Update app.py

app.py

@@ -1,64 +1,116 @@
+import torch
+import pandas as pd
+import re
+import time
+from transformers import pipeline, AutoTokenizer, AutoModelForCausalLM
 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
-"""
-client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
-
-
-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
-
-
-"""
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
-"""
-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)",
-        ),
-    ],
+
+# Disable torch._dynamo
+import torch._dynamo
+torch._dynamo.disable()
+
+# Load model and tokenizer
+model_path = "supp03/E3L-AI-CHATBOT"  
+tokenizer = AutoTokenizer.from_pretrained(model_path)
+model = AutoModelForCausalLM.from_pretrained(
+    model_path,
+    torch_dtype=torch.float16,
+    device_map="auto"
 )
 
+# Load pipeline
+pipe = pipeline(
+    "text-generation",
+    model=model,
+    tokenizer=tokenizer,
+    pad_token_id=tokenizer.eos_token_id
+)
+
+# Context memory
+context_memory = {
+    "last_question": None,
+    "last_answer": None
+}
+
+# Main inference function
+def chat(user_input):
+    user_input = user_input.strip()
+    if not user_input:
+        return "Please enter a question."
+
+    if user_input.lower() == 'exit':
+        return "Exiting. Reload the app to start over."
+
+    exact_answer = get_exact_answer(user_input)
+    if exact_answer:
+        context_memory["last_question"] = user_input
+        context_memory["last_answer"] = exact_answer
+        return f"[Exact Match Found]\n\n{exact_answer}"
+
+    # Determine context reference
+    lower_input = user_input.lower()
+    refer_prev = any(phrase in lower_input for phrase in [
+        "previous question", "previous answer", "elaborate", "detail", "more info", "explain more", "expand on that"
+    ])
+
+    context = ""
+    if refer_prev and context_memory["last_question"] and context_memory["last_answer"]:
+        context = (
+            f"Previous Question: {context_memory['last_question']}\n"
+            f"Previous Answer: {context_memory['last_answer']}\n\n"
+            "Please elaborate on the previous answer with more details, examples, and explanations to deepen understanding.\n\n"
+        )
+
+    if "steps" in lower_input or "step by step" in lower_input:
+        style_instruction = "Organize your response into a clearly numbered step-by-step list (1., 2., 3., ...)."
+    elif "how" in lower_input or "ways" in lower_input:
+        style_instruction = "Organize your response into practical bullet points."
+    else:
+        style_instruction = "Write in a natural, conversational paragraph form."
+
+    generated_prompt = (
+        "You are a warm, thoughtful, and emotionally intelligent mentor. "
+        "Your job is to provide deeply personalized, practical, and supportive guidance. "
+        "Avoid sounding robotic or generic. Respond with empathy, clarity, and human warmth, as if you're speaking directly to the person. "
+        f"{style_instruction}\n\n"
+        f"{context}"
+        f"Question: {user_input}\n\nAnswer:"
+    )
+
+    start_time = time.time()
+    with torch.no_grad():
+        generated = pipe(
+            generated_prompt,
+            max_new_tokens=300,
+            do_sample=True,
+            top_p=0.9,
+            temperature=0.7
+        )
+    elapsed_time = time.time() - start_time
+
+    output_text = generated[0]["generated_text"]
+    match = re.search(r"Answer:\s*(.*)", output_text, re.DOTALL)
+    model_answer = match.group(1).strip() if match else output_text.strip()
+
+    for kw in ["your students", "may also want", "in order to"]:
+        if kw in model_answer:
+            model_answer = model_answer.split(kw)[0].strip()
+            break
+
+    context_memory["last_question"] = user_input
+    context_memory["last_answer"] = model_answer
+
+    seconds = int(elapsed_time % 60)
+    return f"[Response Time]: {seconds}s\n\n{model_answer}"
+
+# Gradio UI
+iface = gr.Interface(
+    fn=chat,
+    inputs=gr.Textbox(label="Ask your question", placeholder="Type your question here..."),
+    outputs="markdown",
+    title="Gemma Mentor Chatbot",
+    description="Ask questions and get warm, supportive, emotionally intelligent responses. Type 'exit' to reset."
+)
 
 if __name__ == "__main__":
-    demo.launch()
+    iface.launch()