Qwen2 7B Int4 Inc

🚀 Intel/Qwen2-7B-int4-inc

This is an int4 auto-round model of Qwen/Qwen2-7B, offering efficient inference and quantization capabilities.

🚀 Quick Start

This model is an int4 auto-round model with group_size 128 of Qwen/Qwen2-7B generated by intel/auto-round. If you need AutoGPTQ format, please load the model with revision 07a117c

✨ Features

• Efficient Inference: Supports INT4 inference on both CPU and Intel Gaudi-2.
• Quantization: Generated by intel/auto-round with group_size 128.
• Evaluation: Provides evaluation results on multiple tasks.
• Reproducibility: Allows users to reproduce the model with the provided command.

📦 Installation

• Install the necessary packages for inference and evaluation:
  • For inference: pip install auto-round (cpu needs version > 0.3.1)
  • For evaluation: pip3 install lm-eval==0.4.4,auto-round

💻 Usage Examples

Basic Usage

INT4 Inference

##pip install auto-round (cpu needs version > 0.3.1))
from auto_round import AutoRoundConfig ##must import for auto_round format
from transformers import AutoModelForCausalLM,AutoTokenizer
quantized_model_dir = "Intel/Qwen2-7B-int4-inc"
tokenizer = AutoTokenizer.from_pretrained(quantized_model_dir)
model = AutoModelForCausalLM.from_pretrained(quantized_model_dir,
                                             device_map="auto"
                                             ## revision="07a117c" ##AutoGPTQ format
                                             )
text = "下面我来介绍一下阿里巴巴公司，"
text = "9.8和9.11哪个数字大？答案是"
text = "Once upon a time,"
text = "There is a girl who likes adventure,"
inputs = tokenizer(text, return_tensors="pt").to(model.device)
print(tokenizer.decode(model.generate(**inputs, max_new_tokens=50, do_sample=False)[0]))
##下面我来介绍一下阿里巴巴公司，阿里巴巴公司是全球领先的电子商务公司，成立于1999年，总部位于中国杭州。阿里巴巴公司致力于为全球中小企业提供一个在线交易平台，帮助他们拓展业务，提高销售额。阿里巴巴公司拥有多个业务板块，包括淘宝、天猫
##
##9.8和9.11哪个数字大？答案是9.8，因为9.8比9.11大0.7。
##Once upon a time, there was a little girl named Alice who loved to read. She had a special book that she had inherited from her grandmother, and it was filled with stories of magical creatures and far-off lands. One day, Alice decided to read the book in a
##There is a girl who likes adventure, and she is always looking for new experiences. She is a bit of a thrill-seeker, and she loves to push herself to the limit. She is always up for a challenge, and she is not afraid to take risks. She is a bit

Intel Gaudi-2 INT4 Inference

import habana_frameworks.torch.core as htcore
import habana_frameworks.torch.hpu as hthpu

from auto_round import AutoRoundConfig
from transformers import AutoModelForCausalLM,AutoTokenizer

quantized_model_dir = "Intel/Qwen2-7B-int4-inc"
tokenizer = AutoTokenizer.from_pretrained(quantized_model_dir)
model = AutoModelForCausalLM.from_pretrained(quantized_model_dir).to('hpu').to(bfloat16)
text = "下面我来介绍一下阿里巴巴公司,"
inputs = tokenizer(text, return_tensors="pt").to(model.device)
print(tokenizer.decode(model.generate(**inputs, max_new_tokens=50, do_sample=False)[0]))

Advanced Usage

Evaluate the model

auto-round  --model "Intel/Qwen2-7B-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

Generate the model

auto-round
--model_name  Qwen/Qwen2-7B \
--device 0 \
--group_size 128 \
--nsamples 512 \
--bits 4 \
--iter 1000 \
--disable_eval \
--model_dtype "float16" \
--format 'auto_round' \
--output_dir "./tmp_autoround" 

📚 Documentation

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:

• Intel Neural Compressor link
• Intel Extension for Transformers link

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.

📄 License

This model is licensed under the Apache-2.0 license.

🔧 Technical Details

The model is an int4 auto-round model with group_size 128 of Qwen/Qwen2-7B generated by intel/auto-round.

📖 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}
}

arxiv github