Spaces:
Sleeping
Sleeping
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,76 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
import pandas as pd
|
3 |
+
import numpy as np
|
4 |
+
import dare
|
5 |
+
import time
|
6 |
+
|
7 |
+
from sklearn.model_selection import train_test_split
|
8 |
+
from sklearn.metrics import accuracy_score
|
9 |
+
|
10 |
+
# Load and prepare data
|
11 |
+
data = pd.read_csv("parkinsons.data")
|
12 |
+
data.columns = data.columns.str.replace(':', '_')
|
13 |
+
X = data.drop(columns=["name", "status"]).values
|
14 |
+
y = data["status"].values
|
15 |
+
|
16 |
+
# Train-test split
|
17 |
+
X_train, X_test, y_train, y_test = train_test_split(
|
18 |
+
X, y, test_size=0.15, stratify=y, random_state=42
|
19 |
+
)
|
20 |
+
|
21 |
+
# Function to train model, delete a sample, and retrain
|
22 |
+
def run_dare_demo(delete_index=25):
|
23 |
+
logs = ""
|
24 |
+
|
25 |
+
# Train initial model
|
26 |
+
model = dare.Forest(n_estimators=25, max_depth=3, random_state=42)
|
27 |
+
start = time.perf_counter()
|
28 |
+
model.fit(X_train, y_train)
|
29 |
+
train_time = time.perf_counter() - start
|
30 |
+
|
31 |
+
y_pred = model.predict(X_test)
|
32 |
+
acc_before = accuracy_score(y_test, y_pred)
|
33 |
+
|
34 |
+
logs += f"✅ Initial training completed in {train_time:.4f} seconds\n"
|
35 |
+
logs += f"🎯 Accuracy before unlearning: {acc_before:.4f}\n"
|
36 |
+
|
37 |
+
# Delete a data point
|
38 |
+
try:
|
39 |
+
start_del = time.perf_counter()
|
40 |
+
model.delete(delete_index)
|
41 |
+
delete_time = time.perf_counter() - start_del
|
42 |
+
y_pred_after = model.predict(X_test.astype(np.float32))
|
43 |
+
acc_after = accuracy_score(y_test, y_pred_after)
|
44 |
+
|
45 |
+
logs += f"\n🧽 Deleted index {delete_index} in {delete_time:.5f} seconds\n"
|
46 |
+
logs += f"🎯 Accuracy after unlearning: {acc_after:.4f}\n"
|
47 |
+
except Exception as e:
|
48 |
+
logs += f"\n⚠️ Error during unlearning: {str(e)}\n"
|
49 |
+
|
50 |
+
# Retrain from scratch for comparison
|
51 |
+
try:
|
52 |
+
X_retrain = np.delete(X_train, delete_index, axis=0)
|
53 |
+
y_retrain = np.delete(y_train, delete_index, axis=0)
|
54 |
+
|
55 |
+
retrain_model = dare.Forest(n_estimators=25, max_depth=3, random_state=42)
|
56 |
+
start_retrain = time.perf_counter()
|
57 |
+
retrain_model.fit(X_retrain, y_retrain)
|
58 |
+
retrain_time = time.perf_counter() - start_retrain
|
59 |
+
|
60 |
+
logs += f"\n🔁 Retraining completed in {retrain_time:.5f} seconds (without index {delete_index})\n"
|
61 |
+
except Exception as e:
|
62 |
+
logs += f"\n⚠️ Error during retraining: {str(e)}\n"
|
63 |
+
|
64 |
+
return logs
|
65 |
+
|
66 |
+
# Gradio Interface
|
67 |
+
iface = gr.Interface(
|
68 |
+
fn=run_dare_demo,
|
69 |
+
inputs=gr.Slider(0, len(X_train)-1, value=25, step=1, label="Data Point Index to Unlearn"),
|
70 |
+
outputs="text",
|
71 |
+
title="DaRE: Unlearning Demo on Parkinson's Dataset",
|
72 |
+
description="This demo shows how to train a DaRE forest, unlearn a data point, and retrain for comparison using the Parkinson's dataset."
|
73 |
+
)
|
74 |
+
|
75 |
+
if __name__ == "__main__":
|
76 |
+
iface.launch()
|