Update app.py

app.py

@@ -41,12 +41,13 @@ def get_image_embedding(image):
             padding=True
         ).to(device, torch_dtype)
         
-        # Generate decoder_input_ids
+        # Generate decoder_input_ids with adjusted parameters
         decoder_input_ids = model.generate(
             **inputs,
-            max_length=1,
+            max_new_tokens=20,  # Increased from max_length
             min_length=1,
             num_beams=1,
+            do_sample=False,
             pad_token_id=processor.tokenizer.pad_token_id,
             return_dict_in_generate=True,
         ).sequences
@@ -55,7 +56,6 @@ def get_image_embedding(image):
         
         with torch.no_grad():
             outputs = model(**inputs)
-            # Use the mean of the last hidden state as the embedding
             image_embeddings = outputs.last_hidden_state.mean(dim=1)
         return image_embeddings.cpu().numpy()
     except Exception as e:
@@ -75,12 +75,13 @@ def get_text_embedding(text):
             padding=True
         ).to(device, torch_dtype)
         
-        # Generate decoder_input_ids
+        # Generate decoder_input_ids with adjusted parameters
         decoder_input_ids = model.generate(
             **inputs,
-            max_length=1,
+            max_new_tokens=20,  # Using max_new_tokens instead of max_length
             min_length=1,
             num_beams=1,
+            do_sample=False,
             pad_token_id=processor.tokenizer.pad_token_id,
             return_dict_in_generate=True,
         ).sequences