Update app.py

app.py

@@ -79,7 +79,7 @@ def process_image(model, tokenizer, transform, device, check_type, image, text):
         else:
             text_embeds = model.tok_embeddings(input_ids)
         
-        vit_embeds, size1 = model.forward_tokenocr(pixel_values.to(device))
+        vit_embeds, size1 = model.forward_tokenocr(pixel_values.to(torch.bfloat16).to(device))
         vit_embeds, size2 = post_process(vit_embeds, target_ratio, check_type)
         
         # 计算相似度