Update app.py from anycoder

app.py

@@ -1,24 +1,28 @@
 import torch
 import gradio as gr
+import spaces
 from transformers import (
     Mistral3ForConditionalGeneration,
     MistralCommonBackend,
 )
 
-# Initialize model and tokenizer with CPU configuration
+# Initialize model and tokenizer with ZeroGPU configuration
 model_id = "mistralai/Devstral-Small-2-24B-Instruct-2512"
 
 # Load tokenizer
 tokenizer = MistralCommonBackend.from_pretrained(model_id)
 
-# Load model on CPU with appropriate configuration for zero-GPU
+# Load model with ZeroGPU compatibility
 model = Mistral3ForConditionalGeneration.from_pretrained(
     model_id,
-    device_map="cpu",  # Force CPU usage
-    torch_dtype=torch.float32,  # Use float32 for CPU compatibility
-    low_cpu_mem_usage=True  # Reduce memory usage during loading
+    torch_dtype=torch.float16,  # Use float16 for better GPU efficiency
+    low_cpu_mem_usage=True
 )
 
+# Move model to GPU when available
+if torch.cuda.is_available():
+    model = model.to('cuda')
+
 # System prompt
 SP = """You are operating as and within Mistral Vibe, a CLI coding-agent built by Mistral AI and powered by default by the Devstral family of models. It wraps Mistral's Devstral models to enable natural language interaction with a local codebase. Use the available tools when helpful.