---
metrics: null
---

Quantized Meta AI's [LLaMA](https://arxiv.org/abs/2302.13971) in 4bit with the help of [GPTQ](https://arxiv.org/abs/2210.17323v2) algorithm v2.

- [**llama13b-4bit-ts-ao-g128-v2.safetensors**](https://huggingface.co/sardukar/llama13b-4bit-v2/blob/main/llama13b-4bit-ts-ao-g128-v2.safetensors)
GPTQ implementation - https://github.com/qwopqwop200/GPTQ-for-LLaMa/tree/49efe0b67db4b40eac2ae963819ebc055da64074

Conversion process:
```sh
CUDA_VISIBLE_DEVICES=0 python llama.py ./llama-13b c4 --wbits 4 --true-sequential --act-order --groupsize 128 --save_safetensors ./q4/llama13b-4bit-ts-ao-g128-v2.safetensors
```


- [llama13b-4bit-v2.safetensors](https://huggingface.co/sardukar/llama13b-4bit-v2/blob/main/llama13b-4bit-v2.safetensors)
GPTQ implementation - https://github.com/qwopqwop200/GPTQ-for-LLaMa/tree/841feedde876785bc8022ca48fd9c3ff626587e2

**Note:** This model will fail to load with current GPTQ-for-LLaMa implementation

Conversion process
```sh
CUDA_VISIBLE_DEVICES=0 python llama.py ./llama-13b c4 --wbits 4 --true-sequential --act-order  --save_safetensors ./q4/llama13b-4bit-v2.safetensors
```