Qwen2-57B-A14B-AutoRound-W4A16-Tuning

Model Details

This model is a int4 weight-only quantization with group_size 128 and symmetric quantization of Qwen/Qwen2-57B-A14B generated by TUNING. Please follow the license of the original model.

Quantization Details

Attribute	Value
Base Model	Qwen/Qwen2-57B-A14B
Quantization Tool	TUNING
Quantization Scheme	W4A16
Quantized Size	29993 MB

Evaluation Results

Task	Accuracy
hellaswag	0.6252
mmlu	0.7381
mmlu_abstract_algebra	0.4600
mmlu_anatomy	0.7481
mmlu_astronomy	0.8289
mmlu_business_ethics	0.8000
mmlu_clinical_knowledge	0.8377
mmlu_college_biology	0.8681
mmlu_college_chemistry	0.5100
mmlu_college_computer_science	0.6300
mmlu_college_mathematics	0.4800
mmlu_college_medicine	0.7457
mmlu_college_physics	0.5098
mmlu_computer_security	0.8400
mmlu_conceptual_physics	0.7872
mmlu_econometrics	0.5965
mmlu_electrical_engineering	0.7310
mmlu_elementary_mathematics	0.6958
mmlu_formal_logic	0.5476
mmlu_global_facts	0.5200
mmlu_high_school_biology	0.8581
mmlu_high_school_chemistry	0.6453
mmlu_high_school_computer_science	0.7800
mmlu_high_school_european_history	0.8121
mmlu_high_school_geography	0.9040
mmlu_high_school_government_and_politics	0.9430
mmlu_high_school_macroeconomics	0.7872
mmlu_high_school_mathematics	0.5074
mmlu_high_school_microeconomics	0.8782
mmlu_high_school_physics	0.5960
mmlu_high_school_psychology	0.8954
mmlu_high_school_statistics	0.6806
mmlu_high_school_us_history	0.8922
mmlu_high_school_world_history	0.8692
mmlu_human_aging	0.7803
mmlu_human_sexuality	0.8626
mmlu_humanities	0.6697
mmlu_international_law	0.8678
mmlu_jurisprudence	0.8241
mmlu_logical_fallacies	0.8589
mmlu_machine_learning	0.6161
mmlu_management	0.8641
mmlu_marketing	0.9316
mmlu_medical_genetics	0.8800
mmlu_miscellaneous	0.8978
mmlu_moral_disputes	0.7803
mmlu_moral_scenarios	0.4715
mmlu_nutrition	0.8072
mmlu_other	0.7976
mmlu_philosophy	0.7717
mmlu_prehistory	0.8117
mmlu_professional_accounting	0.5957
mmlu_professional_law	0.5750
mmlu_professional_medicine	0.8199
mmlu_professional_psychology	0.7876
mmlu_public_relations	0.7364
mmlu_security_studies	0.7714
mmlu_social_sciences	0.8346
mmlu_sociology	0.8955
mmlu_stem	0.6876
mmlu_us_foreign_policy	0.9000
mmlu_virology	0.5120
mmlu_world_religions	0.8713
piqa	0.8003

How to Use

HF Usage

Step 1: Install AutoRound

pip install auto-round

Step 2: Load and run the quantized model

from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "Qwen2-57B-A14B-AutoRound-W4A16-Tuning"

# load the tokenizer and the model
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype="auto", device_map="auto")

# prepare the model input
prompt = "Write a quick sort algorithm."
messages = [{"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)

# conduct text completion
generated_ids = model.generate(**model_inputs, max_new_tokens=512)
output_ids = generated_ids[0][len(model_inputs.input_ids[0]) :].tolist()

content = tokenizer.decode(output_ids, skip_special_tokens=True)
print("content:", content)

VLLM Usage

vllm serve Qwen2-57B-A14B-AutoRound-W4A16-Tuning \
    --trust-remote-code \
    --dtype bfloat16 \
    --tensor_parallel_size 1

If you encounter any issues, feel free to open an issue on the AutoRound GitHub repo or provide feedback on the Low-Bit Open LLM Leaderboard.

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
• AutoRound

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}
}

arxiv github

---

This model is part of the Intel Low-Bit Open LLM Leaderboard initiative.