Update app.py

app.py

@@ -225,7 +225,7 @@ class VisionEncoderDecoder(nn.Module):
                                    input_padding_mask=bool_padding_mask)
         return decoded_seq
 
-model = torch.load("caption_model.pth", weights_only=False)
+model = torch.load("caption_model.pth", weights_only=False,map_location=torch.device('cpu'))
 model.eval()
 tokenizer=AutoTokenizer.from_pretrained("distilbert-base-uncased")