gemma-4-26B-A4B-it-AutoRound-W4A16-RTN

Model Details

This model is a int4 weight-only quantization with group_size 128 and symmetric quantization of google/gemma-4-26B-A4B-it generated by AutoRound. Please follow the license of the original model.

Quantization Details

Attribute	Value
Base Model	google/gemma-4-26B-A4B-it
Quantization Tool	AutoRound
Quantization Scheme	W4A16
Quantized Size	14655 MB

Evaluation Results

Task	Accuracy
hellaswag	0.3310
mmlu	0.4321
mmlu_abstract_algebra	0.3000
mmlu_anatomy	0.4074
mmlu_astronomy	0.4539
mmlu_business_ethics	0.4700
mmlu_clinical_knowledge	0.3585
mmlu_college_biology	0.5278
mmlu_college_chemistry	0.4100
mmlu_college_computer_science	0.4200
mmlu_college_mathematics	0.3000
mmlu_college_medicine	0.3931
mmlu_college_physics	0.2843
mmlu_computer_security	0.4300
mmlu_conceptual_physics	0.3617
mmlu_econometrics	0.4123
mmlu_electrical_engineering	0.4207
mmlu_elementary_mathematics	0.4259
mmlu_formal_logic	0.3333
mmlu_global_facts	0.2200
mmlu_high_school_biology	0.5677
mmlu_high_school_chemistry	0.3892
mmlu_high_school_computer_science	0.5600
mmlu_high_school_european_history	0.6485
mmlu_high_school_geography	0.4747
mmlu_high_school_government_and_politics	0.5751
mmlu_high_school_macroeconomics	0.4333
mmlu_high_school_mathematics	0.3556
mmlu_high_school_microeconomics	0.5042
mmlu_high_school_physics	0.3444
mmlu_high_school_psychology	0.5211
mmlu_high_school_statistics	0.4537
mmlu_high_school_us_history	0.6961
mmlu_high_school_world_history	0.6371
mmlu_human_aging	0.3722
mmlu_human_sexuality	0.4885
mmlu_humanities	0.4421
mmlu_international_law	0.5455
mmlu_jurisprudence	0.4259
mmlu_logical_fallacies	0.4540
mmlu_machine_learning	0.3929
mmlu_management	0.4272
mmlu_marketing	0.4188
mmlu_medical_genetics	0.4100
mmlu_miscellaneous	0.3448
mmlu_moral_disputes	0.3642
mmlu_moral_scenarios	0.3631
mmlu_nutrition	0.4150
mmlu_other	0.3891
mmlu_philosophy	0.3923
mmlu_prehistory	0.4722
mmlu_professional_accounting	0.4078
mmlu_professional_law	0.4139
mmlu_professional_medicine	0.4816
mmlu_professional_psychology	0.4020
mmlu_public_relations	0.4273
mmlu_security_studies	0.5061
mmlu_social_sciences	0.4732
mmlu_sociology	0.4677
mmlu_stem	0.4196
mmlu_us_foreign_policy	0.5600
mmlu_virology	0.4096
mmlu_world_religions	0.5322
piqa	0.5501

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 = "gemma-4-26B-A4B-it-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 gemma-4-26B-A4B-it-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:

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.