Update app.py

app.py

@@ -69,26 +69,12 @@ def warmup(progress=gr.Progress(track_tqdm=True)):
 def _build_inputs(processor, tokenizer, image: Image.Image, prompt: str):
     messages = [{"role": "user", "content": [{"type": "image", "image": image}, {"type": "text", "text": prompt}]}]
     
-    # We explicitly set max_length and truncation here to resolve the token mismatch error.
-    # A value of 2048 is safe, as an image takes up ~1024 tokens.
-    max_len_val = 2048
-
     if tokenizer and hasattr(tokenizer, "apply_chat_template"):
         chat_prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
-        return processor(
-            text=[chat_prompt],
-            images=[image],
-            return_tensors="pt",
-            max_length=max_len_val,
-            truncation=True
-        )
-    return processor(
-        text=[prompt],
-        images=[image],
-        return_tensors="pt",
-        max_length=max_len_val,
-        truncation=True
-    )
+        return processor(text=[chat_prompt], images=[image], return_tensors="pt")
+        
+    return processor(text=[prompt], images=[image], return_tensors="pt")
+
 
 def _decode_text(model, processor, tokenizer, output_ids, prompt: str):
     try: