Update README.md

README.md

@@ -12,81 +12,83 @@ tags:
   - video
 ---
 
-<div align="center">
-  <img width="60%" src="figures/logo.svg" alt="MiniMax">
-</div>
-<hr>
+This model is a MXFP4 mixed model of [MiniMax-M3](https://huggingface.co/MiniMaxAI/MiniMax-M3) generated by [intel/auto-round](https://github.com/intel/auto-round) with llm_compressor format. Please follow the license of the original model.
+- MOE (block_sparse_moe.experts): MXFP4
+- MOE (block_sparse_moe.shared_experts): MXFP8
+- self_attn (q/k/v/o): MXFP8
 
-<p align="center">
-  <a href="https://agent.minimax.io/" target="_blank"><img src="https://img.shields.io/badge/MiniMax%20Agent-FF6C37?logo=minimax&logoColor=white" alt="MiniMax Agent"></a>
-  <a href="https://platform.minimax.io/docs/guides/text-generation" target="_blank"><img src="https://img.shields.io/badge/API-FF6C37?logo=minimax&logoColor=white" alt="API"></a>
-  <a href="https://www.minimax.io" target="_blank"><img src="https://img.shields.io/badge/MiniMax%20Website-FF6C37?logo=minimax&logoColor=white" alt="MiniMax Website"></a>
-  <br>
-  <a href="https://modelscope.cn/organization/minimax" target="_blank" rel="noopener noreferrer"><img alt="ModelScope MiniMax AI" src="https://img.shields.io/badge/ModelScope-MiniMax%20AI-white?labelColor=%23EF3D5D"></a>
-  <a href="https://platform.minimaxi.com/docs/faq/contact-us" target="_blank"><img src="https://img.shields.io/badge/WeChat-07C160?logo=wechat&logoColor=white" alt="WeChat"></a>
-  <a href="https://discord.com/invite/DPC4AHFCBw" target="_blank"><img src="https://img.shields.io/badge/Discord-5865F2?logo=discord&logoColor=white" alt="Discord"></a>
-  <a href="https://huggingface.co/MiniMaxAI" target="_blank"><img src="https://img.shields.io/badge/Hugging%20Face-FFD21E?logo=huggingface&logoColor=black" alt="Hugging Face"></a>
-  <a href="https://github.com/MiniMax-AI/MiniMax-M3" target="_blank"><img src="https://img.shields.io/badge/GitHub-181717?logo=github&logoColor=white" alt="GitHub"></a>
-  <a href="https://arxiv.org/abs/2606.13392" target="_blank"><img src="https://img.shields.io/badge/arXiv-2606.13392-B31B1B?logo=arxiv&logoColor=white" alt="arXiv Paper"></a>
-  <a href="https://huggingface.co/MiniMaxAI/MiniMax-M3/blob/main/LICENSE" target="_blank"><img src="https://img.shields.io/badge/LICENSE-4CAF50?logo=creativecommons&logoColor=white" alt="LICENSE"></a>
-</p>
+| Configuration | GSM8K | MMLU | PIQA | HelleSwag | Average | Relative to BF16 |
+|--------------|-------|----------|----------|----------|----------|------------------|
+| BF16 | 95.30 | 0.8415 | 0.8259 | 0.6714 | 0.82295 | - |
+| MXFP4 (experts) + MXFP8 (shared_experts) + MXFP8 (self_attn) | 0.9507 | 0.8294 | 0.8210 | 0.6522 | 0.813325 | 98.83% |
 
-MiniMax-M3 is a native multimodal model with 1M context. It has ~428B parameters and ~23B activated parameters.
 
-**Highlights:**
-- **Native Multimodality:** M3 undergoes mixed-modality training from the very first step, enabling deeper semantic fusion across text, image, and video.
-- **Context Scaling via Sparse Attention:** M3 introduces MiniMax Sparse Attention (MSA) to improve long context efficiency. M3 delivers 9× prefill and 15× decode speedups compared to M2 at 1M context, reducing per-token compute to 1/20.
-- **Coding & Cowork Capability:** M3 achieves frontier-level performance across long-horizon agentic benchmarks, excelling in both coding and cowork.
+## How to Run Locally
 
 
-<p align="center">
-  <img width="100%" src="figures/benchmark.jpeg">
-</p>
-
-## MiniMax Sparse Attention (MSA)
-
-M3 is powered by [**MiniMax Sparse Attention (MSA)**](https://github.com/MiniMax-AI/MSA), a high-performance sparse attention operator designed for million-token contexts. Compared with GQA, MSA dramatically reduces the attention compute and memory footprint while preserving model quality.
+```
+vllm serve \
+  INCModelSharing/MiniMax-M3-MXFP4-Mixed-MXFP8-Attn-CT-AutoRound \
+  --tensor-parallel-size 2 \
+  --max-model-len 131072 \
+  --tool-call-parser minimax_m3 \
+  --reasoning-parser minimax_m3 \
+  --enable-auto-tool-choice \
+  --port 8000
+````
 
-<p align="center">
-  <img width="100%" src="figures/efficiency_gqa_vs_msa.png" alt="GQA vs MSA Efficiency Comparison">
-</p>
+```
+curl -s http://127.0.0.1:8000/v1/chat/completions   -H "Content-Type: application/json"   -d '{
+    "model": "INCModelSharing/MiniMax-M3-MXFP4-Mixed-MXFP8-Attn-CT-AutoRound",
+    "messages": [
+      {"role":"user","content":"2+3=?"}
+    ],
+    "max_tokens": 10,
+    "extra_body": {
+      "chat_template_kwargs": {
+        "enable_thinking": true
+      }
+    }
+  }' | python3 -m json.tool
+```
 
-> 📄 Read the technical report: [arXiv:2606.13392](https://arxiv.org/abs/2606.13392) · [Hugging Face Papers](https://huggingface.co/papers/2606.13392)
 
-## How to Use
 
-- [MiniMax Agent](https://agent.minimax.io/)
-- [MiniMax API](https://platform.minimax.io/)
+## Generate the Model
 
-M3 supports three reasoning modes through the `thinking` parameter:
-- **`enabled`** — Reasoning is always enabled.
-- **`adaptive`** — M3 automatically determines when additional reasoning is beneficial.
-- **`disabled`** — Reasoning is disabled to minimize latency and maximize throughput.
 
-## Local Deployment
+~~~bash
+auto-round MiniMaxAI/MiniMax-M3 --model_free \
+  --scheme MXFP8 \
+  --ignore_layers vision_tower,lm_head,block_sparse_moe.gate,embed_tokens,\
+patch_merge_mlp,multi_modal_projector,mlp.gate_proj,mlp.up_proj,mlp.down_proj,\
+self_attn.index_q_proj,self_attn.index_k_proj,self_attn.index_q_norm,self_attn.index_k_norm,\
+self_attn.q_norm,self_attn.k_norm \
+  --layer_config "{block_sparse_moe.experts:{bits:4,data_type:mx_fp}}" \
+  --format llm_compressor \
+  --output_dir "./MiniMax-M3-MXFP4MoE-MXFP8-attn"
+~~~
 
-Download the model:
+## Ethical Considerations and Limitations
 
-```bash
-hf download MiniMaxAI/MiniMax-M3 --local-dir MiniMax-M3
-```
+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.
 
-We recommend the following inference frameworks (listed alphabetically) to serve the model:
+Therefore, before deploying any applications of the model, developers should perform safety testing.
 
-- [SGLang](https://docs.sglang.io/) - see  [SGLang cookbook](https://docs.sglang.io/cookbook/autoregressive/MiniMax/MiniMax-M3).
+## Caveats and Recommendations
 
-- [vLLM](https://github.com/vllm-project/vllm) - see [vLLM recipes](https://recipes.vllm.ai/MiniMaxAI/MiniMax-M3).
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model.
 
-- [Transformers](https://github.com/huggingface/transformers) - see [Transformers docs](https://huggingface.co/docs/transformers/model_doc/minimax_m3_vl).
+Here are a couple of useful links to learn more about Intel's AI software:
 
-- [KTransformers](https://github.com/kvcache-ai/ktransformers) - see [KTransformers MiniMax-M3 tutorial](https://github.com/kvcache-ai/ktransformers/blob/main/doc/en/kt-kernel/MiniMax-M3-Tutorial.md).
+- [Intel Neural Compressor](https://github.com/intel/neural-compressor)
 
-- [unsloth](https://unsloth.ai) - see [tutorial](https://unsloth.ai/docs/models/minimax-m3)
+## Disclaimer
 
-### Inference Parameters
+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.
 
-We recommend the following parameters for best performance: `temperature=1.0`, `top_p=0.95`, `top_k=40`.
+## Cite
 
-## Contact Us
+@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} }
 
-Contact us at [model@minimax.io](mailto:model@minimax.io).
+[arxiv](https://arxiv.org/abs/2309.05516) [github](https://github.com/intel/auto-round)