Update app.py

app.py

@@ -2,37 +2,22 @@ import spaces
 import gradio as gr
 from transformers import AutoModelForCausalLM, AutoTokenizer
 
-# Load model and tokenizer
+# Model path
 model_name = "krish10/Qwen3_0.6B_16bit_TA_screen"
+
+# Load model and tokenizer
 tokenizer = AutoTokenizer.from_pretrained(model_name)
 model = AutoModelForCausalLM.from_pretrained(model_name).to("cuda")
 
-# Chat function
+# Raw text-generation function (no chat formatting)
 @spaces.GPU
-def respond(message, history: list[tuple[str, str]], system_message, max_tokens, temperature, top_p):
-    # Format message history
-    messages = []
-    if system_message:
-        messages.append({"role": "system", "content": system_message})
-    for user_msg, bot_msg in history:
-        messages.append({"role": "user", "content": user_msg})
-        messages.append({"role": "assistant", "content": bot_msg})
-    messages.append({"role": "user", "content": message})
-
-    # Format prompt with Qwen's template
-    prompt = tokenizer.apply_chat_template(
-        messages,
-        tokenize=False,
-        add_generation_prompt=True
-    )
-
-    # Optional: print prompt for debugging
-    print("PROMPT:\n", prompt)
+def respond(message, _, __, max_tokens, temperature, top_p):
+    prompt = message  # Use message as-is
 
-    # Tokenize and send to GPU
+    # Tokenize
     inputs = tokenizer(prompt, return_tensors="pt").to("cuda")
 
-    # Generate response
+    # Generate
     outputs = model.generate(
         input_ids=inputs["input_ids"],
         max_new_tokens=max_tokens,
@@ -44,31 +29,27 @@ def respond(message, history: list[tuple[str, str]], system_message, max_tokens,
 
     # Decode and strip prompt
     decoded = tokenizer.decode(outputs[0], skip_special_tokens=True)
-    reply = decoded[len(prompt):]
-    return reply
+    response = decoded[len(prompt):]
+    return response
 
 # Build Gradio UI
 with gr.Blocks() as demo:
-    gr.Markdown("## 🧠 Qwen Chat Interface (with proper chat template)")
+    gr.Markdown("## 🧠 Structured Evaluation Chat (No Template, Matches Fine-Tuning)")
 
-    chatbot = gr.Chatbot()
-    msg = gr.Textbox(label="Enter your message")
-    system_msg = gr.Textbox(value="", label="System message (optional)")
-    max_tokens = gr.Slider(1, 16384, value=4000, step=1, label="Max new tokens")
+    msg = gr.Textbox(lines=15, label="Input your instruction + abstract (exact format as in Colab)")
+    system_msg = gr.Textbox(visible=False)  # ignored
+    max_tokens = gr.Slider(1, 4096, value=512, step=1, label="Max new tokens")
     temperature = gr.Slider(0.1, 4.0, value=0.7, step=0.1, label="Temperature")
     top_p = gr.Slider(0.1, 1.0, value=0.9, step=0.05, label="Top-p")
 
-    state = gr.State([])  # for message history
+    output = gr.Textbox(lines=15, label="Model response")
 
-    def user_submit(user_message, history, sys_msg, max_tokens, temp, top_p):
-        response = respond(user_message, history, sys_msg, max_tokens, temp, top_p)
-        history.append((user_message, response))
-        return history, history
+    btn = gr.Button("Generate")
 
-    msg.submit(
-        user_submit,
-        inputs=[msg, state, system_msg, max_tokens, temperature, top_p],
-        outputs=[chatbot, state]
+    btn.click(
+        fn=respond,
+        inputs=[msg, system_msg, None, max_tokens, temperature, top_p],
+        outputs=[output]
     )
 
 # Launch app