Intel/gemma-7b-int4-inc

Model Details: gemma-7b-int4-inc

This model is an int4 model with group_size 128 of google/gemma-7b generated by intel/auto-round. Inference of this model is compatible with AutoGPTQ's Kernel.

How To Use

Reproduce the model

Here is the sample command to reproduce the model

git clone https://github.com/intel/auto-round
cd auto-round/examples/language-modeling
pip install -r requirements.txt
python3 main.py \
--model_name  google/gemma-7b \
--device 0 \
--group_size 128 \
--bits 4 \
--iters 1000 \
--enable_minmax_tuning \
--minmax_lr 2e-3 \
--low_gpu_mem_usage \
--deployment_device 'gpu' \
--scale_dtype 'fp32' \
--eval_bs 32 \
--output_dir "./tmp_autoround" \
--amp 

Evaluate the model

Install lm-eval-harness 0.4.2 from source. Install the latest AutoGPTQ from source first

lm_eval --model hf --model_args pretrained="Intel/gemma-7b-int4-inc",autogptq=True,gptq_use_triton=True --device cuda:0 --tasks lambada_openai,hellaswag,piqa,winogrande,truthfulqa_mc1,openbookqa,boolq,rte,arc_easy,arc_challenge,mmlu --batch_size 32

Metric	FP16	int4
Avg.	0.6302	0.6262
mmlu	0.6199	0.6111
lambada_openai	0.7316	0.7252
hellaswag	0.6059	0.6033
winogrande	0.7506	0.7451
piqa	0.8014	0.8058
truthfulqa_mc1	0.3121	0.2889
openbookqa	0.3220	0.3380
boolq	0.8339	0.8281
arc_easy	0.8253	0.8152
arc_challenge	0.4991	0.5017

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.

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}, journal={arXiv preprint arXiv:2309.05516}, year={2023} }

arxiv github