Kimi-K2-Instruct-0905-MXFP4

Model Overview

• Model Architecture: Kimi-K2-Instruct
  • Input: Text
  • Output: Text
• Supported Hardware Microarchitecture: AMD MI350/MI355
• ROCm: 7.0
• Operating System(s): Linux
• Inference Engine: vLLM
• Model Optimizer: AMD-Quark (V0.11.1)
  • Quantized layers: experts, shared_experts
  • Weight quantization: OCP MXFP4, Static
  • Activation quantization: OCP MXFP4, Dynamic
• Calibration Dataset: Pile

This model was built with Kimi-K2-Instruct model by applying AMD-Quark for MXFP4 quantization.

Model Quantization

The model was quantized from unsloth/Kimi-K2-Instruct-0905-BF16 using AMD-Quark. The weights and activations are quantized to MXFP4.

Quantization scripts:

cd Quark/examples/torch/language_modeling/llm_ptq/
exclude_layers="*self_attn* *mlp.gate *lm_head *mlp.gate_proj *mlp.up_proj *mlp.down_proj"

python quantize_quark.py \
    --model_dir unsloth/Kimi-K2-Instruct-0905-BF16 \
    --quant_scheme mxfp4 \
    --exclude_layers  $exclude_layers \
    --output_dir amd/Kimi-K2-Instruct-0905-MXFP4 \
    --file2file_quantization

Deployment

Use with vLLM

This model can be deployed efficiently using the vLLM backend.

Evaluation

The model was evaluated on GSM8K benchmarks.

Accuracy

<table> <tr> <td><strong>Benchmark</strong> </td> <td><strong>Kimi-K2-Instruct-0905 </strong> </td> <td><strong>Kimi-K2-Instruct-0905-MXFP4(this model)</strong> </td> <td><strong>Recovery</strong> </td> </tr> <tr> <td>GSM8K (flexible-extract) </td> <td>95.45 </td> <td>93.78 </td> <td>98.25% </td> </tr> </table>

Reproduction

The GSM8K results were obtained using the lm-evaluation-harness framework, based on the Docker image rocm/vllm-private:vllm_dev_base_mxfp4_20260122, with vLLM and lm-eval compiled and installed from source inside the image.

Launching server

export VLLM_ATTENTION_BACKEND="TRITON_MLA"
export VLLM_ROCM_USE_AITER=1
export VLLM_ROCM_USE_AITER_FUSION_SHARED_EXPERTS=0

vllm serve amd/Kimi-K2-Instruct-0905-MXFP4 \
  --port 8000 \
  --served-model-name kimi-k2-mxfp4 \
  --trust-remote-code \
  --tensor-parallel-size 8 \
  --enable-auto-tool-choice \
  --tool-call-parser kimi_k2

Evaluating model in a new terminal

lm_eval \
  --model local-completions \
  --model_args "model=kimi-k2-mxfp4,base_url=http://0.0.0.0:8000/v1/completions,tokenized_requests=False,tokenizer_backend=None,num_concurrent=32" \
  --tasks gsm8k \
  --num_fewshot 5 \
  --batch_size 1

License

Modifications Copyright(c) 2025 Advanced Micro Devices, Inc. All rights reserved.