Improve model card: Add pipeline tag, license, and sample usage

README.md

@@ -1,17 +1,57 @@
 ---
 base_model: Qwen/Qwen2.5-1.5B-Instruct
-library_name: peft
+library_name: transformers
+license: apache-2.0
+pipeline_tag: text-generation
 ---
 
 # Model Summary
 
-<!-- Provide a quick summary of what the model is/does. -->
+This model, `agent-distillation/agent_distilled_Qwen2.5-1.5B-Instruct`, is a distilled version of `Qwen2.5-1.5B-Instruct`. It has been trained on agent trajectories derived from the [agent-distillation/Qwen2.5-32B-Instruct_agent_trajectories_2k dataset](https://huggingface.co/datasets/agent-distillation/Qwen2.5-32B-Instruct_agent_trajectories_2k).
 
-This model is distilled Qwen2.5-1.5B-Instruct from agent trajectories in this [dataset](https://huggingface.co/datasets/agent-distillation/Qwen2.5-32B-Instruct_agent_trajectories_2k).
+The model was presented in the paper [Distilling LLM Agent into Small Models with Retrieval and Code Tools](https://arxiv.org/abs/2505.17612). It focuses on transferring complex reasoning and full task-solving behavior from LLM-based agents into smaller language models (sLMs) by integrating retrieval and code execution capabilities. The method employs techniques like "first-thought prefix" for enhanced teacher-generated trajectories and "self-consistent action generation" for improved robustness in small agents.
 
-- Repository: https://github.com/Nardien/agent-distillation
-- Paper: https://arxiv.org/abs/2505.17612
+-   **Repository**: https://github.com/Nardien/agent-distillation
+-   **Paper**: [Distilling LLM Agent into Small Models with Retrieval and Code Tools](https://arxiv.org/abs/2505.17612)
+-   **Project Page/Related Models**: Explore other models and datasets from this project on the [agent-distillation Hugging Face organization page](https://huggingface.co/agent-distillation).
 
 ### Framework versions
 
-- PEFT 0.15.1
+-   PEFT 0.15.1
+-   Transformers
+
+## Sample Usage
+
+You can quickly try out the distilled 1.5B agent from the Hugging Face Hub using the `smolagents` library, which is introduced in the project's GitHub repository. The `smolagents` library itself builds upon the Hugging Face `transformers` library.
+
+First, ensure `smolagents` and its dependencies are installed as per the [GitHub repository's instructions](https://github.com/Nardien/agent-distillation#installation).
+
+```python
+from smolagents import LlmAgent
+import os
+
+# Initialize the agent
+# The smolagents library builds upon the Hugging Face Transformers library.
+# This model is a PEFT adapter for Qwen2.5-1.5B-Instruct.
+# If you encounter issues with authentication or rate limits, ensure you are logged in
+# or replace "YOUR_HF_TOKEN" with your actual Hugging Face token.
+# You can set it as an environment variable or pass it directly.
+# For example: HUGGING_FACE_HUB_TOKEN="hf_..." python your_script.py
+# Or: agent = LlmAgent.from_pretrained("agent-distillation/agent_distilled_Qwen2.5-1.5B-Instruct", token=os.environ.get("HUGGING_FACE_HUB_TOKEN"))
+agent = LlmAgent.from_pretrained(
+    "agent-distillation/agent_distilled_Qwen2.5-1.5B-Instruct",
+    # Additional arguments like `device_map="auto"` or `trust_remote_code=True`
+    # can be passed here if required by the underlying transformers model loading.
+)
+
+# Chat loop
+print("Agent initialized. Type 'exit' to quit.")
+while True:
+    user_input = input("You: ")
+    if user_input.lower() == 'exit':
+        break
+    # The agent will use its configured tools (e.g., retrieval, code execution)
+    # as needed to respond to the query.
+    response = agent.chat(user_input)
+    print(f"Agent: {response.text}")
+```