metadata

license: apache-2.0
tags:
  - AWQ
inference: false

Falcon-7B-Instruct (4-bit 64g AWQ Quantized)

Falcon-7B-Instruct is a 7B parameters causal decoder-only model built by TII based on Falcon-7B and finetuned on a mixture of chat/instruct datasets.

This model is a 4-bit 64 group size AWQ quantized model. For more information about AWQ quantization, please click here.

Model Date

July 5, 2023

Model License

Please refer to original Falcon 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 80 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/mit-han-lab/llm-awq \
&& cd llm-awq \
&& git checkout 71d8e68df78de6c0c817b029a568c064bf22132d \
&& pip install -e . \
&& cd awq/kernels \
&& export TORCH_CUDA_ARCH_LIST='8.0 8.6 8.7 8.9 9.0' \
&& python setup.py install

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

model_name = "tiiuae/falcon-7b-instruct"

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

# Tokenizer
tokenizer = AutoTokenizer.from_pretrained(model_name)

# Model
w_bit = 4
q_config = {
    "zero_point": True,
    "q_group_size": 64,
}

load_quant = hf_hub_download('abhinavkulkarni/falcon-7b-instruct-w4-g64-awq', 'pytorch_model.bin')

with init_empty_weights():
    model = AutoModelForCausalLM.from_pretrained(model_name, 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
)
print(tokenizer.decode(output[0], skip_special_tokens=True))

Evaluation

This evaluation was done using LM-Eval.

Falcon-7B-Instruct

Task	Version	Metric	Value
wikitext	1	word_perplexity	14.5069
		byte_perplexity	1.6490
		bits_per_byte	0.7216

Falcon-7B-Instruct (4-bit 64-group AWQ)

Task	Version	Metric	Value
wikitext	1	word_perplexity	14.8667
		byte_perplexity	1.6566
		bits_per_byte	0.7282

Acknowledgements

Paper coming soon 😊. In the meanwhile, you can use the following information to cite:

@article{falcon40b,
  title={{Falcon-40B}: an open large language model with state-of-the-art performance},
  author={Almazrouei, Ebtesam and Alobeidli, Hamza and Alshamsi, Abdulaziz and Cappelli, Alessandro and Cojocaru, Ruxandra and Debbah, Merouane and Goffinet, Etienne and Heslow, Daniel and Launay, Julien and Malartic, Quentin and Noune, Badreddine and Pannier, Baptiste and Penedo, Guilherme},
  year={2023}
}

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}
}