update mlx-deploy-guide

README.md

@@ -167,80 +167,6 @@ We look forward to your feedback and to collaborating with developers and resear
 
 Download the model from HuggingFace repository: https://huggingface.co/MiniMaxAI/MiniMax-M2. We recommend using the following inference frameworks (listed alphabetically) to serve the model:
 
-Here's an improved, polished, and professional version of your documentation with better structure, clarity, grammar, accuracy, and usability:
-
-
-### MLX
-
-Run, serve, and fine-tune **MiniMax-M2** locally on your Mac using the **MLX** framework. This guide gets you up and running quickly.
-
-> **Requirements**  
-> - Apple Silicon Mac (M3 Ultra or later)  
-> - **At least 256GB of unified memory (RAM)**  
-
-
-**Installation**
-
-Install the `mlx-lm` package via pip:
-
-```bash
-pip install mlx-lm
-```
-
-**CLI**
-
-Generate text directly from the terminal:
-
-```bash
-mlx_lm.generate \
-  --model mlx-community/MiniMax-M2-4bit \
-  --prompt "How tall is Mount Everest?"
-```
-
-> Add `--max-tokens 256` to control response length, or `--temp 0.7` for creativity.
-
-**Python Script Example**
-
-Use `mlx-lm` in your own Python scripts:
-
-```python
-from mlx_lm import load, generate
-
-# Load the quantized model
-model, tokenizer = load("mlx-community/MiniMax-M2-4bit")
-
-prompt = "Hello, how are you?"
-
-# Apply chat template if available (recommended for chat models)
-if tokenizer.chat_template is not None:
-    messages = [{"role": "user", "content": prompt}]
-    prompt = tokenizer.apply_chat_template(
-        messages,
-        tokenize=False,
-        add_generation_prompt=True
-    )
-
-# Generate response
-response = generate(
-    model,
-    tokenizer,
-    prompt=prompt,
-    max_tokens=256,
-    temp=0.7,
-    verbose=True
-)
-
-print(response)
-```
-
-**Tips**
-- **Model variants**: Check [Hugging Face](https://huggingface.co/collections/mlx-community/minimax-m2) for `MiniMax-M2-4bit`, `6bit`, `8bit`, or `bfloat16` versions.
-- **Fine-tuning**: Use `mlx-lm.lora` for efficient parameter-efficient fine-tuning (PEFT).
-
-**Resources**  
-- GitHub: [https://github.com/ml-explore/mlx-lm](https://github.com/ml-explore/mlx-lm)  
-- Models: [https://huggingface.co/mlx-community](https://huggingface.co/mlx-community)
-
 ### SGLang
 
 We recommend using [SGLang](https://docs.sglang.ai/) to serve MiniMax-M2. SGLang provides solid day-0 support for MiniMax-M2 model. Please refer to our [SGLang Deployment Guide](https://huggingface.co/MiniMaxAI/MiniMax-M2/blob/main/docs/sglang_deploy_guide.md) for more details, and thanks so much for our collaboration with the SGLang team.
@@ -249,6 +175,11 @@ We recommend using [SGLang](https://docs.sglang.ai/) to serve MiniMax-M2. SGLang
 
 We recommend using [vLLM](https://docs.vllm.ai/en/stable/) to serve MiniMax-M2. vLLM provides efficient day-0 support of MiniMax-M2 model, check https://docs.vllm.ai/projects/recipes/en/latest/MiniMax/MiniMax-M2.html for latest deployment guide. We also provide our [vLLM Deployment Guide](https://huggingface.co/MiniMaxAI/MiniMax-M2/blob/main/docs/vllm_deploy_guide.md).
 
+### MLX
+
+We recommend using [MLX-LM](https://github.com/ml-explore/mlx-lm) to serve MiniMax-M2.  Please refer to our [MLX Deployment Guide](https://huggingface.co/MiniMaxAI/MiniMax-M2/blob/main/docs/mlx_deploy_guide.md) for more details.
+
+
 ### Inference Parameters
 We recommend using the following parameters for best performance: `temperature=1.0`, `top_p = 0.95`, `top_k = 40`.
 

docs/mlx_deploy_guide.md

@@ -0,0 +1,73 @@
+Here's an improved, polished, and professional version of your documentation with better structure, clarity, grammar, accuracy, and usability:
+
+
+### MLX
+
+Run, serve, and fine-tune [**MiniMax-M2**](https://huggingface.co/MiniMaxAI/MiniMax-M2) locally on your Mac using the **MLX** framework. This guide gets you up and running quickly.
+
+> **Requirements**  
+> - Apple Silicon Mac (M3 Ultra or later)  
+> - **At least 256GB of unified memory (RAM)**  
+
+
+**Installation**
+
+Install the `mlx-lm` package via pip:
+
+```bash
+pip install mlx-lm
+```
+
+**CLI**
+
+Generate text directly from the terminal:
+
+```bash
+mlx_lm.generate \
+  --model mlx-community/MiniMax-M2-4bit \
+  --prompt "How tall is Mount Everest?"
+```
+
+> Add `--max-tokens 256` to control response length, or `--temp 0.7` for creativity.
+
+**Python Script Example**
+
+Use `mlx-lm` in your own Python scripts:
+
+```python
+from mlx_lm import load, generate
+
+# Load the quantized model
+model, tokenizer = load("mlx-community/MiniMax-M2-4bit")
+
+prompt = "Hello, how are you?"
+
+# Apply chat template if available (recommended for chat models)
+if tokenizer.chat_template is not None:
+    messages = [{"role": "user", "content": prompt}]
+    prompt = tokenizer.apply_chat_template(
+        messages,
+        tokenize=False,
+        add_generation_prompt=True
+    )
+
+# Generate response
+response = generate(
+    model,
+    tokenizer,
+    prompt=prompt,
+    max_tokens=256,
+    temp=0.7,
+    verbose=True
+)
+
+print(response)
+```
+
+**Tips**
+- **Model variants**: Check [Hugging Face](https://huggingface.co/collections/mlx-community/minimax-m2) for `MiniMax-M2-4bit`, `6bit`, `8bit`, or `bfloat16` versions.
+- **Fine-tuning**: Use `mlx-lm.lora` for efficient parameter-efficient fine-tuning (PEFT).
+
+**Resources**  
+- GitHub: [https://github.com/ml-explore/mlx-lm](https://github.com/ml-explore/mlx-lm)  
+- Models: [https://huggingface.co/mlx-community](https://huggingface.co/mlx-community)