Update app.py

app.py

@@ -242,13 +242,14 @@ 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()
+    output_prompt.append.(generated_text)
     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, max_length=64) # Adjust max_length if needed
+        generated_ids = model5.generate(**inputs, max_length=42) # 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)
@@ -259,7 +260,7 @@ def captioning(img):
   #  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)
-    output_prompt.append(response_text)
+    #output_prompt.append(response_text)
     print(output_prompt)
     return output_prompt
 '''
@@ -394,6 +395,7 @@ def generate_30(
         filename= f'rv_IP_{timestamp}.png'
         print("-- using image file --")
         print(caption)
+        print(caption_2)
         print("-- generating further caption --")
 
         #expand_prompt(prompt)