mobiuslabsgmbh/Llama-2-13b-hf-4bit_g64-HQQ

Llama-2-13b-hf-4bit_g64-HQQ

This is a version of the LLama-2-13B-hf model quantized to 4-bit via Half-Quadratic Quantization (HQQ): https://mobiusml.github.io/hqq_blog/

This model matches the quality of AWQ (same perplexity) but didn't use any data for calibration.

To run the model, install the HQQ library:

#This model is deprecated and requires older versions
pip install hqq==0.1.8
pip install transformers==4.46.0

and use it as follows:

model_id = 'mobiuslabsgmbh/Llama-2-13b-hf-4bit_g64-HQQ'

from hqq.engine.hf import HQQModelForCausalLM, AutoTokenizer
tokenizer = AutoTokenizer.from_pretrained(model_id)
model     = HQQModelForCausalLM.from_quantized(model_id)

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