---
license: llama2
train: false
inference: false
pipeline_tag: text-generation
---

## Llama-2-70b-hf-2bit_g16_s128-HQQ 
This is a version of the LLama-2-70B-hf model quantized to 2-bit via Half-Quadratic Quantization (HQQ): https://mobiusml.github.io/hqq/

This model outperforms an fp16 LLama-2-13B (perplexity 4.13 vs. 4.63) for a comparable ~26GB size. 

To run the model, install the HQQ library from https://github.com/mobiusml/hqq/tree/main/code and use it as follows:
``` Python
from hqq.models.llama import LlamaHQQ
import transformers

model_id = 'mobiuslabsgmbh/Llama-2-70b-hf-2bit_g16_s128-HQQ'
#Load the tokenizer
tokenizer = transformers.AutoTokenizer.from_pretrained(model_id)
#Load the model
model = LlamaHQQ.from_quantized(model_id)
```

*Limitations*: <br>
-Only supports single GPU runtime.<br>
-Not compatible with HuggingFace's PEFT.<br>