Instructions to use INCModel/DeepSeek-V4-Flash-MXFP4-LLMC with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use INCModel/DeepSeek-V4-Flash-MXFP4-LLMC with Transformers:

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

pipe = pipeline("text-generation", model="INCModel/DeepSeek-V4-Flash-MXFP4-LLMC")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("INCModel/DeepSeek-V4-Flash-MXFP4-LLMC")
model = AutoModelForCausalLM.from_pretrained("INCModel/DeepSeek-V4-Flash-MXFP4-LLMC")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use INCModel/DeepSeek-V4-Flash-MXFP4-LLMC with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "INCModel/DeepSeek-V4-Flash-MXFP4-LLMC"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "INCModel/DeepSeek-V4-Flash-MXFP4-LLMC",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/INCModel/DeepSeek-V4-Flash-MXFP4-LLMC

SGLang

How to use INCModel/DeepSeek-V4-Flash-MXFP4-LLMC 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 "INCModel/DeepSeek-V4-Flash-MXFP4-LLMC" \
    --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": "INCModel/DeepSeek-V4-Flash-MXFP4-LLMC",
		"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 "INCModel/DeepSeek-V4-Flash-MXFP4-LLMC" \
        --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": "INCModel/DeepSeek-V4-Flash-MXFP4-LLMC",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use INCModel/DeepSeek-V4-Flash-MXFP4-LLMC with Docker Model Runner:
```
docker model run hf.co/INCModel/DeepSeek-V4-Flash-MXFP4-LLMC
```

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

This model is a MXFP4 model of deepseek-ai/DeepSeek-V4-Flash generated by intel/auto-round with llm_compressor format. Please follow the license of the original model.

Model	aime26	gpqa_diamond	math_500	mmlu	piqa	hellaswag	gsm8k	mmlu_pro
DeepSeek-V4-Flash (W4A8 MXFP4 MoE, block FP8 Linear)	0.6097	0.7212	0.9556	0.9043	0.9289	0.8632	0.9684	0.8289
INCModel/DeepSeek-V4-Flash-MXFP4-Mixed-LLMC (W4A4 MXFP4 MoE, MXFP8 Linear)	0.6087	0.7192	0.9472	0.9020	0.9287	0.8650	0.9671	0.8258
INCModel/DeepSeek-V4-Flash-MXFP4-LLMC (W4A4 MXFP4 MoE & Linear)	0.5663	0.7013	0.9440	0.8923	0.9229	0.8574	0.9630	0.8136

How to Run Locally

vLLM is supported here: https://github.com/vllm-project/vllm/pull/45645

CUDA_VISIBLE_DEVICES=0,1 vllm serve INCModel/DeepSeek-V4-Flash-MXFP4-LLMC \
  --trust-remote-code   \
  --kv-cache-dtype fp8   \
  --block-size 256     \
  --tensor-parallel-size 2    \
  --attention_config.use_fp4_indexer_cache=True  \
  --port 8009  \
  --no-enable-flashinfer-autotune  \
  --enforce-eager

curl -s http://127.0.0.1:8009/v1/chat/completions   -H "Content-Type: application/json"   -d '{
    "model": "DeepSeek-V4-Flash-MXFP4-LLMC",
    "messages": [
      {"role":"user","content":"2+3=?"}
    ],
    "max_tokens": 10,
    "extra_body": {
      "chat_template_kwargs": {
        "enable_thinking": true
      }
    }
  }' | python3 -m json.tool

For local deployment, we recommend setting the sampling parameters to temperature = 1.0, top_p = 1.0. For the Think Max reasoning mode, we recommend setting the context window to at least 384K tokens.

Generate the Model

Depends on this PR: https://github.com/intel/auto-round/pull/1921

auto-round deepseek-ai/DeepSeek-V4-Flash \
  --model_free \
  --scheme MXFP4 \
  --ignore_layers compressor,indexer.weights_proj \
  --format llm_compressor \
  --output_dir "./DeepSeek-V4-Flash-MXFP4"

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

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

Downloads last month: 141

Safetensors

Model size

155B params

Tensor type

BF16

I64

F32

Model tree for INCModel/DeepSeek-V4-Flash-MXFP4-LLMC

Base model

deepseek-ai/DeepSeek-V4-Flash

Quantized

(74)

this model

Paper for INCModel/DeepSeek-V4-Flash-MXFP4-LLMC

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

Paper • 2309.05516 • Published Sep 11, 2023 • 12