feat: add ZeroGPU support via @spaces.GPU decorator

app.py

@@ -9,6 +9,7 @@ Deploy:
 """
 
 import os
+import spaces
 import gradio as gr
 import torch
 from transformers import AutoModelForCausalLM, AutoTokenizer
@@ -23,16 +24,13 @@ SYSTEM_PROMPT = (
     "You ALWAYS provide a response — never return empty output."
 )
 
-device = "cuda" if torch.cuda.is_available() else "mps" if torch.backends.mps.is_available() else "cpu"
-dtype = torch.float16 if device != "cpu" else torch.float32
-
-print(f"Loading {MODEL_ID} on {device}...")
+print(f"Loading {MODEL_ID}...")
 tokenizer = AutoTokenizer.from_pretrained(MODEL_ID, trust_remote_code=True)
 if tokenizer.pad_token is None:
     tokenizer.pad_token = tokenizer.eos_token
 
 model = AutoModelForCausalLM.from_pretrained(
-    MODEL_ID, torch_dtype=dtype, device_map="auto", trust_remote_code=True,
+    MODEL_ID, torch_dtype=torch.float16, device_map="auto", trust_remote_code=True,
 )
 
 if ADAPTER_ID:
@@ -43,7 +41,8 @@ model.eval()
 print("Model ready!")
 
 
-# ── Inference ───────────────────────────────────────────────────
+# ── Inference (GPU allocated only during this call) ─────────────
+@spaces.GPU(duration=120)
 def scan_code(language: str, code: str, max_tokens: int = 1024) -> str:
     if not code.strip():
         return "Paste some code to scan."