Update app.py

app.py

@@ -1,17 +1,16 @@
 from transformers import pipeline
 import gradio as gr
 import torch
-import os
 
-# Verify GPU support
+# Verify GPU
 if not torch.cuda.is_available():
-    raise RuntimeError("GPU not available - enable GPU in Space settings")
+    raise RuntimeError("Enable GPU in Space settings")
 
-# Load model with GPU acceleration
-math_pipeline = pipeline(
+# Load model (without bitsandbytes)
+model = pipeline(
     "text-generation",
     model="google/gemma-2b-it",
-    device=0,  # Force GPU usage
+    device=0,  # Force GPU
     torch_dtype=torch.float16,
     model_kwargs={
         "low_cpu_mem_usage": True,
@@ -22,25 +21,24 @@ math_pipeline = pipeline(
 def solve_math(question):
     prompt = f"Solve step by step:\nQ: {question}\nA:"
     try:
-        result = math_pipeline(
+        result = model(
             prompt,
-            max_new_tokens=150,
+            max_new_tokens=100,
             temperature=0.3,
-            do_sample=False,
-            pad_token_id=math_pipeline.tokenizer.eos_token_id
+            do_sample=False
         )
         return result[0]['generated_text'].split("A:")[-1].strip()
     except Exception as e:
         return f"Error: {str(e)}"
 
-# Preload model
+# Preload
 solve_math("2+2=")
 
-# Optimized UI
-with gr.Blocks(title="🚀 Math Solver") as demo:
-    gr.Markdown("## Enter a math problem:")
-    question = gr.Textbox(label="", placeholder="What is 2^8?")
-    answer = gr.Textbox(label="Solution", lines=3)
+# Simple UI
+with gr.Blocks() as demo:
+    gr.Markdown("## Math Solver")
+    question = gr.Textbox(label="Problem")
+    answer = gr.Textbox(label="Solution")
     question.submit(solve_math, question, answer)
 
-demo.launch(server_name="0.0.0.0")
+demo.launch()