|
|
--- |
|
|
title: Neural Network Quantizer |
|
|
emoji: ⚡ |
|
|
colorFrom: indigo |
|
|
colorTo: purple |
|
|
sdk: docker |
|
|
pinned: false |
|
|
license: mit |
|
|
app_port: 7860 |
|
|
--- |
|
|
|
|
|
# Neural Network Weight Quantizer |
|
|
|
|
|
Quantize neural network weights to lower precision formats (INT8, INT4, NF4) with interactive visualizations. |
|
|
|
|
|
## Features |
|
|
|
|
|
- 🔢 Multi-bit quantization (4-bit, 8-bit) |
|
|
- 📊 Interactive weight visualizations |
|
|
- 🤗 HuggingFace model support (optional) |
|
|
- ⚡ GPU acceleration (when available) |
|
|
- 📈 Quantization error analysis |
|
|
- 🔄 Method comparison (INT8 vs INT4 vs NF4) |
|
|
|
|
|
## Quick Start |
|
|
|
|
|
1. Use the **Quantizer** tab to test on random weights |
|
|
2. Compare different methods in the **Analysis** tab |
|
|
3. Optionally load a HuggingFace model in the **Models** tab |
|
|
|
|
|
## API |
|
|
|
|
|
The backend exposes a REST API at `/api`: |
|
|
|
|
|
- `GET /api/system/info` - System capabilities |
|
|
- `POST /api/quantize/weights` - Quantize custom weights |
|
|
- `POST /api/models/load` - Load HuggingFace model |
|
|
- `POST /api/analysis/compare` - Compare methods |
|
|
|
|
|
## 🚀 Deployment |
|
|
|
|
|
### Hugging Face Spaces |
|
|
This project is configured for **Hugging Face Spaces** using the Docker SDK. |
|
|
|
|
|
1. Create a new Space on [Hugging Face](https://huggingface.co/new-space). |
|
|
2. Select **Docker** as the SDK. |
|
|
3. Push this repository to your Space: |
|
|
```bash |
|
|
git remote add space https://huggingface.co/spaces/YOUR_USERNAME/YOUR_SPACE_NAME |
|
|
git push space main |
|
|
``` |
|
|
|
|
|
### Docker |
|
|
Run locally with Docker: |
|
|
```bash |
|
|
docker build -t quantizer . |
|
|
docker run -p 7860:7860 quantizer |
|
|
``` |
|
|
Open `http://localhost:7860`. |
|
|
|
