README.md · abhinavkulkarni/lmsys-vicuna-33b-v1.3-w4-g128-awq at 619ba25f890fa9f17c62c5e0095aff82972ce3f9

metadata

inference: false

Vicuña 33b v1.3 (4-bit 128g AWQ Quantized)

Vicuna is a chat assistant trained by fine-tuning LLaMA on user-shared conversations collected from ShareGPT.

Developed by: LMSYS
Model type: An auto-regressive language model based on the transformer architecture.
License: Non-commercial license
Finetuned from model: LLaMA.

This model is a 4-bit 128 group size AWQ quantized model. For more information about AWQ quantization, please click here.

Model Date

July 14, 2023

Model License

Please refer to original Vicuna model license (link).

Please refer to the AWQ quantization license (link).

CUDA Version

This model was successfully tested on CUDA driver v530.30.02 and runtime v11.7 with Python v3.10.11. Please note that AWQ requires NVIDIA GPUs with compute capability of 8.0 or higher.

For Docker users, the nvcr.io/nvidia/pytorch:23.06-py3 image is runtime v12.1 but otherwise the same as the configuration above and has also been verified to work.

How to Use

git clone https://github.com/abhinavkulkarni/llm-awq \
&& cd llm-awq \
&& git checkout ba01560f21516805fc5ceba5c2566dcbd1cf66d8 \
&& pip install -e . \
&& cd awq/kernels \
&& python setup.py install

import torch
from awq.quantize.quantizer import real_quantize_model_weight
from transformers import AutoModelForCausalLM, AutoConfig, AutoTokenizer, TextStreamer
from accelerate import init_empty_weights, load_checkpoint_and_dispatch
from huggingface_hub import snapshot_download

model_name = "abhinavkulkarni/lmsys-vicuna-33b-v1.3-w4-g128-awq"

# Config
config = AutoConfig.from_pretrained(model_name, trust_remote_code=True)

# Tokenizer
tokenizer = AutoTokenizer.from_pretrained(config.tokenizer_name)
streamer = TextStreamer(tokenizer, skip_special_tokens=True)

# Model
w_bit = 4
q_config = {
    "zero_point": True,
    "q_group_size": 128,
}

load_quant = snapshot_download(model_name)

with init_empty_weights():
    model = AutoModelForCausalLM.from_config(config=config, 
                                                 torch_dtype=torch.float16, trust_remote_code=True)

real_quantize_model_weight(model, w_bit=w_bit, q_config=q_config, init_only=True)

model = load_checkpoint_and_dispatch(model, load_quant, device_map="balanced")

# Inference
prompt = f'''What is the difference between nuclear fusion and fission?
###Response:'''

input_ids = tokenizer(prompt, return_tensors='pt').input_ids.cuda()
output = model.generate(
    inputs=input_ids, 
    temperature=0.7,
    max_new_tokens=512,
    top_p=0.15,
    top_k=0,
    repetition_penalty=1.1,
    eos_token_id=tokenizer.eos_token_id,
    streamer=streamer)

Evaluation

This evaluation was done using LM-Eval.

vicuna-33b-v1.3

Task	Version	Metric	Value
wikitext	1	word_perplexity	9.8210
		byte_perplexity	1.5330
		bits_per_byte	0.6163

vicuna-33b-v1.3 (4-bit 128-group AWQ)

Task	Version	Metric	Value
wikitext	1	word_perplexity	9.9924
		byte_perplexity	1.5380
		bits_per_byte	0.6210

Acknowledgements

The model was quantized with AWQ technique. If you find AWQ useful or relevant to your research, please kindly cite the paper:

@article{lin2023awq,
  title={AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration},
  author={Lin, Ji and Tang, Jiaming and Tang, Haotian and Yang, Shang and Dang, Xingyu and Han, Song},
  journal={arXiv},
  year={2023}
}