Update app.py

app.py

@@ -250,14 +250,14 @@ 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=42, max_length=64)
+    generated_ids = model5.generate(**inputsa, min_length=42, max_length=128)
     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, min_length=16, max_length=64) # Adjust max_length if needed
+        generated_ids = model5.generate(**inputs, min_length=32, max_length=64) # 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)
@@ -271,6 +271,9 @@ def captioning(img):
     #output_prompt.append(response_text)
     print(output_prompt)
     return output_prompt
+    
+def flatten_and_stringify(data):
+    return [str(item) for sublist in data if isinstance(sublist, list) for item in flatten_and_stringify(sublist) ] + [str(item) for item in data if not isinstance(item, list)]
 
 def expand_prompt(prompt):
         system_prompt_rewrite = (
@@ -309,7 +312,7 @@ def expand_prompt(prompt):
         outputs_2 = model.generate(
             input_ids=input_ids_2,
             attention_mask=attention_mask_2,
-            max_new_tokens=512,
+            max_new_tokens=1024,
             temperature=0.2,
             top_p=0.9,
             do_sample=True,
@@ -404,6 +407,10 @@ def generate_30(
         print("-- using image file --")
         print(caption)
         print(caption_2)
+        caption = flatten_and_stringify(caption)
+        caption = " ".join(caption)
+        caption_2 = flatten_and_stringify(caption_2)
+        caption_2 = " ".join(caption_2)
         print("-- generating further caption --")
         global model5
         global processor5