Update my_model/tabs/run_inference.py

my_model/tabs/run_inference.py

@@ -5,7 +5,8 @@ import accelerate
 import scipy
 import copy
 import time
-from typing import Tuple, Dict
+from typing import Tuple, Dict, List
+from streamlit.delta_generator import DeltaGenerator
 from PIL import Image
 import torch.nn as nn
 import pandas as pd
@@ -17,9 +18,9 @@ from my_model.config import inference_config as config
 
 
 class InferenceRunner(StateManager):
-
     """
-    Manages the user interface and interactions for a Streamlit-based Knowledge-Based Visual Question Answering (KBVQA) application.
+    Manages the user interface and interactions for running inference using the Streamlit-based Knowledge-Based Visual Question Answering (KBVQA) application.
+    
     This class handles image uploads, displays sample images, and facilitates the question-answering process using the KBVQA model.
     Inherits from the StateManager class.
     """
@@ -28,10 +29,10 @@ class InferenceRunner(StateManager):
         """
         Initializes the InferenceRunner instance, setting up the necessary state.
         """
-        
-        super().__init__()
 
+        super().__init__()
 
+    
     def answer_question(self, caption: str, detected_objects_str: str, question: str) -> Tuple[str, int]:
         """
         Generates an answer to a user's question based on the image's caption and detected objects.
@@ -42,8 +43,9 @@ class InferenceRunner(StateManager):
             question (str): User's question about the image.
 
         Returns:
-            tuple: A tuple containing the answer to the question and the prompt length.
+            Tuple[str, int]: A tuple containing the answer to the question and the prompt length.
         """
+        
         free_gpu_resources()
         answer = st.session_state.kbvqa.generate_answer(question, caption, detected_objects_str)
         prompt_length  = st.session_state.kbvqa.current_prompt_length
@@ -54,6 +56,9 @@ class InferenceRunner(StateManager):
     def display_sample_images(self) -> None:
         """
         Displays sample images as clickable thumbnails for the user to select.
+
+        Returns:
+            None
         """
         
         self.col1.write("Choose from sample images:")
@@ -66,37 +71,50 @@ class InferenceRunner(StateManager):
                 if st.button(f'Select Sample Image {idx + 1}', key=f'sample_{idx+1}'):
                     self.process_new_image(sample_image_path, image)
 
+    
     def handle_image_upload(self) -> None:
         """
         Provides an image uploader widget for the user to upload their own images.
+
+        Returns:
+            None
         """
+        
         uploaded_image = self.col1.file_uploader("Or upload an Image", type=["png", "jpg", "jpeg"])
         if uploaded_image is not None:
             self.process_new_image(uploaded_image.name, Image.open(uploaded_image))
+            
 
-    def display_image_and_analysis(self, image_key: str, image_data: dict, nested_col21, nested_col22) -> None:
+    def display_image_and_analysis(self, image_key: str, image_data: Dict, nested_col21: DeltaGenerator, nested_col22: DeltaGenerator) -> None:
         """
         Displays the uploaded or selected image and provides an option to analyze the image.
 
         Args:
             image_key (str): Unique key identifying the image.
-            image_data (dict): Data associated with the image.
-            nested_col21 (streamlit column): Column for displaying the image.
-            nested_col22 (streamlit column): Column for displaying the analysis button.
+            image_data (Dict): Data associated with the image.
+            nested_col21 (DeltaGenerator): Column for displaying the image.
+            nested_col22 (DeltaGenerator): Column for displaying the analysis button.
+        
+        Returns:
+            None
         """
         
         image_for_display = self.resize_image(image_data['image'], 600)
         nested_col21.image(image_for_display, caption=f'Uploaded Image: {image_key[-11:]}')
         self.handle_analysis_button(image_key, image_data, nested_col22)
+        
 
-    def handle_analysis_button(self, image_key: str, image_data: dict, nested_col22) -> None:
+    def handle_analysis_button(self, image_key: str, image_data: Dict, nested_col22: DeltaGenerator) -> None:
         """
         Provides an 'Analyze Image' button and processes the image analysis upon click.
 
         Args:
             image_key (str): Unique key identifying the image.
-            image_data (dict): Data associated with the image.
-            nested_col22 (streamlit column): Column for displaying the analysis button.
+            image_data (Dict): Data associated with the image.
+            nested_col22 (DeltaGenerator): Column for displaying the analysis button.
+        
+        Returns:
+            None
         """
         
         if not image_data['analysis_done'] or self.settings_changed or self.confidance_change:
@@ -109,14 +127,18 @@ class InferenceRunner(StateManager):
                         self.update_image_data(image_key, caption, detected_objects_str, True)
             st.session_state['loading_in_progress'] = False
 
-    def handle_question_answering(self, image_key: str, image_data: dict, nested_col22) -> None:
+    
+    def handle_question_answering(self, image_key: str, image_data: Dict, nested_col22: DeltaGenerator) -> None:
         """
         Manages the question-answering interface for each image.
 
         Args:
             image_key (str): Unique key identifying the image.
-            image_data (dict): Data associated with the image.
-            nested_col22 (streamlit column): Column for displaying the question-answering interface.
+            image_data (Dict): Data associated with the image.
+            nested_col22 (DeltaGenerator): Column for displaying the question-answering interface.
+        
+        Returns:
+            None
         """
         
         if image_data['analysis_done']:
@@ -126,14 +148,17 @@ class InferenceRunner(StateManager):
             nested_col22.warning("Confidence level changed, please click 'Analyze Image' each time you change it.")
 
 
-    def display_question_answering_interface(self, image_key: str, image_data: Dict, nested_col22: st.columns) -> None:
+    def display_question_answering_interface(self, image_key: str, image_data: Dict, nested_col22: DeltaGenerator) -> None:
         """
         Displays the interface for question answering, including sample questions and a custom question input.
-    
+
         Args:
             image_key (str): Unique key identifying the image.
-            image_data (dict): Data associated with the image.
-            nested_col22 (streamlit column): The column where the interface will be displayed.
+            image_data (Dict): Data associated with the image.
+            nested_col22 (DeltaGenerator): The column where the interface will be displayed.
+        
+        Returns:
+            None
         """
         
         sample_questions = config.SAMPLE_QUESTIONS.get(image_key, [])
@@ -152,19 +177,20 @@ class InferenceRunner(StateManager):
             nested_col22.text(f"Q{num+1}: {q}\nA{num+1}: {a}\nPrompt Length: {p}\n")
 
 
-            
-    def process_question(self, image_key: str, question: str, image_data: Dict, nested_col22: st.columns) -> None:
+    def process_question(self, image_key: str, question: str, image_data: Dict, nested_col22: DeltaGenerator) -> None:
         """
         Processes the user's question, generates an answer, and updates the question-answer history.
+        This method checks if the question is new or if settings have changed, and if so, generates an answer using the KBVQA model.
+        It then updates the question-answer history for the image.
 
         Args:
             image_key (str): Unique key identifying the image.
             question (str): The question asked by the user.
             image_data (Dict): Data associated with the image.
-            nested_col22 (streamlit column): The column where the answer will be displayed.
-
-        This method checks if the question is new or if settings have changed, and if so, generates an answer using the KBVQA model.
-        It then updates the question-answer history for the image.
+            nested_col22 (DeltaGenerator): The column where the answer will be displayed.
+        
+        Returns:
+            None
         """
         
         qa_history = image_data.get('qa_history', [])
@@ -172,7 +198,7 @@ class InferenceRunner(StateManager):
             if nested_col22.button('Get Answer', key=f'answer_{image_key}', disabled=self.is_widget_disabled):
                 answer, prompt_length = self.answer_question(image_data['caption'], image_data['detected_objects_str'], question)
                 self.add_to_qa_history(image_key, question, answer, prompt_length)
-               # nested_col22.text(f"Q: {question}\nA: {answer}\nPrompt Length: {prompt_length}")
+                
 
     def image_qa_app(self) -> None:
         """
@@ -180,6 +206,9 @@ class InferenceRunner(StateManager):
 
         This method orchestrates the display of sample images, handles image uploads, and facilitates the question-answering process.
         It iterates through each image in the session state, displaying the image and providing interfaces for image analysis and question answering.
+        
+        Returns:
+            None
         """
         
         self.display_sample_images()
@@ -192,29 +221,27 @@ class InferenceRunner(StateManager):
                     self.display_image_and_analysis(image_key, image_data, nested_col21, nested_col22)
                     self.handle_question_answering(image_key, image_data, nested_col22)
                     
-
         
-    def run_inference(self):
+    def run_inference(self) -> None:
         """
-        Sets up widgets and manages the inference process, including model loading and reloading,
-        based on user interactions.
+        Sets up widgets and manages the inference process, including model loading and reloading, based on user interactions.
 
         This method orchestrates the overall flow of the inference process.
-        """
         
-        self.set_up_widgets()
+        Returns:
+            None
+        """
+
+        self.set_up_widgets()  # Inherent from the StateManager Class
   
         load_fine_tuned_model = False
         fine_tuned_model_already_loaded = False
         reload_detection_model = False
         force_reload_full_model = False
         
-     
         if self.is_model_loaded and self.settings_changed:
             self.col1.warning("Model settings have changed, please reload the model, this will take a second .. ")
             self.update_prev_state()
-         
-           
         st.session_state.button_label = "Reload Model" if self.is_model_loaded and st.session_state.kbvqa.detection_model != st.session_state['detection_model'] else "Load Model"
       
         with self.col1:
@@ -232,25 +259,20 @@ class InferenceRunner(StateManager):
                             reload_detection_model = True
                     if nested_col12.button("Force Reload", on_click=self.disable_widgets, disabled=self.is_widget_disabled):
                         force_reload_full_model = True
-                        
-
                 if load_fine_tuned_model:
                     t1=time.time()
                     free_gpu_resources()
                     self.load_model()
                     st.session_state['time_taken_to_load_model'] = int(time.time()-t1)
                     st.session_state['loading_in_progress'] = False
-                    
                 elif fine_tuned_model_already_loaded:
                     free_gpu_resources()
                     self.col1.text("Model already loaded and no settings were changed:)")
                     st.session_state['loading_in_progress'] = False
-                    
                 elif reload_detection_model:
                     free_gpu_resources()
                     self.reload_detection_model()
                     st.session_state['loading_in_progress'] = False
-                    
                 elif force_reload_full_model:
                     free_gpu_resources()
                     t1=time.time()
@@ -258,19 +280,11 @@ class InferenceRunner(StateManager):
                     st.session_state['time_taken_to_load_model'] = int(time.time()-t1)
                     st.session_state['loading_in_progress'] = False
                     st.session_state['model_loaded'] = True
-                    
-           # elif st.session_state.method == "13b-Fine-Tuned Model":
-           #     self.col1.warning(f'Model using {st.session_state.method} is not deployed yet, will be ready later.')
-                
-
             elif st.session_state.method == "Vision-Language Embeddings Alignment":
                 self.col1.warning(f'Model using {st.session_state.method} is desgined but requires large scale data and multiple high-end GPUs, implementation will be explored in the future.')
-            
-        
         if self.is_model_loaded:
             free_gpu_resources()
             st.session_state['loading_in_progress'] = False
-            
             self.image_qa_app() # this is the main Q/A Application