+
- `config.json` - Model configuration
|
| 75 |
+
- `layer_report.json` - Layer information
|
| 76 |
+
|
| 77 |
+
## Performance
|
| 78 |
+
|
| 79 |
+
| Resolution | Frames | Steps |
|
| 80 |
+
|------------|--------|------|
|
| 81 |
+
| 512x512 | 33 | ~40 steps |
|
| 82 |
+
| 768x512 | 33 | ~40 steps |
|
| 83 |
+
|
| 84 |
+
## License
|
| 85 |
+
|
| 86 |
+
This model is released under the [LTX Video License](https://huggingface.co/Lightricks/LTX-2/blob/main/LICENSE.md).
|
| 87 |
+
|
| 88 |
+
## Acknowledgements
|
| 89 |
+
|
| 90 |
+
- [Lightricks](https://lightricks.com/) for the original LTX-2 model
|
| 91 |
+
- [MLX](https://github.com/ml-explore/mlx) team at Apple for the framework
|
| 92 |
+
- [mlx-video](https://github.com/CharafChnioune/mlx-video) for the MLX conversion
|