Model Details
This model is an int4 model with group_size 128 of Qwen/Qwen2.5-72B-Instruct generated by intel/auto-round, auto-round is needed to run this model
How To Use
INT4 Inference
##git clone https://github.com/intel/auto-round.git
##cd auto-round && pip install -vvv --no-build-isolation -e .
from auto_round import AutoHfQuantizer ##must import
import torch
from transformers import AutoModelForCausalLM,AutoTokenizer
quantized_model_dir = "Intel/Qwen2.5-72B-Instruct-int4-inc"
tokenizer = AutoTokenizer.from_pretrained(quantized_model_dir)
model = AutoModelForCausalLM.from_pretrained(
quantized_model_dir,
torch_dtype='auto',
device_map="auto",
)
##import habana_frameworks.torch.core as htcore ## uncommnet it for HPU
##import habana_frameworks.torch.hpu as hthpu ## uncommnet it for HPU
##model = model.to(torch.bfloat16).to("hpu") ## uncommnet it for HPU
prompt = "There is a girl who likes adventure,"
messages = [
{"role": "system", "content": "You are a helpful assistant."},
{"role": "user", "content": prompt}
]
text = tokenizer.apply_chat_template(
messages,
tokenize=False,
add_generation_prompt=True
)
model_inputs = tokenizer([text], return_tensors="pt").to(model.device)
generated_ids = model.generate(
model_inputs.input_ids,
max_new_tokens=50, ##change this to align with the official usage
do_sample=False ##change this to align with the official usage
)
generated_ids = [
output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
]
response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
print(response)
##prompt = "There is a girl who likes adventure,
##That sounds like a wonderful trait! A girl who enjoys adventure likely has a spirit of curiosity, bravery, and a willingness to explore the unknown. Whether it's trying new activities, traveling to new places, or simply seeking out new experiences, her love
##prompt = "Which one is bigger, 9.11 or 9.8"
##To determine which number is bigger between 9.11 and 9.8, you can compare them digit by digit:
##1. Compare the whole number parts:
## - Both numbers have the same whole number part, which is 9.
##prompt = "Once upon a time,"
##Once upon a time, in a land far, far away, there was a kingdom known for its beauty and prosperity. The kingdom was ruled by a wise and just king who loved his people dearly. In the heart of the kingdom stood a magnificent castle
##prompt = "请介绍一下阿里巴巴公司"
##阿里巴巴集团是一家中国跨国科技公司，成立于1999年，总部位于中国杭州。阿里巴巴的业务涵盖了电子商务、零售、金融、物流、云计算等多个领域，旗下拥有包括淘宝网、天猫、菜鸟网络、阿里云等
Evaluate the model
pip3 install lm-eval==0.4.4
git clone https://github.com/intel/auto-round
cd auto-round
python -m auto_round --model "Intel/Qwen2.5-72B-Instruct-int4-inc" --eval --eval_bs 16 --tasks lambada_openai,hellaswag,piqa,winogrande,truthfulqa_mc1,openbookqa,boolq,arc_easy,arc_challenge,mmlu,gsm8k,cmmlu,ceval-valid
| Metric | BF16 | INT4 |
|---|---|---|
| Avg | 0.7582 | 0.7567 |
| mmlu | 0.8336 | 0.8306 |
| cmmlu | 0.8722 | 0.8638 |
| ceval-valid | 0.8982 | 0.8938 |
| lambada_openai | 0.7518 | 0.7603 |
| hellaswag | 0.7040 | 0.6970 |
| winogrande | 0.7577 | 0.7695 |
| piqa | 0.8335 | 0.8270 |
| truthfulqa_mc1 | 0.5288 | 0.5202 |
| openbookqa | 0.3860 | 0.3900 |
| boolq | 0.9046 | 0.9080 |
| arc_easy | 0.8603 | 0.8577 |
| arc_challenge | 0.6169 | 0.6109 |
| gsm8k 5 shots | 0.9090 | 0.9083 |
Reproduce the model
Here is the sample command to reproduce the model. We observed a larger accuracy drop in Chinese tasks and recommend using a high-quality Chinese dataset for calibration. However, we did not achieve better accuracy with some public datasets.
git clone https://github.com/intel/auto-round
cd auto-round
python -m auto_round \
--model_name Qwen/Qwen2.5-72B-Instruct \
--device 0 \
--group_size 128 \
--nsamples 512 \
--bits 4 \
--iter 1000 \
--disable_eval \
--model_dtype "float16" \
--format 'auto_round' \
--output_dir "./tmp_autoround"
Ethical Considerations and Limitations
The model can produce factually incorrect output, and should not be relied on to produce factually accurate information. Because of the limitations of the pretrained model and the finetuning datasets, it is possible that this model could generate lewd, biased or otherwise offensive outputs.
Therefore, before deploying any applications of the model, developers should perform safety testing.
Caveats and Recommendations
Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model.
Here are a couple of useful links to learn more about Intel's AI software:
Disclaimer
The license on this model does not constitute legal advice. We are not responsible for the actions of third parties who use this model. Please consult an attorney before using this model for commercial purposes.
Cite
@article{cheng2023optimize, title={Optimize weight rounding via signed gradient descent for the quantization of llms}, author={Cheng, Wenhua and Zhang, Weiwei and Shen, Haihao and Cai, Yiyang and He, Xin and Lv, Kaokao and Liu, Yi}, journal={arXiv preprint arXiv:2309.05516}, year={2023} }