Update

app.py

@@ -85,12 +85,11 @@ def process(text: str) -> str:
         add_special_tokens=True, 
         return_tensors="pt"
     ).input_ids
-    with torch.no_grad():
-        output = model.generate(
-            input_ids.to(zero.device), # GPU
-            do_sample=False,
-            max_new_tokens=512,
-        )
+    output = model.generate(
+        input_ids.to(zero.device), # GPU
+        do_sample=False,
+        max_new_tokens=512,
+    )
     return tokenizer.decode(output[0], skip_special_tokens=True)[len(prompt):].strip()