Update app.py

app.py

@@ -1,40 +1,38 @@
 #!/usr/bin/env python3
 # -*- coding: utf-8 -*-
-"""
-Gemma-3 Iraqi Arabic Chat App
-"""
-
-import os
 import torch
 from transformers import AutoModelForCausalLM, AutoTokenizer
 import gradio as gr
+import spaces  # مهم جداً في Hugging Face Spaces
 
 MODEL_NAME = "anaspro/gemma3-iraqi"
 
-print("🔐 Loading model:", MODEL_NAME)
+# نحدد أن هذا التطبيق يحتاج GPU لتشغيله
+@spaces.GPU
+def load_model():
+    print("🔐 Loading model:", MODEL_NAME)
+    tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
+    model = AutoModelForCausalLM.from_pretrained(
+        MODEL_NAME,
+        torch_dtype=torch.bfloat16 if torch.cuda.is_available() else torch.float32,
+        device_map="auto",
+    )
+    print("✅ Model loaded and ready!")
+    return tokenizer, model
 
-tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
-model = AutoModelForCausalLM.from_pretrained(
-    MODEL_NAME,
-    torch_dtype=torch.bfloat16 if torch.cuda.is_available() else torch.float32,
-    device_map="auto",
-)
-print("✅ Model loaded and ready!")
+tokenizer, model = load_model()
 
 def chat_fn(user_message, history):
-    # سجل رسالة المستخدم
     history = history or []
     history.append((user_message, None))
-    
-    # إعداد الإدخال
+
     inputs = tokenizer.apply_chat_template(
         [{"role": "user", "content": user_message}],
         tokenize=True,
         add_generation_prompt=True,
         return_tensors="pt"
     ).to(model.device)
-    
-    # توليد الرد
+
     outputs = model.generate(
         **inputs,
         max_new_tokens=1024,
@@ -45,22 +43,25 @@ def chat_fn(user_message, history):
         eos_token_id=tokenizer.eos_token_id,
     )
     resp = tokenizer.decode(outputs[0], skip_special_tokens=True)
-    
-    # اقتطاع الجزء بعد “assistant” إذا موجود
     if "assistant" in resp:
         resp = resp.split("assistant")[-1].strip()
-    
-    # سجل الرد
+
     history[-1] = (user_message, resp)
     return history, history
 
 with gr.Blocks() as demo:
-    gr.Markdown("## 🗣️ Iraqi Arabic Chat with Gemma-3\nاكتب باللهجة العراقية أي سؤال وتفاعل.")
-    chatbot = gr.Chatbot(label="Chatbot")
-    msg = gr.Textbox(placeholder="اكتب نصك هنا...", lines=1)
-    clear = gr.Button("🧹 مسح المحادثة")
-    
+    gr.Markdown("## 🧠 Chat with Iraqi Gemma 3 – (GPU Optimized)")
+    chatbot = gr.Chatbot(label="💬 Chatbot", type="messages")
+    msg = gr.Textbox(placeholder="اكتب باللهجة العراقية...")
+    clear = gr.Button("🧹 مسح")
+
     msg.submit(chat_fn, [msg, chatbot], [chatbot, chatbot])
     clear.click(lambda: None, None, chatbot, queue=False)
-    
-demo.launch(server_name="0.0.0.0", server_port=7860, share=True)
+
+# Spaces يعرف أن نقطة البداية هي GPU
+@spaces.GPU
+def main():
+    demo.launch(server_name="0.0.0.0", server_port=7860)
+
+if __name__ == "__main__":
+    main()