Improve model card: add pipeline tag, library name, code link, and usage example

This PR significantly enhances the model card for `ycwu97/mamba2-distilled-small` by:
- Adding the `pipeline_tag: text-generation` to improve discoverability and classification.
- Specifying `library_name: transformers` as evidenced by the `config.json` and GitHub repository's setup, enabling automated "How to use" snippets on the Hub.
- Including a direct link to the GitHub repository.
- Integrating a comprehensive sample usage code snippet, derived from the project's official GitHub repository, demonstrating how to use this model as a draft model within the STree framework.

Please review these changes.

README.md

@@ -1,12 +1,61 @@
 ---
-license: apache-2.0
-language:
-- en
 base_model:
 - JunxiongWang/Llama3.2-Mamba2-3B-distill
+language:
+- en
+license: apache-2.0
+pipeline_tag: text-generation
+library_name: transformers
 ---
+
 # Description
 
 2 layer mamba2 models distilled from JunxiongWang/Llama3.2-Mamba2-3B-distill. Early stop at 48000 step.
 
-Used in [STree](https://arxiv.org/abs/2505.14969) as a draft model for speculative decoding for hybrid models. 
+Used in [STree: Speculative Tree Decoding for Hybrid State-Space Models](https://arxiv.org/abs/2505.14969) as a draft model for speculative decoding for hybrid models.
+
+For more details on installation, training, and evaluation, please refer to the [GitHub repository](https://github.com/wyc1997/stree).
+
+## Usage
+
+You can use `EaModel.from_pretrained` for accelerated text generation, similar to `generate` from Hugging Face Transformers. Here is an example:
+
+```python
+import torch
+from eagle.model.ea_model import EaModel
+from fastchat.model import get_conversation_template
+
+# Load the base model and EAGLE acceleration model
+base_model_path = "JunxiongWang/Llama3.2-Mamba2-3B-distill" # Replace with your base model path
+EAGLE_model_path = "ycwu97/mamba2-distilled-small" # Replace with your EAGLE weights path
+
+model = EaModel.from_pretrained(
+    base_model_path=base_model_path,
+    ea_model_path=EAGLE_model_path,
+    torch_dtype=torch.float16,
+    low_cpu_mem_usage=True,
+    device_map="auto",
+    total_token=-1 # -1 for auto configuration of draft tokens
+)
+model.eval()
+
+# Prepare your message using a conversation template (e.g., Vicuna)
+your_message="Hello"
+conv = get_conversation_template("vicuna") # Use the correct chat template for your base model
+conv.append_message(conv.roles[0], your_message)
+conv.append_message(conv.roles[1], None)
+prompt = conv.get_prompt()
+
+# Tokenize the input prompt
+input_ids = model.tokenizer([prompt]).input_ids
+input_ids = torch.as_tensor(input_ids).cuda()
+
+# Generate output using EAGLE's accelerated decoding
+output_ids = model.eagenerate(input_ids, temperature=0.5, max_new_tokens=512)
+
+# Decode and print the generated text
+output = model.tokenizer.decode(output_ids[0])
+print(output)
+```
+
+**Note:** For chat models like Vicuna, LLaMA2-Chat, and LLaMA3-Instruct, you must use the correct chat template to ensure proper model output and EAGLE's performance.