|
--- |
|
inference: false |
|
language: |
|
- bg |
|
license: mit |
|
tags: |
|
- torch |
|
--- |
|
|
|
# LLaMA-7B |
|
|
|
This repo contains a low-rank adapter for LLaMA-7b trained on Bulgarian dataset. |
|
|
|
This model was introduced in [this paper](https://arxiv.org/abs/2106.09685). |
|
|
|
## Model description |
|
|
|
The training data is private Bulgarian dataset. |
|
|
|
## Intended uses & limitations |
|
|
|
This is an instruction-based model similart to ChatGPT but in Bulgarian. |
|
|
|
### How to use |
|
|
|
Here is how to use this model in PyTorch: |
|
|
|
```bash |
|
>>> git clone https://github.com/tloen/alpaca-lora.git |
|
>>> cd alpaca-lora |
|
>>> pip install -r requirements.txt |
|
>>> |
|
>>> python generate.py \ |
|
--load_8bit \ |
|
--base_model 'yahma/llama-7b-hf' \ |
|
--lora_weights 'rmihaylov/alpaca-lora-bg-7b' \ |
|
--share_gradio |
|
``` |
|
|
|
This will download both a base model and an adapter from huggingface. Then it will run a gradio interface for chatting. |
|
|
|
Example using this model: [Colab](https://colab.research.google.com/drive/1IPz8QqOa5ZUBz7ZyXE4hhh7XwMEH-D9S?usp=sharing). You need a Colab Pro because the model needs high ram when loading. |
|
|
|
### Interface |
|
|
|
 |
