Update app.py

app.py

@@ -1,97 +1,83 @@
-import gradio as gr
-from huggingface_hub import InferenceClient
-
-
-"""
-For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
-"""
-
 import os
-client = InferenceClient(model="HuggingFaceH4/zephyr-7b-beta", token=os.getenv("HUGGINGFACEHUB_API_TOKEN"))
-
-
-def respond(
-    message,
-    history: list[tuple[str, str]],
-    system_message,
-    max_tokens,
-    temperature,
-    top_p,
-):
-    messages = [{"role": "system", "content": system_message}]
-
-    for val in history:
-        if val[0]:
-            messages.append({"role": "user", "content": val[0]})
-        if val[1]:
-            messages.append({"role": "assistant", "content": val[1]})
-
-    messages.append({"role": "user", "content": message})
-
-    response = ""
-
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        token = message.choices[0].delta.content
-
-        response += token
-        yield response
-
+import torch
+import gradio as gr
+from transformers import AutoModelForCausalLM, AutoTokenizer
+from peft import PeftModel
 
 """
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
+Root_Math fine-tuned model chat app for Hugging Face Spaces.
 """
-demo = gr.ChatInterface(
-    respond,
-    additional_inputs=[
-        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
-        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
-        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
-        gr.Slider(
-            minimum=0.1,
-            maximum=1.0,
-            value=0.95,
-            step=0.05,
-            label="Top-p (nucleus sampling)",
-        ),
-    ],
-)
-
-
-if __name__ == "__main__":
-    demo.launch()
-
-import os
-from transformers import AutoModelForCausalLM, AutoTokenizer
-from peft import PeftModel
-import torch
 
-# Load Hugging Face API token securely
+# ✅ Load Hugging Face API token securely
 api_token = os.getenv("HUGGINGFACEHUB_API_TOKEN")
-
 if not api_token:
     raise ValueError("❌ ERROR: Hugging Face API token is not set. Please set it as an environment variable.")
 
-# Define model names
+# ✅ Define model names
 base_model_name = "unsloth/qwen2.5-math-7b-bnb-4bit"
-peft_model_name = "Hrushi02/Root_Math"
+peft_model_name = "Hrushi02/Root_Math"  # <-- stays the same
 
-# Load base model with authentication
+# ✅ Load base model
+print("🔄 Loading base model...")
 base_model = AutoModelForCausalLM.from_pretrained(
     base_model_name,
     torch_dtype=torch.float16,
     device_map="auto",
-    use_auth_token=api_token  # ✅ Correct
+    use_auth_token=api_token
 )
 
-
-# Load fine-tuned model
+# ✅ Load your fine-tuned PEFT adapter
+print("🔄 Loading fine-tuned adapter...")
 model = PeftModel.from_pretrained(base_model, peft_model_name, token=api_token)
 
-# Load tokenizer
+# ✅ Load tokenizer
+print("🔄 Loading tokenizer...")
 tokenizer = AutoTokenizer.from_pretrained(base_model_name, token=api_token)
+
+# ✅ Define chat response function
+def respond(message, history, system_message, max_tokens, temperature, top_p):
+    """Generate responses from your fine-tuned model."""
+    full_prompt = system_message + "\n\n"
+    for user_msg, bot_msg in history:
+        if user_msg:
+            full_prompt += f"User: {user_msg}\n"
+        if bot_msg:
+            full_prompt += f"Assistant: {bot_msg}\n"
+    full_prompt += f"User: {message}\nAssistant:"
+
+    inputs = tokenizer(full_prompt, return_tensors="pt").to(model.device)
+
+    with torch.no_grad():
+        outputs = model.generate(
+            **inputs,
+            max_new_tokens=max_tokens,
+            temperature=temperature,
+            top_p=top_p,
+            do_sample=True
+        )
+
+    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
+    # extract only the assistant's last message
+    if "Assistant:" in response:
+        response = response.split("Assistant:")[-1].strip()
+
+    yield response
+
+
+# ✅ Create Gradio interface
+demo = gr.ChatInterface(
+    respond,
+    additional_inputs=[
+        gr.Textbox(value="You are a helpful math assistant.", label="System message"),
+        gr.Slider(minimum=1, maximum=1024, value=256, step=1, label="Max new tokens"),
+        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
+        gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p (nucleus sampling)"),
+    ],
+    title="🧮 Root Math Assistant",
+    description="A fine-tuned math reasoning model by Hrushi02 using Unsloth + PEFT."
+)
+
+
+# ✅ Launch app
+if __name__ == "__main__":
+    demo.launch()