Update app.py

app.py

@@ -1,70 +1,102 @@
+# app.py — Corrected for Hugging Face ZeroGPU Spaces
+# ---------------------------------------------------------------
+# This version is adapted for the ZeroGPU environment by using
+# the @spaces.GPU decorator.
+# ---------------------------------------------------------------
+import os
+import torch
 import gradio as gr
-from huggingface_hub import InferenceClient
+from transformers import pipeline
+import spaces # 1. Import the spaces library
 
+# ── Configuration ──────────────────────────────────────────────────────────────
+MODEL_ID = "Reubencf/gemma3-konkani"
+HF_TOKEN = os.getenv("HF_TOKEN")
 
-def respond(
-    message,
-    history: list[dict[str, str]],
-    system_message,
-    max_tokens,
-    temperature,
-    top_p,
-    hf_token: gr.OAuthToken,
-):
+TITLE = "🌴 Gemma Goan Q&A Bot (ZeroGPU)"
+DESCRIPTION = (
+    "Konkni uloi re babab.\n"
+    "ulo re uloi re"
+)
+# We define the pipeline object globally but initialize it inside the function
+pipe = None
+
+# ── Generation Function ──────────────────────────────────────────────────────
+@spaces.GPU(duration=60) # 2. Decorate the function that needs the GPU
+def generate_response(message, history):
     """
-    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
+    This function is called for each user message.
+    The @spaces.GPU decorator ensures a GPU is allocated when this runs.
     """
-    client = InferenceClient(token=hf_token.token, model="openai/gpt-oss-20b")
+    global pipe # Use the global pipe variable
 
-    messages = [{"role": "system", "content": system_message}]
+    # 3. Load the model inside the decorated function
+    # This ensures the model is loaded only when a GPU is active.
+    # We check if it's already loaded to avoid reloading on every call.
+    if pipe is None:
+        print(f"[Init] Loading model pipeline for the first time: {MODEL_ID}...")
+        try:
+            pipe = pipeline(
+                "text-generation",
+                model=MODEL_ID,
+                torch_dtype=torch.bfloat16,
+                device_map="auto", # This will now correctly map to the allocated GPU
+                token=HF_TOKEN,
+            )
+            print("[Init] Model pipeline loaded successfully.")
+        except Exception as e:
+            # If model loading fails, we can't proceed.
+            print(f"[Fatal] Could not load model: {e}")
+            return f"❌ Model failed to load: {e}"
 
-    messages.extend(history)
+    # Format the conversation history
+    conversation = []
+    for user_msg, assistant_msg in history:
+        conversation.append({"role": "user", "content": user_msg})
+        if assistant_msg:
+            conversation.append({"role": "assistant", "content": assistant_msg})
 
-    messages.append({"role": "user", "content": message})
+    # Add the current user's message
+    conversation.append({"role": "user", "content": message})
 
-    response = ""
+    # Apply the chat template
+    prompt = pipe.tokenizer.apply_chat_template(
+        conversation,
+        tokenize=False,
+        add_generation_prompt=True
+    )
 
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        choices = message.choices
-        token = ""
-        if len(choices) and choices[0].delta.content:
-            token = choices[0].delta.content
+    # Generate the response
+    outputs = pipe(
+        prompt,
+        max_new_tokens=256, # It's good practice to set a max token limit
+        do_sample=True,
+        temperature=0.7,
+        top_k=50,
+        top_p=0.95
+    )
+    
+    # Extract only the newly generated text
+    response = outputs[0]["generated_text"]
+    new_response = response[len(prompt):].strip()
 
-        response += token
-        yield response
+    return new_response
 
+# ── UI ────────────────────────────────────────────────────────────────────────
+examples = [
+    "Translate From English to Devnagri Konkani: what is color?",
+    "घरांत विजेचो वापर उणो करपाची येवजण तयार करप.",
+]
 
-"""
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
-"""
-chatbot = gr.ChatInterface(
-    respond,
-    type="messages",
-    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)",
-        ),
-    ],
+demo = gr.ChatInterface(
+    fn=generate_response,
+    title=TITLE,
+    description=DESCRIPTION,
+    examples=examples,
+    theme="soft",
 )
 
-with gr.Blocks() as demo:
-    with gr.Sidebar():
-        gr.LoginButton()
-    chatbot.render()
-
-
+# ── Launch ────────────────────────────────────────────────────────────────────
 if __name__ == "__main__":
+    print("🚀 Starting Gradio app for ZeroGPU...")
     demo.launch()