Update app.py

app.py

@@ -32,6 +32,9 @@ def transcribe_image(image):
         return_tensors="pt"
     ).to(device)
 
+    # FIX: Ensure the image tensor's dtype matches the model's dtype
+    inputs["pixel_values"] = inputs["pixel_values"].to(model.dtype)
+
     with torch.inference_mode():
         output = model.generate(
             **inputs,
@@ -67,6 +70,9 @@ def visual_question_answer(image, question):
         return_tensors="pt"
     ).to(device)
 
+    # FIX: Ensure the image tensor's dtype matches the model's dtype
+    inputs["pixel_values"] = inputs["pixel_values"].to(model.dtype)
+
     with torch.inference_mode():
         output = model.generate(
             **inputs,