experiment024b-AutoRound-W4A16-RTN

Model Details

This model is a int4 weight-only quantization with group_size 128 and symmetric quantization of groxaxo/experiment024b generated by AutoRound. Please follow the license of the original model.

Quantization Details

Attribute	Value
Base Model	groxaxo/experiment024b
Quantization Tool	AutoRound
Quantization Scheme	W4A16
Original Size	48960 MB
Quantized Size	13575 MB

Evaluation Results

Task	Accuracy
hellaswag	0.5830
mmlu	0.6724
mmlu_abstract_algebra	0.4200
mmlu_anatomy	0.7556
mmlu_astronomy	0.8421
mmlu_business_ethics	0.6700
mmlu_clinical_knowledge	0.7547
mmlu_college_biology	0.8264
mmlu_college_chemistry	0.5400
mmlu_college_computer_science	0.6200
mmlu_college_mathematics	0.4700
mmlu_college_medicine	0.6590
mmlu_college_physics	0.5000
mmlu_computer_security	0.7300
mmlu_conceptual_physics	0.6383
mmlu_econometrics	0.4737
mmlu_electrical_engineering	0.7034
mmlu_elementary_mathematics	0.5714
mmlu_formal_logic	0.3810
mmlu_global_facts	0.5000
mmlu_high_school_biology	0.8323
mmlu_high_school_chemistry	0.6108
mmlu_high_school_computer_science	0.7300
mmlu_high_school_european_history	0.8303
mmlu_high_school_geography	0.8283
mmlu_high_school_government_and_politics	0.8756
mmlu_high_school_macroeconomics	0.7231
mmlu_high_school_mathematics	0.4370
mmlu_high_school_microeconomics	0.7437
mmlu_high_school_physics	0.4834
mmlu_high_school_psychology	0.8606
mmlu_high_school_statistics	0.5694
mmlu_high_school_us_history	0.8333
mmlu_high_school_world_history	0.8186
mmlu_human_aging	0.7040
mmlu_human_sexuality	0.8092
mmlu_humanities	0.5894
mmlu_international_law	0.8430
mmlu_jurisprudence	0.8241
mmlu_logical_fallacies	0.8160
mmlu_machine_learning	0.4911
mmlu_management	0.8350
mmlu_marketing	0.9103
mmlu_medical_genetics	0.7400
mmlu_miscellaneous	0.8301
mmlu_moral_disputes	0.7514
mmlu_moral_scenarios	0.2927
mmlu_nutrition	0.8105
mmlu_other	0.7399
mmlu_philosophy	0.7331
mmlu_prehistory	0.7747
mmlu_professional_accounting	0.5284
mmlu_professional_law	0.4941
mmlu_professional_medicine	0.7500
mmlu_professional_psychology	0.7320
mmlu_public_relations	0.6545
mmlu_security_studies	0.7918
mmlu_social_sciences	0.7800
mmlu_sociology	0.8557
mmlu_stem	0.6248
mmlu_us_foreign_policy	0.9300
mmlu_virology	0.5241
mmlu_world_religions	0.8246
piqa	0.7878

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 = "experiment024b-AutoRound-W4A16-RTN"

# 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 experiment024b-AutoRound-W4A16-RTN \
    --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.