Spaces:
Running
on
CPU Upgrade
Running
on
CPU Upgrade
chore: update
Browse files
app.py
CHANGED
@@ -27,38 +27,39 @@ time.sleep(3)
|
|
27 |
# pylint: disable=c-extension-no-member,invalid-name
|
28 |
|
29 |
|
30 |
-
def
|
31 |
"""
|
32 |
-
Check if the
|
33 |
|
34 |
Args:
|
35 |
-
|
36 |
|
37 |
Returns:
|
38 |
-
bool: True if the
|
39 |
"""
|
40 |
-
return
|
41 |
|
42 |
|
|
|
|
|
43 |
# def fill_in_fn(default_disease: str, *checkbox_symptoms: Tuple[str]) -> Dict:
|
44 |
# """
|
45 |
-
# Fill in the gr.CheckBoxGroup list with
|
46 |
-
|
47 |
# Args:
|
48 |
-
# default_disease (str): The default disease
|
49 |
-
# *checkbox_symptoms (Tuple[str]):
|
50 |
-
|
51 |
# Returns:
|
52 |
# dict: The updated gr.CheckBoxesGroup.
|
53 |
# """
|
|
|
|
|
54 |
# df = pd.read_csv(TRAINING_FILENAME)
|
55 |
# df_filtred = df[df[TARGET_COLUMNS[1]] == default_disease]
|
56 |
# symptoms = pretty_print(df_filtred.columns[df_filtred.eq(1).any()].to_list())
|
57 |
-
|
58 |
# if any(lst for lst in checkbox_symptoms if lst):
|
59 |
# for sublist in checkbox_symptoms:
|
60 |
# symptoms.extend(sublist)
|
61 |
-
|
62 |
# return {box: symptoms for box in check_boxes}
|
63 |
|
64 |
|
@@ -67,7 +68,7 @@ def get_user_symptoms_from_checkboxgroup(checkbox_symptoms: List) -> np.array:
|
|
67 |
Convert the user symptoms into a binary vector representation.
|
68 |
|
69 |
Args:
|
70 |
-
checkbox_symptoms (
|
71 |
|
72 |
Returns:
|
73 |
np.array: A binary vector representing the user's symptoms.
|
@@ -76,7 +77,7 @@ def get_user_symptoms_from_checkboxgroup(checkbox_symptoms: List) -> np.array:
|
|
76 |
KeyError: If a provided symptom is not recognized as a valid symptom.
|
77 |
|
78 |
"""
|
79 |
-
symptoms_vector = {key: 0 for key in
|
80 |
for pretty_symptom in checkbox_symptoms:
|
81 |
original_symptom = "_".join((pretty_symptom.lower().split(" ")))
|
82 |
if original_symptom not in symptoms_vector.keys():
|
@@ -130,10 +131,10 @@ def key_gen_fn(user_symptoms: List[str]) -> Dict:
|
|
130 |
"""
|
131 |
clean_directory()
|
132 |
|
133 |
-
if
|
134 |
print("Error: Please submit your symptoms or select a default disease.")
|
135 |
return {
|
136 |
-
error_box2: gr.update(visible=True, value="Please submit your symptoms first"),
|
137 |
}
|
138 |
|
139 |
# Generate a random user ID
|
@@ -142,7 +143,6 @@ def key_gen_fn(user_symptoms: List[str]) -> Dict:
|
|
142 |
|
143 |
client = FHEModelClient(path_dir=DEPLOYMENT_DIR, key_dir=KEYS_DIR / f"{user_id}")
|
144 |
client.load()
|
145 |
-
print("CLIENT_LOADED")
|
146 |
|
147 |
# Creates the private and evaluation keys on the client side
|
148 |
client.generate_private_and_evaluation_keys()
|
@@ -175,7 +175,7 @@ def encrypt_fn(user_symptoms: np.ndarray, user_id: str) -> None:
|
|
175 |
user_id (user): The current user's ID
|
176 |
"""
|
177 |
|
178 |
-
if
|
179 |
print("Error in encryption step: Provide your symptoms and generate the evaluation keys.")
|
180 |
return {
|
181 |
error_box3: gr.update(
|
@@ -192,7 +192,7 @@ def encrypt_fn(user_symptoms: np.ndarray, user_id: str) -> None:
|
|
192 |
|
193 |
encrypted_quantized_user_symptoms = client.quantize_encrypt_serialize(user_symptoms)
|
194 |
assert isinstance(encrypted_quantized_user_symptoms, bytes)
|
195 |
-
encrypted_input_path = KEYS_DIR / f"{user_id}/
|
196 |
|
197 |
with encrypted_input_path.open("wb") as f:
|
198 |
f.write(encrypted_quantized_user_symptoms)
|
@@ -213,11 +213,11 @@ def send_input_fn(user_id: str, user_symptoms: np.ndarray) -> Dict:
|
|
213 |
"""Send the encrypted data and the evaluation key to the server.
|
214 |
|
215 |
Args:
|
216 |
-
user_id (
|
217 |
-
user_symptoms (
|
218 |
"""
|
219 |
|
220 |
-
if
|
221 |
return {
|
222 |
error_box4: gr.update(
|
223 |
visible=True,
|
@@ -227,7 +227,7 @@ def send_input_fn(user_id: str, user_symptoms: np.ndarray) -> Dict:
|
|
227 |
}
|
228 |
|
229 |
evaluation_key_path = KEYS_DIR / f"{user_id}/evaluation_key"
|
230 |
-
encrypted_input_path = KEYS_DIR / f"{user_id}/
|
231 |
|
232 |
if not evaluation_key_path.is_file():
|
233 |
print(
|
@@ -252,7 +252,7 @@ def send_input_fn(user_id: str, user_symptoms: np.ndarray) -> Dict:
|
|
252 |
# Define the data and files to post
|
253 |
data = {
|
254 |
"user_id": user_id,
|
255 |
-
"
|
256 |
}
|
257 |
|
258 |
files = [
|
@@ -268,18 +268,20 @@ def send_input_fn(user_id: str, user_symptoms: np.ndarray) -> Dict:
|
|
268 |
files=files,
|
269 |
) as response:
|
270 |
print(f"Sending Data: {response.ok=}")
|
271 |
-
return {
|
272 |
-
|
273 |
-
|
|
|
|
|
274 |
|
275 |
|
276 |
def run_fhe_fn(user_id: str) -> Dict:
|
277 |
-
"""Send the encrypted input
|
278 |
|
279 |
Args:
|
280 |
user_id (int): The current user's ID.
|
281 |
"""
|
282 |
-
if
|
283 |
return {
|
284 |
error_box5: gr.update(
|
285 |
visible=True,
|
@@ -292,8 +294,6 @@ def run_fhe_fn(user_id: str) -> Dict:
|
|
292 |
"user_id": user_id,
|
293 |
}
|
294 |
|
295 |
-
# Trigger the FHE execution on the encrypted previously sent
|
296 |
-
|
297 |
url = SERVER_URL + "run_fhe"
|
298 |
|
299 |
with requests.post(
|
@@ -302,7 +302,13 @@ def run_fhe_fn(user_id: str) -> Dict:
|
|
302 |
) as response:
|
303 |
if not response.ok:
|
304 |
return {
|
305 |
-
error_box5: gr.update(
|
|
|
|
|
|
|
|
|
|
|
|
|
306 |
fhe_execution_time_box: gr.update(visible=True),
|
307 |
}
|
308 |
else:
|
@@ -311,7 +317,7 @@ def run_fhe_fn(user_id: str) -> Dict:
|
|
311 |
return {
|
312 |
error_box5: gr.update(visible=False),
|
313 |
fhe_execution_time_box: gr.update(value=f"{response.json()} seconds"),
|
314 |
-
next_step_tab4: gr.update(visible=True)
|
315 |
}
|
316 |
|
317 |
|
@@ -319,11 +325,11 @@ def get_output_fn(user_id: str, user_symptoms: np.ndarray) -> Dict:
|
|
319 |
"""Retreive the encrypted data from the server.
|
320 |
|
321 |
Args:
|
322 |
-
user_id (
|
323 |
-
user_symptoms (
|
324 |
"""
|
325 |
|
326 |
-
if
|
327 |
return {
|
328 |
error_box6: gr.update(
|
329 |
visible=True,
|
@@ -360,14 +366,14 @@ def decrypt_fn(user_id: str, user_symptoms: np.ndarray) -> Dict:
|
|
360 |
"""Dencrypt the data on the `Client Side`.
|
361 |
|
362 |
Args:
|
363 |
-
user_id (
|
364 |
-
user_symptoms (
|
365 |
|
366 |
Returns:
|
367 |
Decrypted output
|
368 |
"""
|
369 |
|
370 |
-
if
|
371 |
return {
|
372 |
error_box7: gr.update(
|
373 |
visible=True,
|
@@ -377,7 +383,6 @@ def decrypt_fn(user_id: str, user_symptoms: np.ndarray) -> Dict:
|
|
377 |
}
|
378 |
|
379 |
# Get the encrypted output path
|
380 |
-
|
381 |
encrypted_output_path = CLIENT_DIR / f"{user_id}_encrypted_output"
|
382 |
|
383 |
if not encrypted_output_path.is_file():
|
@@ -398,6 +403,7 @@ def decrypt_fn(user_id: str, user_symptoms: np.ndarray) -> Dict:
|
|
398 |
# Retrieve the client API
|
399 |
client = FHEModelClient(path_dir=DEPLOYMENT_DIR, key_dir=KEYS_DIR / f"{user_id}")
|
400 |
client.load()
|
|
|
401 |
# Deserialize, decrypt and post-process the encrypted output
|
402 |
output = client.deserialize_decrypt_dequantize(encrypted_output)
|
403 |
|
@@ -408,7 +414,7 @@ def decrypt_fn(user_id: str, user_symptoms: np.ndarray) -> Dict:
|
|
408 |
|
409 |
|
410 |
def reset_fn():
|
411 |
-
"""
|
412 |
|
413 |
clean_directory()
|
414 |
|
@@ -455,9 +461,7 @@ if __name__ == "__main__":
|
|
455 |
|
456 |
clean_directory()
|
457 |
|
458 |
-
(X_train, X_test), (y_train, y_test) = load_data()
|
459 |
-
|
460 |
-
valid_columns = X_train.columns.to_list()
|
461 |
|
462 |
with gr.Blocks(css=CSS) as demo:
|
463 |
|
@@ -512,6 +516,8 @@ if __name__ == "__main__":
|
|
512 |
|
513 |
error_box1 = gr.Textbox(label="Error", visible=False)
|
514 |
|
|
|
|
|
515 |
# Default disease, picked from the dataframe
|
516 |
# disease_box = gr.Dropdown(list(sorted(set(df_test["prognosis"]))),
|
517 |
# label="Disease:")
|
@@ -535,7 +541,8 @@ if __name__ == "__main__":
|
|
535 |
<p align="center">
|
536 |
<img width="80%" height="20%" src="https://raw.githubusercontent.com/kcelia/Img/main/Go-To-Step2.png">
|
537 |
</p>
|
538 |
-
""",
|
|
|
539 |
)
|
540 |
|
541 |
submit_button.click(
|
@@ -628,7 +635,8 @@ if __name__ == "__main__":
|
|
628 |
<p align="center">
|
629 |
<img width="80%" height="20%" src="https://raw.githubusercontent.com/kcelia/Img/main/Go-To-Step3.png">
|
630 |
</p>
|
631 |
-
""",
|
|
|
632 |
)
|
633 |
|
634 |
send_input_btn.click(
|
@@ -652,7 +660,8 @@ if __name__ == "__main__":
|
|
652 |
<p align="center">
|
653 |
<img width="80%" height="20%" src="https://raw.githubusercontent.com/kcelia/Img/main/Go-to-Step4.png">
|
654 |
</p>
|
655 |
-
""",
|
|
|
656 |
)
|
657 |
run_fhe_btn.click(
|
658 |
run_fhe_fn,
|
|
|
27 |
# pylint: disable=c-extension-no-member,invalid-name
|
28 |
|
29 |
|
30 |
+
def is_none(obj) -> bool:
|
31 |
"""
|
32 |
+
Check if the object is None.
|
33 |
|
34 |
Args:
|
35 |
+
obj (any): The input to be checked.
|
36 |
|
37 |
Returns:
|
38 |
+
bool: True if the object is None or empty, False otherwise.
|
39 |
"""
|
40 |
+
return all((obj is None, (obj is not None and len(obj) < 1)))
|
41 |
|
42 |
|
43 |
+
# <!> This function has been paused due to UI issues.
|
44 |
+
|
45 |
# def fill_in_fn(default_disease: str, *checkbox_symptoms: Tuple[str]) -> Dict:
|
46 |
# """
|
47 |
+
# Fill in the gr.CheckBoxGroup list with predefined symptoms of a selected default disease.
|
|
|
48 |
# Args:
|
49 |
+
# default_disease (str): The default selected disease
|
50 |
+
# *checkbox_symptoms (Tuple[str]): Existing checked symptoms
|
|
|
51 |
# Returns:
|
52 |
# dict: The updated gr.CheckBoxesGroup.
|
53 |
# """
|
54 |
+
#
|
55 |
+
# # Figure out the symptoms of the disease, selected by the user
|
56 |
# df = pd.read_csv(TRAINING_FILENAME)
|
57 |
# df_filtred = df[df[TARGET_COLUMNS[1]] == default_disease]
|
58 |
# symptoms = pretty_print(df_filtred.columns[df_filtred.eq(1).any()].to_list())
|
59 |
+
# # Check if there are existing symptoms, in the CheckbBxGroup list
|
60 |
# if any(lst for lst in checkbox_symptoms if lst):
|
61 |
# for sublist in checkbox_symptoms:
|
62 |
# symptoms.extend(sublist)
|
|
|
63 |
# return {box: symptoms for box in check_boxes}
|
64 |
|
65 |
|
|
|
68 |
Convert the user symptoms into a binary vector representation.
|
69 |
|
70 |
Args:
|
71 |
+
checkbox_symptoms (List): A list of user symptoms.
|
72 |
|
73 |
Returns:
|
74 |
np.array: A binary vector representing the user's symptoms.
|
|
|
77 |
KeyError: If a provided symptom is not recognized as a valid symptom.
|
78 |
|
79 |
"""
|
80 |
+
symptoms_vector = {key: 0 for key in valid_symptoms}
|
81 |
for pretty_symptom in checkbox_symptoms:
|
82 |
original_symptom = "_".join((pretty_symptom.lower().split(" ")))
|
83 |
if original_symptom not in symptoms_vector.keys():
|
|
|
131 |
"""
|
132 |
clean_directory()
|
133 |
|
134 |
+
if is_none(user_symptoms):
|
135 |
print("Error: Please submit your symptoms or select a default disease.")
|
136 |
return {
|
137 |
+
error_box2: gr.update(visible=True, value="Please submit your symptoms first!."),
|
138 |
}
|
139 |
|
140 |
# Generate a random user ID
|
|
|
143 |
|
144 |
client = FHEModelClient(path_dir=DEPLOYMENT_DIR, key_dir=KEYS_DIR / f"{user_id}")
|
145 |
client.load()
|
|
|
146 |
|
147 |
# Creates the private and evaluation keys on the client side
|
148 |
client.generate_private_and_evaluation_keys()
|
|
|
175 |
user_id (user): The current user's ID
|
176 |
"""
|
177 |
|
178 |
+
if is_none(user_id) or is_none(user_symptoms):
|
179 |
print("Error in encryption step: Provide your symptoms and generate the evaluation keys.")
|
180 |
return {
|
181 |
error_box3: gr.update(
|
|
|
192 |
|
193 |
encrypted_quantized_user_symptoms = client.quantize_encrypt_serialize(user_symptoms)
|
194 |
assert isinstance(encrypted_quantized_user_symptoms, bytes)
|
195 |
+
encrypted_input_path = KEYS_DIR / f"{user_id}/encrypted_input"
|
196 |
|
197 |
with encrypted_input_path.open("wb") as f:
|
198 |
f.write(encrypted_quantized_user_symptoms)
|
|
|
213 |
"""Send the encrypted data and the evaluation key to the server.
|
214 |
|
215 |
Args:
|
216 |
+
user_id (str): The current user's ID
|
217 |
+
user_symptoms (np.ndarray): The user symptoms
|
218 |
"""
|
219 |
|
220 |
+
if is_none(user_id) or is_none(user_symptoms):
|
221 |
return {
|
222 |
error_box4: gr.update(
|
223 |
visible=True,
|
|
|
227 |
}
|
228 |
|
229 |
evaluation_key_path = KEYS_DIR / f"{user_id}/evaluation_key"
|
230 |
+
encrypted_input_path = KEYS_DIR / f"{user_id}/encrypted_input"
|
231 |
|
232 |
if not evaluation_key_path.is_file():
|
233 |
print(
|
|
|
252 |
# Define the data and files to post
|
253 |
data = {
|
254 |
"user_id": user_id,
|
255 |
+
"input": user_symptoms,
|
256 |
}
|
257 |
|
258 |
files = [
|
|
|
268 |
files=files,
|
269 |
) as response:
|
270 |
print(f"Sending Data: {response.ok=}")
|
271 |
+
return {
|
272 |
+
error_box4: gr.update(visible=False),
|
273 |
+
next_step_tab3: gr.update(visible=True),
|
274 |
+
srv_resp_send_data_box: "Data sent",
|
275 |
+
}
|
276 |
|
277 |
|
278 |
def run_fhe_fn(user_id: str) -> Dict:
|
279 |
+
"""Send the encrypted input and the evaluation key to the server.
|
280 |
|
281 |
Args:
|
282 |
user_id (int): The current user's ID.
|
283 |
"""
|
284 |
+
if is_none(user_id):
|
285 |
return {
|
286 |
error_box5: gr.update(
|
287 |
visible=True,
|
|
|
294 |
"user_id": user_id,
|
295 |
}
|
296 |
|
|
|
|
|
297 |
url = SERVER_URL + "run_fhe"
|
298 |
|
299 |
with requests.post(
|
|
|
302 |
) as response:
|
303 |
if not response.ok:
|
304 |
return {
|
305 |
+
error_box5: gr.update(
|
306 |
+
visible=True,
|
307 |
+
value=(
|
308 |
+
"An error occurred on the Server Side. "
|
309 |
+
"Please check connectivity and data transmission."
|
310 |
+
),
|
311 |
+
),
|
312 |
fhe_execution_time_box: gr.update(visible=True),
|
313 |
}
|
314 |
else:
|
|
|
317 |
return {
|
318 |
error_box5: gr.update(visible=False),
|
319 |
fhe_execution_time_box: gr.update(value=f"{response.json()} seconds"),
|
320 |
+
next_step_tab4: gr.update(visible=True),
|
321 |
}
|
322 |
|
323 |
|
|
|
325 |
"""Retreive the encrypted data from the server.
|
326 |
|
327 |
Args:
|
328 |
+
user_id (str): The current user's ID
|
329 |
+
user_symptoms (np.ndarray): The user symptoms
|
330 |
"""
|
331 |
|
332 |
+
if is_none(user_id) or is_none(user_symptoms):
|
333 |
return {
|
334 |
error_box6: gr.update(
|
335 |
visible=True,
|
|
|
366 |
"""Dencrypt the data on the `Client Side`.
|
367 |
|
368 |
Args:
|
369 |
+
user_id (str): The current user's ID
|
370 |
+
user_symptoms (np.ndarray): The user symptoms
|
371 |
|
372 |
Returns:
|
373 |
Decrypted output
|
374 |
"""
|
375 |
|
376 |
+
if is_none(user_id) or is_none(user_symptoms):
|
377 |
return {
|
378 |
error_box7: gr.update(
|
379 |
visible=True,
|
|
|
383 |
}
|
384 |
|
385 |
# Get the encrypted output path
|
|
|
386 |
encrypted_output_path = CLIENT_DIR / f"{user_id}_encrypted_output"
|
387 |
|
388 |
if not encrypted_output_path.is_file():
|
|
|
403 |
# Retrieve the client API
|
404 |
client = FHEModelClient(path_dir=DEPLOYMENT_DIR, key_dir=KEYS_DIR / f"{user_id}")
|
405 |
client.load()
|
406 |
+
|
407 |
# Deserialize, decrypt and post-process the encrypted output
|
408 |
output = client.deserialize_decrypt_dequantize(encrypted_output)
|
409 |
|
|
|
414 |
|
415 |
|
416 |
def reset_fn():
|
417 |
+
"""Reset the space and clear all the box outputs."""
|
418 |
|
419 |
clean_directory()
|
420 |
|
|
|
461 |
|
462 |
clean_directory()
|
463 |
|
464 |
+
(X_train, X_test), (y_train, y_test), valid_symptoms = load_data()
|
|
|
|
|
465 |
|
466 |
with gr.Blocks(css=CSS) as demo:
|
467 |
|
|
|
516 |
|
517 |
error_box1 = gr.Textbox(label="Error", visible=False)
|
518 |
|
519 |
+
# <!> This part has been paused due to UI issues.
|
520 |
+
|
521 |
# Default disease, picked from the dataframe
|
522 |
# disease_box = gr.Dropdown(list(sorted(set(df_test["prognosis"]))),
|
523 |
# label="Disease:")
|
|
|
541 |
<p align="center">
|
542 |
<img width="80%" height="20%" src="https://raw.githubusercontent.com/kcelia/Img/main/Go-To-Step2.png">
|
543 |
</p>
|
544 |
+
""",
|
545 |
+
visible=False,
|
546 |
)
|
547 |
|
548 |
submit_button.click(
|
|
|
635 |
<p align="center">
|
636 |
<img width="80%" height="20%" src="https://raw.githubusercontent.com/kcelia/Img/main/Go-To-Step3.png">
|
637 |
</p>
|
638 |
+
""",
|
639 |
+
visible=False,
|
640 |
)
|
641 |
|
642 |
send_input_btn.click(
|
|
|
660 |
<p align="center">
|
661 |
<img width="80%" height="20%" src="https://raw.githubusercontent.com/kcelia/Img/main/Go-to-Step4.png">
|
662 |
</p>
|
663 |
+
""",
|
664 |
+
visible=False,
|
665 |
)
|
666 |
run_fhe_btn.click(
|
667 |
run_fhe_fn,
|
utils.py
CHANGED
@@ -1,7 +1,7 @@
|
|
1 |
import os
|
2 |
import shutil
|
3 |
from pathlib import Path
|
4 |
-
from typing import
|
5 |
|
6 |
import numpy
|
7 |
import pandas
|
@@ -87,7 +87,7 @@ def get_disease_name(encoded_prediction: int, file_name: str = TRAINING_FILENAME
|
|
87 |
return disease_name
|
88 |
|
89 |
|
90 |
-
def load_data() -> Tuple[pandas.DataFrame,
|
91 |
"""
|
92 |
Return the data
|
93 |
|
@@ -95,7 +95,7 @@ def load_data() -> Tuple[pandas.DataFrame, pandas.DataFrame, numpy.ndarray]:
|
|
95 |
None
|
96 |
|
97 |
Return:
|
98 |
-
|
99 |
|
100 |
|
101 |
"""
|
@@ -113,7 +113,7 @@ def load_data() -> Tuple[pandas.DataFrame, pandas.DataFrame, numpy.ndarray]:
|
|
113 |
y_test = df_test[TARGET_COLUMNS[0]]
|
114 |
X_test = df_test.drop(columns=TARGET_COLUMNS, axis=1, errors="ignore")
|
115 |
|
116 |
-
return (X_train, X_test), (y_train, y_test)
|
117 |
|
118 |
|
119 |
def load_model(X_train: pandas.DataFrame, y_train: numpy.ndarray):
|
|
|
1 |
import os
|
2 |
import shutil
|
3 |
from pathlib import Path
|
4 |
+
from typing import List, Tuple, Union
|
5 |
|
6 |
import numpy
|
7 |
import pandas
|
|
|
87 |
return disease_name
|
88 |
|
89 |
|
90 |
+
def load_data() -> Union[Tuple[pandas.DataFrame, numpy.ndarray], List]:
|
91 |
"""
|
92 |
Return the data
|
93 |
|
|
|
95 |
None
|
96 |
|
97 |
Return:
|
98 |
+
The train, testing set and valid symptoms.
|
99 |
|
100 |
|
101 |
"""
|
|
|
113 |
y_test = df_test[TARGET_COLUMNS[0]]
|
114 |
X_test = df_test.drop(columns=TARGET_COLUMNS, axis=1, errors="ignore")
|
115 |
|
116 |
+
return (X_train, X_test), (y_train, y_test), X_train.columns.to_list()
|
117 |
|
118 |
|
119 |
def load_model(X_train: pandas.DataFrame, y_train: numpy.ndarray):
|