|
|
--- |
|
|
license: mit |
|
|
tags: |
|
|
- nlp |
|
|
- text-generation |
|
|
- bitnet |
|
|
- pytorch |
|
|
- tensorflow |
|
|
pipeline_tag: text-generation |
|
|
--- |
|
|
|
|
|
# PlayWeBit (BitNet b1.58-2B-4T) |
|
|
|
|
|
This is an implementation of the `microsoft/bitnet-b1.58-2B-4T` model, optimized for text generation. It supports PyTorch and TensorFlow and can run on CPU or GPU. A custom model class (`custom_bitnet.py`) is included for compatibility with the official Hugging Face `transformers` library. |
|
|
|
|
|
## Model Description |
|
|
- **Architecture**: BitNet b1.58-2B with 4-bit quantization |
|
|
- **Parameters**: 2 billion |
|
|
- **Training Data**: Not disclosed (refer to original Microsoft repository) |
|
|
- **Intended Use**: Text generation, chat applications |
|
|
|
|
|
## Installation |
|
|
Install the Hugging Face `transformers` library: |
|
|
```bash |
|
|
pip install transformers |
