---
pipeline_tag: image-text-to-text
license: other
license_name: minimax-community
license_link: LICENSE
library_name: transformers
tags:
  - multimodal
  - moe
  - agent
  - coding
  - video
---

# Model Overview

- **Model Architecture:** MiniMaxM3SparseForConditionalGeneration
  - **Input:** Text, Image
  - **Output:** Text
- **Supported Hardware Microarchitecture:** AMD MI350/MI355
- **ROCm**: 7.1.1
- **PyTorch**: 2.10.0
- **Transformers**: 5.2.0
- **Operating System(s):** Linux
- **Inference Engine:** [vLLM](https://docs.vllm.ai/en/latest/)
- **Model Optimizer:** [AMD-Quark](https://quark.docs.amd.com/latest/index.html)
  - **Weight quantization:** OCP MXFP4, Static
  - **Activation quantization:** OCP MXFP4, Dynamic


# Model Quantization

The model was quantized from [MiniMaxAI/MiniMax-M3](https://huggingface.co/MiniMaxAI/MiniMax-M3) using [AMD-Quark](https://quark.docs.amd.com/latest/index.html). The weights are quantized to MXFP4 and activations are quantized to MXFP4.


**Quantization scripts:**
```
cd Quark/examples/torch/language_modeling/llm_ptq/
exclude_layers="*lm_head *vision_tower* *multi_modal_projector* *patch_merge_mlp* *block_sparse_moe.gate *self_attn* *mlp.gate_proj *mlp.up_proj *mlp.down_proj"
CUDA_VISIBLE_DEVICES=0 python3 quantize_quark.py \
  --model_dir MiniMaxAI/MiniMax-M3 \
  --quant_scheme mxfp4 \
  --exclude_layers $exclude_layers \
  --output_dir /mnt/amd/MiniMax-M3-MXFP4 \
  --file2file_quantization
```

For further details or issues, please refer to the AMD-Quark documentation or contact the respective developers.

# Evaluation
The model was evaluated on gsm8k benchmarks using the vllm framework.

### Accuracy

<table>
  <tr>
   <td><strong>Benchmark</strong>
   </td>
   <td><strong>MiniMaxAI/MiniMax-M3 </strong>
   </td>
   <td><strong>amd/MiniMax-M3-MXFP4(this model)</strong>
   </td>
   <td><strong>Recovery</strong>
   </td>
  </tr>
  <tr>
   <td>gsm8k (flexible-extract) 
   </td>
   <td>95.30
   </td>
   <td>94.19
   </td>
   <td>98.84%
   </td>
  </tr>
</table>


### Reproduction

The GSM8K results were obtained using the lm-eval framework, based on the
Docker image `rocm/pytorch-private:vllm-hy-mm-06112026`. The vLLM shipped in
that image was used as-is, with the patch from this PR ([#45794](https://github.com/vllm-project/vllm/pull/45794/changes)) applied on top.

#### Launching server
```
vllm serve /mnt/amd/MiniMax-M3-MXFP4 \
  --trust-remote-code \
  --block-size 128 \
  --tensor-parallel-size 8 \
  --attention-backend TRITON_ATTN \
  --mm-encoder-tp-mode data \
  --mm-encoder-attn-backend ROCM_AITER_FA \
  --tool-call-parser minimax_m3 \
  --enable-auto-tool-choice \
  --reasoning-parser minimax_m3 \
  --moe-backend emulation
```


#### Evaluating model in a new terminal
```
lm_eval \
  --model local-chat-completions \
  --model_args "model=/mnt/amd/MiniMax-M3-MXFP4,base_url=http://127.0.0.1:8000/v1/chat/completions,num_concurrent=32,max_gen_toks=16384" \
  --tasks gsm8k \
  --num_fewshot 5 \
  --batch_size 1 \
  --apply_chat_template \
  --fewshot_as_multiturn
```