Update app.py

app.py

@@ -242,20 +242,20 @@ def captioning(img):
     output_prompt=[]
     # Initial caption generation without a prompt:
     inputsa = processor5(images=img, return_tensors="pt").to('cuda')
-    generated_ids = model5.generate(**inputsa, min_length=32, max_length=64)
+    generated_ids = model5.generate(**inputsa, min_length=24, max_length=42)
     generated_text = processor5.batch_decode(generated_ids, skip_special_tokens=True)[0].strip()
     print(generated_text)  
     # Loop through prompts array:
     for prompt in prompts_array:
         inputs = processor5(images=img, text=prompt, return_tensors="pt").to('cuda')
-        generated_ids = model5.generate(**inputs, min_length=32, max_length=64) # Adjust max_length if needed
+        generated_ids = model5.generate(**inputs, min_length=16, max_length=32) # Adjust max_length if needed
         generated_text = processor5.batch_decode(generated_ids, skip_special_tokens=True)[0].strip()
         response_text = generated_text.replace(prompt, "").strip() #Or could try .split(prompt, 1)[-1].strip()
         output_prompt.append(response_text)
         print(f"{response_text}\n") # Print only the response text
     # Continue conversation:
     inputf = processor5(images=img, text=generated_text + 'So therefore', return_tensors="pt").to('cuda')
-    generated_ids = model5.generate(**inputf, min_length=32, max_length=64)
+    generated_ids = model5.generate(**inputf, min_length=24, max_length=42)
     generated_text = processor5.batch_decode(generated_ids, skip_special_tokens=True)[0].strip()
     response_text = generated_text.replace(generated_text, "").strip()  # Remove the previous text plus 'So therefore'
     print(response_text)