Update README.md

README.md

@@ -9,7 +9,7 @@ base_model:
 
 [English](README.md) | [中文](README_CN.md)
 
-## 1. Model Description
+## Model Description
 
 We introduce **MindLink**, a new family of large language models developed by **Kunlun Inc**. Built on **Qwen**, these models incorporate our latest advances in post-training techniques. MindLink demonstrates strong performance across various common benchmarks and is widely applicable in diverse AI scenarios. We welcome feedback to help us continuously optimize and improve our models.
 
@@ -31,7 +31,7 @@ Our training methodology and evaluation: [MindLink](https://github.com/SkyworkAI
 
 ---
 
-## 2. Highlights
+##  Highlights
 
 * **Plan-based Reasoning**: Without the "think" tag, MindLink achieves competitive performance with leading proprietary models across a wide range of reasoning and general tasks. It significantly reduces inference cost, and improves multi-turn capabilities.
 * **Mathematical Framework**: It analyzes the effectiveness of both **Chain-of-Thought (CoT)** and **Plan-based Reasoning**.
@@ -39,7 +39,55 @@ Our training methodology and evaluation: [MindLink](https://github.com/SkyworkAI
 
 ---
 
-## 2. API Access
+## Quickstart
+
+Here provides a code snippet with apply_chat_template to show you how to load the tokenizer and model and how to generate contents.
+
+> ⚠️ Please make sure you have installed `transformers>=4.51.0`. Lower versions are not supported.
+
+```python
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+model_name = "Skywork/MindLink-32B-0801"
+
+model = AutoModelForCausalLM.from_pretrained(
+    model_name,
+    torch_dtype="auto",
+    device_map="auto"
+)
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+
+prompt = "What is the capital of China?"
+messages = [
+    {"role": "user", "content": prompt}
+]
+text = tokenizer.apply_chat_template(
+    messages,
+    tokenize=False,
+    add_generation_prompt=True
+)
+model_inputs = tokenizer([text], return_tensors="pt").to(model.device)
+
+generated_ids = model.generate(
+    **model_inputs,
+    max_new_tokens=512
+)
+generated_ids = [
+    output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
+]
+
+response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
+```
+
+For deployment, you can use sglang>=0.4.6.post1 to create an OpenAI-compatible API endpoint:
+- SGLang:
+    ```shell
+    python -m sglang.launch_server --model-path Skywork/MindLink-32B-0801
+    ```
+
+---
+
+##  API Access
 
 📢 We provide developers with a **one-month free trial** of our API for exploring and testing our models. To request access to an **Open WebUI account** (https://sd1svahsfo0m61h76e190.apigateway-cn-beijing.volceapi.com), please contact us at: **[mindlink@skywork.ai](mailto:mindlink@skywork.ai)**
 
@@ -113,7 +161,7 @@ else:
 
 ---
 
-## 3. Evaluation
+## Evaluation
 
 The results are shown below:
 ![Comparison between MindLink (ML) and other frontier models across various benchmarks.](./figure1.png)