Spaces:
Sleeping
Sleeping
Update my_model/tabs/run_inference.py
Browse files- my_model/tabs/run_inference.py +63 -49
my_model/tabs/run_inference.py
CHANGED
@@ -5,7 +5,8 @@ import accelerate
|
|
5 |
import scipy
|
6 |
import copy
|
7 |
import time
|
8 |
-
from typing import Tuple, Dict
|
|
|
9 |
from PIL import Image
|
10 |
import torch.nn as nn
|
11 |
import pandas as pd
|
@@ -17,9 +18,9 @@ from my_model.config import inference_config as config
|
|
17 |
|
18 |
|
19 |
class InferenceRunner(StateManager):
|
20 |
-
|
21 |
"""
|
22 |
-
Manages the user interface and interactions for
|
|
|
23 |
This class handles image uploads, displays sample images, and facilitates the question-answering process using the KBVQA model.
|
24 |
Inherits from the StateManager class.
|
25 |
"""
|
@@ -28,10 +29,10 @@ class InferenceRunner(StateManager):
|
|
28 |
"""
|
29 |
Initializes the InferenceRunner instance, setting up the necessary state.
|
30 |
"""
|
31 |
-
|
32 |
-
super().__init__()
|
33 |
|
|
|
34 |
|
|
|
35 |
def answer_question(self, caption: str, detected_objects_str: str, question: str) -> Tuple[str, int]:
|
36 |
"""
|
37 |
Generates an answer to a user's question based on the image's caption and detected objects.
|
@@ -42,8 +43,9 @@ class InferenceRunner(StateManager):
|
|
42 |
question (str): User's question about the image.
|
43 |
|
44 |
Returns:
|
45 |
-
|
46 |
"""
|
|
|
47 |
free_gpu_resources()
|
48 |
answer = st.session_state.kbvqa.generate_answer(question, caption, detected_objects_str)
|
49 |
prompt_length = st.session_state.kbvqa.current_prompt_length
|
@@ -54,6 +56,9 @@ class InferenceRunner(StateManager):
|
|
54 |
def display_sample_images(self) -> None:
|
55 |
"""
|
56 |
Displays sample images as clickable thumbnails for the user to select.
|
|
|
|
|
|
|
57 |
"""
|
58 |
|
59 |
self.col1.write("Choose from sample images:")
|
@@ -66,37 +71,50 @@ class InferenceRunner(StateManager):
|
|
66 |
if st.button(f'Select Sample Image {idx + 1}', key=f'sample_{idx+1}'):
|
67 |
self.process_new_image(sample_image_path, image)
|
68 |
|
|
|
69 |
def handle_image_upload(self) -> None:
|
70 |
"""
|
71 |
Provides an image uploader widget for the user to upload their own images.
|
|
|
|
|
|
|
72 |
"""
|
|
|
73 |
uploaded_image = self.col1.file_uploader("Or upload an Image", type=["png", "jpg", "jpeg"])
|
74 |
if uploaded_image is not None:
|
75 |
self.process_new_image(uploaded_image.name, Image.open(uploaded_image))
|
|
|
76 |
|
77 |
-
def display_image_and_analysis(self, image_key: str, image_data:
|
78 |
"""
|
79 |
Displays the uploaded or selected image and provides an option to analyze the image.
|
80 |
|
81 |
Args:
|
82 |
image_key (str): Unique key identifying the image.
|
83 |
-
image_data (
|
84 |
-
nested_col21 (
|
85 |
-
nested_col22 (
|
|
|
|
|
|
|
86 |
"""
|
87 |
|
88 |
image_for_display = self.resize_image(image_data['image'], 600)
|
89 |
nested_col21.image(image_for_display, caption=f'Uploaded Image: {image_key[-11:]}')
|
90 |
self.handle_analysis_button(image_key, image_data, nested_col22)
|
|
|
91 |
|
92 |
-
def handle_analysis_button(self, image_key: str, image_data:
|
93 |
"""
|
94 |
Provides an 'Analyze Image' button and processes the image analysis upon click.
|
95 |
|
96 |
Args:
|
97 |
image_key (str): Unique key identifying the image.
|
98 |
-
image_data (
|
99 |
-
nested_col22 (
|
|
|
|
|
|
|
100 |
"""
|
101 |
|
102 |
if not image_data['analysis_done'] or self.settings_changed or self.confidance_change:
|
@@ -109,14 +127,18 @@ class InferenceRunner(StateManager):
|
|
109 |
self.update_image_data(image_key, caption, detected_objects_str, True)
|
110 |
st.session_state['loading_in_progress'] = False
|
111 |
|
112 |
-
|
|
|
113 |
"""
|
114 |
Manages the question-answering interface for each image.
|
115 |
|
116 |
Args:
|
117 |
image_key (str): Unique key identifying the image.
|
118 |
-
image_data (
|
119 |
-
nested_col22 (
|
|
|
|
|
|
|
120 |
"""
|
121 |
|
122 |
if image_data['analysis_done']:
|
@@ -126,14 +148,17 @@ class InferenceRunner(StateManager):
|
|
126 |
nested_col22.warning("Confidence level changed, please click 'Analyze Image' each time you change it.")
|
127 |
|
128 |
|
129 |
-
def display_question_answering_interface(self, image_key: str, image_data: Dict, nested_col22:
|
130 |
"""
|
131 |
Displays the interface for question answering, including sample questions and a custom question input.
|
132 |
-
|
133 |
Args:
|
134 |
image_key (str): Unique key identifying the image.
|
135 |
-
image_data (
|
136 |
-
nested_col22 (
|
|
|
|
|
|
|
137 |
"""
|
138 |
|
139 |
sample_questions = config.SAMPLE_QUESTIONS.get(image_key, [])
|
@@ -152,19 +177,20 @@ class InferenceRunner(StateManager):
|
|
152 |
nested_col22.text(f"Q{num+1}: {q}\nA{num+1}: {a}\nPrompt Length: {p}\n")
|
153 |
|
154 |
|
155 |
-
|
156 |
-
def process_question(self, image_key: str, question: str, image_data: Dict, nested_col22: st.columns) -> None:
|
157 |
"""
|
158 |
Processes the user's question, generates an answer, and updates the question-answer history.
|
|
|
|
|
159 |
|
160 |
Args:
|
161 |
image_key (str): Unique key identifying the image.
|
162 |
question (str): The question asked by the user.
|
163 |
image_data (Dict): Data associated with the image.
|
164 |
-
nested_col22 (
|
165 |
-
|
166 |
-
|
167 |
-
|
168 |
"""
|
169 |
|
170 |
qa_history = image_data.get('qa_history', [])
|
@@ -172,7 +198,7 @@ class InferenceRunner(StateManager):
|
|
172 |
if nested_col22.button('Get Answer', key=f'answer_{image_key}', disabled=self.is_widget_disabled):
|
173 |
answer, prompt_length = self.answer_question(image_data['caption'], image_data['detected_objects_str'], question)
|
174 |
self.add_to_qa_history(image_key, question, answer, prompt_length)
|
175 |
-
|
176 |
|
177 |
def image_qa_app(self) -> None:
|
178 |
"""
|
@@ -180,6 +206,9 @@ class InferenceRunner(StateManager):
|
|
180 |
|
181 |
This method orchestrates the display of sample images, handles image uploads, and facilitates the question-answering process.
|
182 |
It iterates through each image in the session state, displaying the image and providing interfaces for image analysis and question answering.
|
|
|
|
|
|
|
183 |
"""
|
184 |
|
185 |
self.display_sample_images()
|
@@ -192,29 +221,27 @@ class InferenceRunner(StateManager):
|
|
192 |
self.display_image_and_analysis(image_key, image_data, nested_col21, nested_col22)
|
193 |
self.handle_question_answering(image_key, image_data, nested_col22)
|
194 |
|
195 |
-
|
196 |
|
197 |
-
def run_inference(self):
|
198 |
"""
|
199 |
-
Sets up widgets and manages the inference process, including model loading and reloading,
|
200 |
-
based on user interactions.
|
201 |
|
202 |
This method orchestrates the overall flow of the inference process.
|
203 |
-
"""
|
204 |
|
205 |
-
|
|
|
|
|
|
|
|
|
206 |
|
207 |
load_fine_tuned_model = False
|
208 |
fine_tuned_model_already_loaded = False
|
209 |
reload_detection_model = False
|
210 |
force_reload_full_model = False
|
211 |
|
212 |
-
|
213 |
if self.is_model_loaded and self.settings_changed:
|
214 |
self.col1.warning("Model settings have changed, please reload the model, this will take a second .. ")
|
215 |
self.update_prev_state()
|
216 |
-
|
217 |
-
|
218 |
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"
|
219 |
|
220 |
with self.col1:
|
@@ -232,25 +259,20 @@ class InferenceRunner(StateManager):
|
|
232 |
reload_detection_model = True
|
233 |
if nested_col12.button("Force Reload", on_click=self.disable_widgets, disabled=self.is_widget_disabled):
|
234 |
force_reload_full_model = True
|
235 |
-
|
236 |
-
|
237 |
if load_fine_tuned_model:
|
238 |
t1=time.time()
|
239 |
free_gpu_resources()
|
240 |
self.load_model()
|
241 |
st.session_state['time_taken_to_load_model'] = int(time.time()-t1)
|
242 |
st.session_state['loading_in_progress'] = False
|
243 |
-
|
244 |
elif fine_tuned_model_already_loaded:
|
245 |
free_gpu_resources()
|
246 |
self.col1.text("Model already loaded and no settings were changed:)")
|
247 |
st.session_state['loading_in_progress'] = False
|
248 |
-
|
249 |
elif reload_detection_model:
|
250 |
free_gpu_resources()
|
251 |
self.reload_detection_model()
|
252 |
st.session_state['loading_in_progress'] = False
|
253 |
-
|
254 |
elif force_reload_full_model:
|
255 |
free_gpu_resources()
|
256 |
t1=time.time()
|
@@ -258,19 +280,11 @@ class InferenceRunner(StateManager):
|
|
258 |
st.session_state['time_taken_to_load_model'] = int(time.time()-t1)
|
259 |
st.session_state['loading_in_progress'] = False
|
260 |
st.session_state['model_loaded'] = True
|
261 |
-
|
262 |
-
# elif st.session_state.method == "13b-Fine-Tuned Model":
|
263 |
-
# self.col1.warning(f'Model using {st.session_state.method} is not deployed yet, will be ready later.')
|
264 |
-
|
265 |
-
|
266 |
elif st.session_state.method == "Vision-Language Embeddings Alignment":
|
267 |
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.')
|
268 |
-
|
269 |
-
|
270 |
if self.is_model_loaded:
|
271 |
free_gpu_resources()
|
272 |
st.session_state['loading_in_progress'] = False
|
273 |
-
|
274 |
self.image_qa_app() # this is the main Q/A Application
|
275 |
|
276 |
|
|
|
5 |
import scipy
|
6 |
import copy
|
7 |
import time
|
8 |
+
from typing import Tuple, Dict, List
|
9 |
+
from streamlit.delta_generator import DeltaGenerator
|
10 |
from PIL import Image
|
11 |
import torch.nn as nn
|
12 |
import pandas as pd
|
|
|
18 |
|
19 |
|
20 |
class InferenceRunner(StateManager):
|
|
|
21 |
"""
|
22 |
+
Manages the user interface and interactions for running inference using the Streamlit-based Knowledge-Based Visual Question Answering (KBVQA) application.
|
23 |
+
|
24 |
This class handles image uploads, displays sample images, and facilitates the question-answering process using the KBVQA model.
|
25 |
Inherits from the StateManager class.
|
26 |
"""
|
|
|
29 |
"""
|
30 |
Initializes the InferenceRunner instance, setting up the necessary state.
|
31 |
"""
|
|
|
|
|
32 |
|
33 |
+
super().__init__()
|
34 |
|
35 |
+
|
36 |
def answer_question(self, caption: str, detected_objects_str: str, question: str) -> Tuple[str, int]:
|
37 |
"""
|
38 |
Generates an answer to a user's question based on the image's caption and detected objects.
|
|
|
43 |
question (str): User's question about the image.
|
44 |
|
45 |
Returns:
|
46 |
+
Tuple[str, int]: A tuple containing the answer to the question and the prompt length.
|
47 |
"""
|
48 |
+
|
49 |
free_gpu_resources()
|
50 |
answer = st.session_state.kbvqa.generate_answer(question, caption, detected_objects_str)
|
51 |
prompt_length = st.session_state.kbvqa.current_prompt_length
|
|
|
56 |
def display_sample_images(self) -> None:
|
57 |
"""
|
58 |
Displays sample images as clickable thumbnails for the user to select.
|
59 |
+
|
60 |
+
Returns:
|
61 |
+
None
|
62 |
"""
|
63 |
|
64 |
self.col1.write("Choose from sample images:")
|
|
|
71 |
if st.button(f'Select Sample Image {idx + 1}', key=f'sample_{idx+1}'):
|
72 |
self.process_new_image(sample_image_path, image)
|
73 |
|
74 |
+
|
75 |
def handle_image_upload(self) -> None:
|
76 |
"""
|
77 |
Provides an image uploader widget for the user to upload their own images.
|
78 |
+
|
79 |
+
Returns:
|
80 |
+
None
|
81 |
"""
|
82 |
+
|
83 |
uploaded_image = self.col1.file_uploader("Or upload an Image", type=["png", "jpg", "jpeg"])
|
84 |
if uploaded_image is not None:
|
85 |
self.process_new_image(uploaded_image.name, Image.open(uploaded_image))
|
86 |
+
|
87 |
|
88 |
+
def display_image_and_analysis(self, image_key: str, image_data: Dict, nested_col21: DeltaGenerator, nested_col22: DeltaGenerator) -> None:
|
89 |
"""
|
90 |
Displays the uploaded or selected image and provides an option to analyze the image.
|
91 |
|
92 |
Args:
|
93 |
image_key (str): Unique key identifying the image.
|
94 |
+
image_data (Dict): Data associated with the image.
|
95 |
+
nested_col21 (DeltaGenerator): Column for displaying the image.
|
96 |
+
nested_col22 (DeltaGenerator): Column for displaying the analysis button.
|
97 |
+
|
98 |
+
Returns:
|
99 |
+
None
|
100 |
"""
|
101 |
|
102 |
image_for_display = self.resize_image(image_data['image'], 600)
|
103 |
nested_col21.image(image_for_display, caption=f'Uploaded Image: {image_key[-11:]}')
|
104 |
self.handle_analysis_button(image_key, image_data, nested_col22)
|
105 |
+
|
106 |
|
107 |
+
def handle_analysis_button(self, image_key: str, image_data: Dict, nested_col22: DeltaGenerator) -> None:
|
108 |
"""
|
109 |
Provides an 'Analyze Image' button and processes the image analysis upon click.
|
110 |
|
111 |
Args:
|
112 |
image_key (str): Unique key identifying the image.
|
113 |
+
image_data (Dict): Data associated with the image.
|
114 |
+
nested_col22 (DeltaGenerator): Column for displaying the analysis button.
|
115 |
+
|
116 |
+
Returns:
|
117 |
+
None
|
118 |
"""
|
119 |
|
120 |
if not image_data['analysis_done'] or self.settings_changed or self.confidance_change:
|
|
|
127 |
self.update_image_data(image_key, caption, detected_objects_str, True)
|
128 |
st.session_state['loading_in_progress'] = False
|
129 |
|
130 |
+
|
131 |
+
def handle_question_answering(self, image_key: str, image_data: Dict, nested_col22: DeltaGenerator) -> None:
|
132 |
"""
|
133 |
Manages the question-answering interface for each image.
|
134 |
|
135 |
Args:
|
136 |
image_key (str): Unique key identifying the image.
|
137 |
+
image_data (Dict): Data associated with the image.
|
138 |
+
nested_col22 (DeltaGenerator): Column for displaying the question-answering interface.
|
139 |
+
|
140 |
+
Returns:
|
141 |
+
None
|
142 |
"""
|
143 |
|
144 |
if image_data['analysis_done']:
|
|
|
148 |
nested_col22.warning("Confidence level changed, please click 'Analyze Image' each time you change it.")
|
149 |
|
150 |
|
151 |
+
def display_question_answering_interface(self, image_key: str, image_data: Dict, nested_col22: DeltaGenerator) -> None:
|
152 |
"""
|
153 |
Displays the interface for question answering, including sample questions and a custom question input.
|
154 |
+
|
155 |
Args:
|
156 |
image_key (str): Unique key identifying the image.
|
157 |
+
image_data (Dict): Data associated with the image.
|
158 |
+
nested_col22 (DeltaGenerator): The column where the interface will be displayed.
|
159 |
+
|
160 |
+
Returns:
|
161 |
+
None
|
162 |
"""
|
163 |
|
164 |
sample_questions = config.SAMPLE_QUESTIONS.get(image_key, [])
|
|
|
177 |
nested_col22.text(f"Q{num+1}: {q}\nA{num+1}: {a}\nPrompt Length: {p}\n")
|
178 |
|
179 |
|
180 |
+
def process_question(self, image_key: str, question: str, image_data: Dict, nested_col22: DeltaGenerator) -> None:
|
|
|
181 |
"""
|
182 |
Processes the user's question, generates an answer, and updates the question-answer history.
|
183 |
+
This method checks if the question is new or if settings have changed, and if so, generates an answer using the KBVQA model.
|
184 |
+
It then updates the question-answer history for the image.
|
185 |
|
186 |
Args:
|
187 |
image_key (str): Unique key identifying the image.
|
188 |
question (str): The question asked by the user.
|
189 |
image_data (Dict): Data associated with the image.
|
190 |
+
nested_col22 (DeltaGenerator): The column where the answer will be displayed.
|
191 |
+
|
192 |
+
Returns:
|
193 |
+
None
|
194 |
"""
|
195 |
|
196 |
qa_history = image_data.get('qa_history', [])
|
|
|
198 |
if nested_col22.button('Get Answer', key=f'answer_{image_key}', disabled=self.is_widget_disabled):
|
199 |
answer, prompt_length = self.answer_question(image_data['caption'], image_data['detected_objects_str'], question)
|
200 |
self.add_to_qa_history(image_key, question, answer, prompt_length)
|
201 |
+
|
202 |
|
203 |
def image_qa_app(self) -> None:
|
204 |
"""
|
|
|
206 |
|
207 |
This method orchestrates the display of sample images, handles image uploads, and facilitates the question-answering process.
|
208 |
It iterates through each image in the session state, displaying the image and providing interfaces for image analysis and question answering.
|
209 |
+
|
210 |
+
Returns:
|
211 |
+
None
|
212 |
"""
|
213 |
|
214 |
self.display_sample_images()
|
|
|
221 |
self.display_image_and_analysis(image_key, image_data, nested_col21, nested_col22)
|
222 |
self.handle_question_answering(image_key, image_data, nested_col22)
|
223 |
|
|
|
224 |
|
225 |
+
def run_inference(self) -> None:
|
226 |
"""
|
227 |
+
Sets up widgets and manages the inference process, including model loading and reloading, based on user interactions.
|
|
|
228 |
|
229 |
This method orchestrates the overall flow of the inference process.
|
|
|
230 |
|
231 |
+
Returns:
|
232 |
+
None
|
233 |
+
"""
|
234 |
+
|
235 |
+
self.set_up_widgets() # Inherent from the StateManager Class
|
236 |
|
237 |
load_fine_tuned_model = False
|
238 |
fine_tuned_model_already_loaded = False
|
239 |
reload_detection_model = False
|
240 |
force_reload_full_model = False
|
241 |
|
|
|
242 |
if self.is_model_loaded and self.settings_changed:
|
243 |
self.col1.warning("Model settings have changed, please reload the model, this will take a second .. ")
|
244 |
self.update_prev_state()
|
|
|
|
|
245 |
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"
|
246 |
|
247 |
with self.col1:
|
|
|
259 |
reload_detection_model = True
|
260 |
if nested_col12.button("Force Reload", on_click=self.disable_widgets, disabled=self.is_widget_disabled):
|
261 |
force_reload_full_model = True
|
|
|
|
|
262 |
if load_fine_tuned_model:
|
263 |
t1=time.time()
|
264 |
free_gpu_resources()
|
265 |
self.load_model()
|
266 |
st.session_state['time_taken_to_load_model'] = int(time.time()-t1)
|
267 |
st.session_state['loading_in_progress'] = False
|
|
|
268 |
elif fine_tuned_model_already_loaded:
|
269 |
free_gpu_resources()
|
270 |
self.col1.text("Model already loaded and no settings were changed:)")
|
271 |
st.session_state['loading_in_progress'] = False
|
|
|
272 |
elif reload_detection_model:
|
273 |
free_gpu_resources()
|
274 |
self.reload_detection_model()
|
275 |
st.session_state['loading_in_progress'] = False
|
|
|
276 |
elif force_reload_full_model:
|
277 |
free_gpu_resources()
|
278 |
t1=time.time()
|
|
|
280 |
st.session_state['time_taken_to_load_model'] = int(time.time()-t1)
|
281 |
st.session_state['loading_in_progress'] = False
|
282 |
st.session_state['model_loaded'] = True
|
|
|
|
|
|
|
|
|
|
|
283 |
elif st.session_state.method == "Vision-Language Embeddings Alignment":
|
284 |
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.')
|
|
|
|
|
285 |
if self.is_model_loaded:
|
286 |
free_gpu_resources()
|
287 |
st.session_state['loading_in_progress'] = False
|
|
|
288 |
self.image_qa_app() # this is the main Q/A Application
|
289 |
|
290 |
|