Set inputs to CPU for compatibility without GPU

Updated the input configuration to use the CPU instead of GPU, ensuring broader compatibility for environments lacking a GPU. This change enhances the application's accessibility, allowing users to run predictions without specific hardware requirements.

app.py

@@ -11,7 +11,7 @@ processor = AutoProcessor.from_pretrained("Qwen/Qwen2-VL-2B-Instruct")
 async def predict(messages: list):
     # Processamento e inferência
     text = processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
-    inputs = processor(text=[text], padding=True, return_tensors="pt").to("cuda")  # Altere para "cpu" se não tiver GPU
+    inputs = processor(text=[text], padding=True, return_tensors="pt").to("cpu")  # Altere para "cpu" se não tiver GPU
     generated_ids = model.generate(**inputs, max_new_tokens=128)
     output_text = processor.batch_decode(generated_ids, skip_special_tokens=True)
     return {"response": output_text}