This model is a int4 weight-only quantization with group_size 128 and symmetric quantization of OrionLLM/GRM-3.2-Turf generated by TUNING. Please follow the license of the original model.

Quantization Details

Attribute	Value
Base Model	OrionLLM/GRM-3.2-Turf
Quantization Tool	TUNING
Quantization Scheme	W4A16
Quantized Size	770 MB

Evaluation Results

Task	Accuracy
hellaswag	0.3998
mmlu	0.2763
mmlu_abstract_algebra	0.2800
mmlu_anatomy	0.2519
mmlu_astronomy	0.2237
mmlu_business_ethics	0.3100
mmlu_clinical_knowledge	0.3358
mmlu_college_biology	0.2708
mmlu_college_chemistry	0.2500
mmlu_college_computer_science	0.2600
mmlu_college_mathematics	0.2600
mmlu_college_medicine	0.3121
mmlu_college_physics	0.2157
mmlu_computer_security	0.4400
mmlu_conceptual_physics	0.2851
mmlu_econometrics	0.2632
mmlu_electrical_engineering	0.2621
mmlu_elementary_mathematics	0.2354
mmlu_formal_logic	0.2698
mmlu_global_facts	0.3700
mmlu_high_school_biology	0.2742
mmlu_high_school_chemistry	0.1970
mmlu_high_school_computer_science	0.3000
mmlu_high_school_european_history	0.2364
mmlu_high_school_geography	0.2828
mmlu_high_school_government_and_politics	0.2953
mmlu_high_school_macroeconomics	0.2487
mmlu_high_school_mathematics	0.2481
mmlu_high_school_microeconomics	0.2353
mmlu_high_school_physics	0.2185
mmlu_high_school_psychology	0.2936
mmlu_high_school_statistics	0.2037
mmlu_high_school_us_history	0.2745
mmlu_high_school_world_history	0.2827
mmlu_human_aging	0.2915
mmlu_human_sexuality	0.2977
mmlu_humanities	0.2638
mmlu_international_law	0.3388
mmlu_jurisprudence	0.3241
mmlu_logical_fallacies	0.2822
mmlu_machine_learning	0.3304
mmlu_management	0.2816
mmlu_marketing	0.3547
mmlu_medical_genetics	0.3400
mmlu_miscellaneous	0.3436
mmlu_moral_disputes	0.2861
mmlu_moral_scenarios	0.2425
mmlu_nutrition	0.3333
mmlu_other	0.3109
mmlu_philosophy	0.2187
mmlu_prehistory	0.2716
mmlu_professional_accounting	0.2305
mmlu_professional_law	0.2536
mmlu_professional_medicine	0.2022
mmlu_professional_psychology	0.2958
mmlu_public_relations	0.3273
mmlu_security_studies	0.2367
mmlu_social_sciences	0.2811
mmlu_sociology	0.3234
mmlu_stem	0.2563
mmlu_us_foreign_policy	0.3000
mmlu_virology	0.3193
mmlu_world_religions	0.3626
piqa	0.6861

How to Use

HF Usage

Step 1: Install AutoRound

pip install auto-round

Step 2: Load and run the quantized model

python

from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "GRM-3.2-Turf-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

bash

vllm serve GRM-3.2-Turf-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.

LeaderboardModel1/GRM-3.2-Turf-AutoRound-W4A16-Tuning

GRM-3.2-Turf-AutoRound-W4A16-Tuning

Model Details