Modelpublic

amd/Kimi-K2.5-MXFP4

sourceHugging Faceotherupdated 3mo agoView on Hugging Face

4likes66kdownloads

Model Card

Model Overview

—Model Architecture: Kimi-K2.5
—Input: Text, Image, Video
—Output: Text
—Supported Hardware Microarchitecture: AMD MI350/MI355
—ROCm: 7.1.0
—Operating System(s): Linux
—Inference Engine: vLLM
—Model Optimizer: AMD-Quark (V0.11.1)
—Quantized layers: layers.0.mlp, experts and shared_experts
—Weight quantization: OCP MXFP4, Static
—Activation quantization: OCP MXFP4, Dynamic
—Calibration Dataset: Pile

This model was built with Kimi-K2.5 model by applying AMD-Quark for MXFP4 quantization.

Model Quantization

The model was quantized from moonshotai/Kimi-K2.5 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 *mm_projector* *vision_tower*"

python quantize_quark.py \
    --model_dir moonshotai/Kimi-K2.5 \
    --quant_scheme mxfp4 \
    --exclude_layers  $exclude_layers \
    --output_dir amd/Kimi-K2.5-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-K2.5 </strong> </td> <td><strong>Kimi-K2.5-MXFP4(this model) </strong> </td> <td><strong>Recovery</strong> </td> </tr> <tr> <td>GSM8K (flexible-extract) </td> <td>94.09 </td> <td>93.1 </td> <td>98.95% </td> </tr> </table>

Reproduction

The GSM8K results were obtained using the lm-evaluation-harness framework, based on the Docker image vllm/vllm-openai-rocm:v0.17.0.

Install the lm-eval (Version: 0.4.11) in container first.

pip install lm-eval[api]

Launching server

export VLLM_ROCM_USE_AITER=1

vllm serve amd/Kimi-K2.5-MXFP4 -tp 4 \
  --mm-encoder-tp-mode data \
  --tool-call-parser kimi_k2 \
  --reasoning-parser kimi_k2 \
  --enforce-eager \
  --trust-remote-code

Evaluating model in a new terminal

lm_eval \
  --model local-completions \
  --model_args "model=amd/Kimi-K2.5-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