Instructions to use intel-ai/gemma-4-31B-MXFP8-FP8KV-CT-AutoRound with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use intel-ai/gemma-4-31B-MXFP8-FP8KV-CT-AutoRound with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("image-text-to-text", model="intel-ai/gemma-4-31B-MXFP8-FP8KV-CT-AutoRound")

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("intel-ai/gemma-4-31B-MXFP8-FP8KV-CT-AutoRound")
model = AutoModelForMultimodalLM.from_pretrained("intel-ai/gemma-4-31B-MXFP8-FP8KV-CT-AutoRound", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use intel-ai/gemma-4-31B-MXFP8-FP8KV-CT-AutoRound with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "intel-ai/gemma-4-31B-MXFP8-FP8KV-CT-AutoRound"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "intel-ai/gemma-4-31B-MXFP8-FP8KV-CT-AutoRound",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/intel-ai/gemma-4-31B-MXFP8-FP8KV-CT-AutoRound

SGLang

How to use intel-ai/gemma-4-31B-MXFP8-FP8KV-CT-AutoRound with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "intel-ai/gemma-4-31B-MXFP8-FP8KV-CT-AutoRound" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "intel-ai/gemma-4-31B-MXFP8-FP8KV-CT-AutoRound",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "intel-ai/gemma-4-31B-MXFP8-FP8KV-CT-AutoRound" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "intel-ai/gemma-4-31B-MXFP8-FP8KV-CT-AutoRound",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use intel-ai/gemma-4-31B-MXFP8-FP8KV-CT-AutoRound with Docker Model Runner:
```
docker model run hf.co/intel-ai/gemma-4-31B-MXFP8-FP8KV-CT-AutoRound
```

Configuration Parsing Warning:In UNKNOWN_FILENAME: "quantization_config.config_groups.group_0.format" must be a string

Model Details

This model is an MXFP8 (W8A8, group_size 32) model of google/gemma-4-31B generated by intel/auto-round with 200 tuning iterations. Please follow the license of the original model.

Quantized (MXFP8): attention, MLP
Kept in BF16: embeddings, vision tower, vision projection
Static FP8: KV cache

How to Use

Environment

uv pip install vllm==0.29.0
uv pip install flashinfer-python==0.7.0

vLLM Usage

vllm serve intel-ai/gemma-4-31B-MXFP8-FP8KV-CT-AutoRound \
    --attention-backend FLASHINFER \
    --language-model-only \
    --dtype bfloat16 \
    --kv-cache-dtype fp8

Evaluation

Benchmark	BF16	MXFP8	Recovery
PIQA	82.81	82.70	99.87%
HellaSwag	65.96	65.56	99.40%
MMLU	81.17	81.13	99.95%
GSM8K	86.28	86.35	100.09%
Average	79.05	78.94	99.85%
niah_multiquery (128K)	100.00	100.00	100.00%
ruler_qa_squad (128K)	39.42	40.08	101.69%
RULER Average	69.71	70.04	100.48%

Generate the Model

uv pip install git+https://github.com/intel/auto-round.git@20710256296c4e129de5ef9639755bc8a03edcbe

auto-round \
    --model_name google/gemma-4-31B \
    --scheme MXFP8 \
    --algorithm auto_round \
    --iters 200 \
    --nsamples 128 \
    --static_kv_dtype fp8 \
    --ignore_layers embed_tokens,vision_tower,embed_vision \
    --format llm_compressor \
    --output_dir gemma-4-31B-MXFP8-FP8KV-CT-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:

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: Sun, Xuehao xuehao.sun@intel.com

Downloads last month: 87

Safetensors

Model size

31B params

Tensor type

F32

BF16

F8_E4M3

Model tree for intel-ai/gemma-4-31B-MXFP8-FP8KV-CT-AutoRound

Base model

google/gemma-4-31B

Quantized

(45)

this model

Paper for intel-ai/gemma-4-31B-MXFP8-FP8KV-CT-AutoRound

Optimize Weight Rounding via Signed Gradient Descent for the Quantization of LLMs

Paper • 2309.05516 • Published Sep 11, 2023 • 16