README.md

---
language:
- code
pipeline_tag: text-generation
tags:
- llama-2
- mlx
license: llama2
library_name: mlx
base_model: codellama/CodeLlama-7b-Instruct-hf
---

# mlx-community/CodeLlama-7b-Instruct-hf-4bit-mlx-2

As opposed to [mlx-community/CodeLlama-7b-Instruct-hf-4bit-MLX](https://huggingface.co/mlx-community/CodeLlama-7b-Instruct-hf-4bit-MLX), this
one is converted from the base model using a newer MLX-LM version which uses newer and improved quantization and generates model
in a more standard format. For context see [Issue#130](https://github.com/ml-explore/mlx-lm/issues/130) and [PR#114](https://github.com/ml-explore/mlx-lm/pull/114)
of the [MLX-LM](https://github.com/ml-explore/mlx-lm) repo.

This model [mlx-community/CodeLlama-7b-Instruct-hf-4bit-mlx-2](https://huggingface.co/mlx-community/CodeLlama-7b-Instruct-hf-4bit-mlx-2) was
converted to MLX format from [codellama/CodeLlama-7b-Instruct-hf](https://huggingface.co/codellama/CodeLlama-7b-Instruct-hf)
using mlx-lm version **0.23.2**.

## Use with mlx

```bash
pip install mlx-lm
```

```python
from mlx_lm import load, generate

model, tokenizer = load("mlx-community/CodeLlama-7b-Instruct-hf-4bit-mlx-2")

prompt = "hello"

if tokenizer.chat_template is not None:
    messages = [{"role": "user", "content": prompt}]
    prompt = tokenizer.apply_chat_template(
        messages, add_generation_prompt=True
    )

response = generate(model, tokenizer, prompt=prompt, verbose=True)
```