GRM-2.6-Plus-AutoRound-W4A16-Tuning

Model Details

This model is a int4 weight-only quantization with group_size 128 and symmetric quantization of OrionLLM/GRM-2.6-Plus generated by TUNING. Please follow the license of the original model.

Quantization Details

Attribute	Value
Base Model	OrionLLM/GRM-2.6-Plus
Quantization Tool	TUNING
Quantization Scheme	W4A16
Original Size	4900 MB
Quantized Size	17537 MB

Evaluation Results

Task	Accuracy
hellaswag	0.6372
mmlu	0.8407
mmlu_abstract_algebra	0.7100
mmlu_anatomy	0.8370
mmlu_astronomy	0.9342
mmlu_business_ethics	0.8200
mmlu_clinical_knowledge	0.8642
mmlu_college_biology	0.9514
mmlu_college_chemistry	0.6400
mmlu_college_computer_science	0.8300
mmlu_college_mathematics	0.7200
mmlu_college_medicine	0.8208
mmlu_college_physics	0.7059
mmlu_computer_security	0.8600
mmlu_conceptual_physics	0.9149
mmlu_econometrics	0.7632
mmlu_electrical_engineering	0.8276
mmlu_elementary_mathematics	0.8836
mmlu_formal_logic	0.8175
mmlu_global_facts	0.6500
mmlu_high_school_biology	0.9290
mmlu_high_school_chemistry	0.8374
mmlu_high_school_computer_science	0.9100
mmlu_high_school_european_history	0.8485
mmlu_high_school_geography	0.9293
mmlu_high_school_government_and_politics	0.9793
mmlu_high_school_macroeconomics	0.9410
mmlu_high_school_mathematics	0.6444
mmlu_high_school_microeconomics	0.9454
mmlu_high_school_physics	0.8079
mmlu_high_school_psychology	0.9413
mmlu_high_school_statistics	0.8843
mmlu_high_school_us_history	0.9314
mmlu_high_school_world_history	0.9325
mmlu_human_aging	0.8206
mmlu_human_sexuality	0.8931
mmlu_humanities	0.7928
mmlu_international_law	0.9174
mmlu_jurisprudence	0.8704
mmlu_logical_fallacies	0.9264
mmlu_machine_learning	0.7768
mmlu_management	0.8350
mmlu_marketing	0.9530
mmlu_medical_genetics	0.9300
mmlu_miscellaneous	0.9132
mmlu_moral_disputes	0.7746
mmlu_moral_scenarios	0.7542
mmlu_nutrition	0.8987
mmlu_other	0.8597
mmlu_philosophy	0.8264
mmlu_prehistory	0.8827
mmlu_professional_accounting	0.8085
mmlu_professional_law	0.7040
mmlu_professional_medicine	0.9338
mmlu_professional_psychology	0.8578
mmlu_public_relations	0.7727
mmlu_security_studies	0.8204
mmlu_social_sciences	0.9009
mmlu_sociology	0.9254
mmlu_stem	0.8348
mmlu_us_foreign_policy	0.9300
mmlu_virology	0.5783
mmlu_world_religions	0.9006
piqa	0.8161

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 = "GRM-2.6-Plus-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 GRM-2.6-Plus-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:

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.