Update app_dialogue.py

app_dialogue.py

@@ -70,13 +70,20 @@ import tempfile
 
 
 def convert_to_rgb_pil(image):
-    print(f"***** convert_to_rgb_pil ******")
-    print(f"params: image is - {image}")
-    #if image.mode == "RGB":
-    #    return image
+    """
+    Convert a PIL Image object to RGB mode and save it locally.
+
+    The function ensures that images with transparency (alpha channel)
+    are overlaid on a white background before saving.
+
+    Parameters:
+    - image (PIL.Image.Image): The input image to be processed.
+
+    Returns:
+    - str: The path to the saved RGB image.
+
+    """
     # Save the converted image to a temporary file
-    #temp_file = tempfile.NamedTemporaryFile(delete=False, suffix=".jpg")
-    #temp_file_path = temp_file.name
     filename = f"{uuid.uuid4()}.jpg"
     local_path = f"{filename}"
     
@@ -89,16 +96,30 @@ def convert_to_rgb_pil(image):
     else:
         image.save(local_path)
         
-    #temp_file.close()
-    print(f"# Return the path to the saved image as - {local_path}")
     return local_path  # Return the path to the saved image
     
 
 def convert_to_rgb(filepath_or_pilimg):
+    """
+    Convert an image to RGB mode, handling transparency for non-RGB images.
+
+    This function can accept either a file path to an image or a PIL Image object.
+    For transparent images, the function overlays the image onto a white background 
+    to handle the transparency before converting it to RGB mode.
+
+    Parameters:
+    - filepath_or_pilimg (str or PIL.Image.Image): The file path to an image or a PIL 
+                                                  Image object to be processed.
+
+    Returns:
+    - str: If the input was a file path, the return will be the path to the original 
+           image (if it's already in RGB) or the path to the saved RGB image.
+           If the input was a PIL Image object, the return will be the path to the saved 
+           RGB image.
+
+    """
     # `image.convert("RGB")` would only work for .jpg images, as it creates a wrong background
     # for transparent images. The call to `alpha_composite` handles this case
-    print(f"***** convert_to_rgb ******")
-    print(f"params: image is - {filepath_or_pilimg}")
     
     if isinstance(filepath_or_pilimg, PIL.Image.Image):
         return convert_to_rgb_pil(filepath_or_pilimg)
@@ -123,26 +144,15 @@ def convert_to_rgb(filepath_or_pilimg):
         # Save the converted image to a temporary file
         filename = f"{uuid.uuid4()}.jpg"
         local_path = f"{filename}"
-        #temp_file = tempfile.NamedTemporaryFile(delete=False, suffix=".jpg")
-        #temp_file_path = temp_file.name
         alpha_composite.save(local_path)
-        #temp_file.close()
 
-        print(f"# Return the path to the saved image as - {local_path}")
         return local_path  # Return the path to the saved image
 
+# XXXX REMOVE
 def pil_to_markdown_im(image):
     """
     Convert a PIL image into markdown filled with the base64 string representation.
     """
-    print(f"***** pil_to_markdown_im ******")
-    print(f"params: image is - {image}")
-    #if isinstance(image, PIL.Image.Image):
-    #img_b64_str = pil_to_base64(image)
-    #img_str = f'<img src="data:image/png;base64,{img_b64_str}" />'
-    #if path_or_url.startswith(("http://", "https://")):
-    #response = requests.get(image)
-    #image = Image.open(BytesIO(response.content))
     # Generate a unique filename using UUID
     filename = f"{uuid.uuid4()}.jpg"
     local_path = f"{filename}" 
@@ -156,7 +166,7 @@ def base64_to_pil(encoded_image):
     pil_image = Image.open(BytesIO(decoded_image))
     return pil_image
 
-
+# XXXXX REMOVE
 def im_markdown_to_pil(im_markdown_str):
     pattern = r'<img src="data:image/png;base64,([^"]+)" />'
     match = re.search(pattern, im_markdown_str)
@@ -218,33 +228,20 @@ def isolate_images_urls(prompt_list):
     ]
     ```
     """
-    print(f"******* isolate_images_urls *******")
-    print(f"params: prompt_list is - {prompt_list}")
     
     linearized_list = []
     for prompt in prompt_list:
-        print(f"inside FOR loop: prompt in prompt_list is - {prompt}")
         # Prompt can be either a string, or a PIL image
         if isinstance(prompt, PIL.Image.Image):
-            print(f"inside first IF in FOR loop: prompt is of type PIL.Image.Image")
             linearized_list.append(prompt)
-            print(f"linearized_list after append is - {linearized_list}")
         elif isinstance(prompt, str) and "/tmp/gradio/" in prompt: #isinstance(prompt, PIL.Image.Image):
-            print(f"inside IF in FOR loop: prompt is a string and is a path for temporary file")
             linearized_list.append(prompt)
-            print(f"linearized_list after append is - {linearized_list}")
         elif isinstance(prompt, str) and "/tmp/gradio/" not in prompt:
-            print(f"inside ELIF in FOR loop: prompt is a string and is NOT a path for temporary file")
             if "<fake_token_around_image>" not in prompt:
-                print(f"inside IF inside ELIF in FOR loop: '<fake_token_around_image>' is NOT in prompt")
                 linearized_list.append(prompt)
-                print(f"linearized_list after append is - {linearized_list}")
             else:
-                print(f"inside ELSE inside ELIF in FOR loop: '<fake_token_around_image>' IS IN prompt")
                 prompt_splitted = prompt.split("<fake_token_around_image>")
-                print(f"prompt_splitted is - {prompt_splitted}")
                 for ps in prompt_splitted:
-                    print(f"Inside FOR loop inside FOR loop: ps in prompt_split is {ps}")
                     if ps == "":
                         continue
                     if ps.startswith("<image:"):
@@ -256,7 +253,6 @@ def isolate_images_urls(prompt_list):
                 f"Unrecognized type for `prompt`. Got {type(type(prompt))}. Was expecting something in [`str`,"
                 " `PIL.Image.Image`]"
             )
-            print(f"linearized_list to be returned is - {linearized_list}")
     return linearized_list
 
 
@@ -285,31 +281,20 @@ def user_prompt_list_to_markdown(user_prompt_list: List[Union[str, PIL.Image.Ima
     Convert a user prompt in the list format (i.e. elements are either a PIL image or a string) into
     the markdown format that is used for the chatbot history and rendering.
     """
-    print("********** user_prompt_list_to_markdown *********")
-    print(f" param : user_prompt_list is - {user_prompt_list}")
     resulting_string = ""
     for elem in user_prompt_list:
-        print(f"inside user_prompt_list_to_markdown, for loop on user_prompt_list")
-        print(f"elem is - {elem} ")
         if isinstance(elem, str):
             if "/tmp/gradio/" not in elem:
                 resulting_string += elem
-                print(f"inside IF - when elem is string and is not temp image filepath. resulting_string is - {resulting_string}")
             elif "/tmp/gradio/" in elem:            
                 resulting_string += f"![](/file={convert_to_rgb(elem)})"
-                print(f"inside IF - when elem is string and is a temp image filepath. resulting_string is - {resulting_string}")       
-        #elif isinstance(elem, str) and "/tmp/gradio/" in elem:
-        #    resulting_string += f"![](/file={convert_to_rgb(elem)})"  #f"![](/file={image})" 
-        #    print(f"inside first ELIF - when elem is string and is the temp image filepath. resulting_string is - {resulting_string}")
         elif isinstance(elem, PIL.Image.Image): #or "/tmp/gradio/" in elem:  #and "/tmp/gradio/" in elem:
             resulting_string += f"![](/file={convert_to_rgb(elem)})" #pil_to_markdown_im(convert_to_rgb(elem)) <---------------
-            print(f"inside the ELIF - when elem is an instance of  PIL.Image.Image. The resulting_string after convert_to_rgb() operation is - {resulting_string}")
         else:
             raise ValueError(
                 "Unknown type for `user_prompt_list`. Expected an element of type `str` or `PIL.Image.Image` and got"
                 f" `{type(elem)}`"
             )
-    print(f" final resulting_string that will be returned is - {resulting_string}")
     return resulting_string
 
 
@@ -348,8 +333,6 @@ def load_processor_tokenizer_model(model_name):
         max_memory=max_memory_map,
     )
     model.eval()
-    print("Current device map:", model.hf_device_map)
-    print("Model default generation config:", model.generation_config)
     # TODO: the device_map looks very inefficien right now. that could be improved
     return processor, tokenizer, model
 
@@ -361,60 +344,41 @@ def format_user_prompt_with_im_history_and_system_conditioning(
     Produces the resulting list that needs to go inside the processor.
     It handles the potential image box input, the history and the system conditionning.
     """
-    print(f"*********format_user_prompt_with_im_history_and_system_conditioning*********")
-    print(f"format_user_prompt_with_im_history_and_system_conditioning -- param current_user_prompt_str is - {current_user_prompt_str} ")
-    print(f"format_user_prompt_with_im_history_and_system_conditioning -- param current_image is - {current_image} ")
-    print(f"format_user_prompt_with_im_history_and_system_conditioning -- param history is - {history} ")
     
     resulting_list = copy.deepcopy(SYSTEM_PROMPT)
 
     # Format history
     for turn in history:
-        print(f"inside for loop, turn is - {turn}")
         user_utterance, assistant_utterance = turn
-        print("calling split_str_on_im_markdown from inside for loop inside format_user_prompt_with_im_history_and_system_conditioning")
         splitted_user_utterance = split_str_on_im_markdown(user_utterance)
-        print(f"splitted_user_utterance from split_str_on_im_markdown is - {splitted_user_utterance} ")
-        splitted_user_utterance = [
-            im_markdown_to_pil(s) if s.startswith('<img src="data:image/png;base64,') else s
-            for s in splitted_user_utterance
-            if s != ""
-        ]
-        print(f"splitted_user_utterance after im_markdown_to_pil() is - {splitted_user_utterance} ")
+        #splitted_user_utterance = [
+        #    im_markdown_to_pil(s) if s.startswith('<img src="data:image/png;base64,') else s
+        #    for s in splitted_user_utterance
+        #    if s != ""
+        #]
         
         if isinstance(splitted_user_utterance[0], str):
             resulting_list.append("\nUser: ")
         else:
             resulting_list.append("\nUser:")
-        print(f"resulting_list after if..else block is - {resulting_list}")
         resulting_list.extend(splitted_user_utterance)
-        print(f"resulting_list after extend is - {resulting_list}")
         resulting_list.append(f"<end_of_utterance>\nAssistant: {assistant_utterance}")
-        print(f"resulting_list after append is - {resulting_list}")
         
 
     # Format current input
     current_user_prompt_str = remove_spaces_around_token(current_user_prompt_str)
-    print(f"current_user_prompt_str is - {current_user_prompt_str}")
     
     if current_image is None:
-        print("inside IF : current_image is NONE")
         if "<img src=data:image/png;base64" in current_user_prompt_str:
             raise ValueError("The UI does not support inputing via the text box an image in base64.")
         current_user_prompt_list = handle_manual_images_in_user_prompt(current_user_prompt_str)
-        print(f"current_user_prompt_list (or [user_prompt]/resulting_user_prompt((most likely this one)) from handle_manual_images_in_user_prompt ) is - {current_user_prompt_list}")
         resulting_list.append("\nUser: ")
-        print(f"resulting_list with append user - {resulting_list}")
         resulting_list.extend(current_user_prompt_list)
-        print(f"resulting_list after extend with current_user_prompt_list is - {resulting_list}")
         resulting_list.append("<end_of_utterance>\nAssistant:")
-        print(f"resulting_list after append with end_of_utteranceAssistant is - {resulting_list}")
         return resulting_list, current_user_prompt_list
     else:
-        print("inside ELSE : current_image is not NONE")
-        # Choosing to put the image first when the image is inputted through the UI, but this is an arbiratrary choice.
+        # Choosing to put the image first when the image is inputted through the UI, but this is an arbitrary choice.
         resulting_list.extend(["\nUser:", Image.open(current_image), f"{current_user_prompt_str}<end_of_utterance>\nAssistant:"]) #current_image
-        print(f"final resulting_list passed on to calling function is - {resulting_list}")
         return resulting_list, [current_user_prompt_str]
 
 
@@ -836,11 +800,6 @@ And so, the story of Mulan and Shrek's romance came to an end, leaving a lasting
         penalty_alpha,
     ):
         # global processor, model, tokenizer
-        print("***********Model_inference*************")
-        print(f"Inside Model_inference, user_prompt_str is - {user_prompt_str} ")
-        print(f"Inside Model_inference, chat_history is - {chat_history} ")
-        print(f"Inside Model_inference, image type is - {type(image)} ")
-        print(f"Inside Model_inference, image is - {image} ")
 
         force_words = ""
         hide_special_tokens = False
@@ -851,9 +810,6 @@ And so, the story of Mulan and Shrek's romance came to an end, leaving a lasting
             history=chat_history,
         )
 
-        print(f"formated_prompt_list (or resulting_list)  is {formated_prompt_list}")
-        print(f"user_prompt_list (or [current_user_prompt_str]) is {user_prompt_list}")
-        
         generated_text = model_generation(
             prompt_list=formated_prompt_list,
             processor=processor,
@@ -881,7 +837,6 @@ And so, the story of Mulan and Shrek's romance came to an end, leaving a lasting
             chat_history.append(
                 (user_prompt_list_to_markdown(user_prompt_list), generated_text.strip("<end_of_utterance>"))
             )
-            print(f"chat_history (IF image is None or is with fake token) is -{chat_history}")
         else:
             # Case where the image is passed through the Image Box.
             # Convert the image into base64 for both passing it through the chat history and
@@ -892,7 +847,6 @@ And so, the story of Mulan and Shrek's romance came to an end, leaving a lasting
                     generated_text.strip("<end_of_utterance>"),
                 )
             )
-            print(f"chat_history (ELSE IF image is available) is -{chat_history}")
         return "", None, chat_history
 
 
@@ -1045,52 +999,52 @@ And so, the story of Mulan and Shrek's romance came to an end, leaving a lasting
         examples=[
             ["What are the armed baguettes guarding?", f"{examples_path}/example_images/baguettes_guarding_paris.png"],
             [
-                "Can you tell me a very short story based on this image?",
-                f"{examples_path}/example_images/chicken_on_money.png",
-            ],
-            ["Can you describe the image?", f"{examples_path}/example_images/bear_costume.png"],
-            ["What is this animal and why is it unusual?", f"{examples_path}/example_images/blue_dog.png"],
-            [
-                "What is this object and do you think it is horrifying?",
-                f"{examples_path}/example_images/can_horror.png",
-            ],
-            ["What is this sketch for? How would you make an argument to prove this sketch was made by Picasso himself?", f"{examples_path}/example_images/cat_sketch.png"],
-            ["Which celebrity does this claymation figure look like?", f"{examples_path}/example_images/kanye.jpg"],
-            [
-                "Which famous person does the person in the image look like? Could you craft an engaging narrative featuring this character from the image as the main protagonist?",
-                f"{examples_path}/example_images/obama-harry-potter.jpg",
-            ],
-            [
-                "Is there a celebrity look-alike in this image? What is happening to the person?",
-                f"{examples_path}/example_images/ryan-reynolds-borg.jpg",
-            ],
-            ["Can you describe this image in details please?", f"{examples_path}/example_images/dragons_playing.png"],
-            ["What can you tell me about the cap in this image?", f"{examples_path}/example_images/ironman_cap.png"],
-            [
-                "Can you write an advertisement for Coca-Cola based on this image?",
-                f"{examples_path}/example_images/polar_bear_coke.png",
-            ],
-            [
-                "What is the rabbit doing in this image? Do you think this image is real?",
-                f"{examples_path}/example_images/rabbit_force.png",
-            ],
-            ["What is happening in this image and why is it unusual?", f"{examples_path}/example_images/ramen.png"],
-            [
-                "What I should look most forward to when I visit this place?",
-                f"{examples_path}/example_images/tree_fortress.jpg",
-            ],
-            ["Who is the person in the image and what is he doing?", f"{examples_path}/example_images/tom-cruise-astronaut-pegasus.jpg"],
-            [
-                "What is happening in this image? Which famous personality does this person in center looks like?",
-                f"{examples_path}/example_images/gandhi_selfie.jpg",
-            ],
-            [
-                (
-                    "<fake_token_around_image><image:https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/blog/stable-diffusion-xl-coreml/a_high_quality_photo_of_a_surfing_dog.7667.final_float16_original.jpg><fake_token_around_image>What"
-                    " do you think the dog is doing and is it unusual?"
-                ),
-                None,
-            ],
+                 "Can you tell me a very short story based on this image?",
+                 f"{examples_path}/example_images/chicken_on_money.png",
+             ],
+            # ["Can you describe the image?", f"{examples_path}/example_images/bear_costume.png"],
+            # ["What is this animal and why is it unusual?", f"{examples_path}/example_images/blue_dog.png"],
+            # [
+            #     "What is this object and do you think it is horrifying?",
+            #     f"{examples_path}/example_images/can_horror.png",
+            # ],
+            # ["What is this sketch for? How would you make an argument to prove this sketch was made by Picasso himself?", f"{examples_path}/example_images/cat_sketch.png"],
+            # ["Which celebrity does this claymation figure look like?", f"{examples_path}/example_images/kanye.jpg"],
+            # [
+            #     "Which famous person does the person in the image look like? Could you craft an engaging narrative featuring this character from the image as the main protagonist?",
+            #     f"{examples_path}/example_images/obama-harry-potter.jpg",
+            # ],
+            # [
+            #     "Is there a celebrity look-alike in this image? What is happening to the person?",
+            #     f"{examples_path}/example_images/ryan-reynolds-borg.jpg",
+            # ],
+            # ["Can you describe this image in details please?", f"{examples_path}/example_images/dragons_playing.png"],
+            # ["What can you tell me about the cap in this image?", f"{examples_path}/example_images/ironman_cap.png"],
+            # [
+            #     "Can you write an advertisement for Coca-Cola based on this image?",
+            #     f"{examples_path}/example_images/polar_bear_coke.png",
+            # ],
+            # [
+            #     "What is the rabbit doing in this image? Do you think this image is real?",
+            #     f"{examples_path}/example_images/rabbit_force.png",
+            # ],
+            # ["What is happening in this image and why is it unusual?", f"{examples_path}/example_images/ramen.png"],
+            # [
+            #     "What I should look most forward to when I visit this place?",
+            #     f"{examples_path}/example_images/tree_fortress.jpg",
+            # ],
+            # ["Who is the person in the image and what is he doing?", f"{examples_path}/example_images/tom-cruise-astronaut-pegasus.jpg"],
+            # [
+            #     "What is happening in this image? Which famous personality does this person in center looks like?",
+            #     f"{examples_path}/example_images/gandhi_selfie.jpg",
+            # ],
+            # [
+            #     (
+            #         "<fake_token_around_image><image:https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/blog/stable-diffusion-xl-coreml/a_high_quality_photo_of_a_surfing_dog.7667.final_float16_original.jpg><fake_token_around_image>What"
+            #         " do you think the dog is doing and is it unusual?"
+            #     ),
+            #     None,
+            # ],
         ],
         inputs=[textbox, imagebox],
         outputs=[textbox, imagebox, chatbot],