Update app.py

app.py

@@ -49,11 +49,12 @@ def answer_question(image, question, max_crops, num_tokens):
 <|im_start|>assistant
 """
     streamer = TextIteratorStreamer(processor.tokenizer, skip_special_tokens=True)
-    inputs = processor(prompt, [image], model, max_crops=max_crops, num_tokens=num_tokens)
+    with torch.inference_mode():
+        inputs = processor(prompt, [image], model, max_crops=max_crops, num_tokens=num_tokens)
     generation_kwargs = {
         "input_ids": inputs["input_ids"],
         "attention_mask": inputs["attention_mask"],
-        "image_features": cached_vision_process(image, max_crops, num_tokens),
+        "image_features": inputs["image_features"],
         "streamer": streamer,
         "max_length": 1000,
         "use_cache": True,