intel-ai/FLUX.2-klein-4B-MXFP4-RTN-AutoRound

Model Details

This model is an MXFP4 (W4A4, group_size 32) model of black-forest-labs/FLUX.2-klein-4B generated by intel/auto-round using RTN. Please follow the license of the original model.

• Quantized (MXFP4): transformer blocks
• Kept in BF16: text encoders / VAE

How to Use

Environment

uv pip install vllm==0.28.0
uv pip install git+https://github.com/lkk12014402/vllm-omni.git@5aa7e3d9c84a5e450aa037e142a7df1bfc9addb5
uv pip install flashinfer-python==0.6.18

vLLM Usage

vllm serve intel-ai/FLUX.2-klein-4B-MXFP4-RTN-AutoRound \
    --omni

Evaluation

Benchmark	BF16	MXFP4	Recovery
CLIP	32.07	32.44	101.15%
CLIP-IQA	69.87	66.97	95.85%
ImageReward	104.00	100.00	96.15%
GenEval	82.19	79.97	97.30%
Average	72.03	69.84	96.96%

Generate the Model

auto-round \
    --model_name black-forest-labs/FLUX.2-klein-4B \
    --scheme MXFP4 \
    --iters 0 \
    --format auto_round \
    --output_dir FLUX.2-klein-4B-MXFP4-RTN-AutoRound

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

Signed-off-by: Lv, Kaokao kaokao.lv@intel.com