Update app.py
Browse files
app.py
CHANGED
@@ -1,97 +1,102 @@
|
|
1 |
# Import the libraries
|
2 |
|
3 |
-
import os
|
4 |
import uuid
|
5 |
import joblib
|
6 |
import json
|
7 |
|
8 |
import gradio as gr
|
9 |
-
import pandas as pd
|
|
|
|
|
|
|
10 |
|
11 |
# Run the training script placed in the same directory as app.py
|
12 |
-
# The training script will train and persist a
|
13 |
# model with the filename 'model.joblib'
|
14 |
|
15 |
os.system("python train.py")
|
16 |
|
17 |
-
|
18 |
# Load the freshly trained model from disk
|
19 |
|
20 |
-
insurance_charge_predictor = joblib.load(
|
21 |
|
22 |
# Prepare the logging functionality
|
23 |
-
# log_file = Path("logs/") / f"data_{uuid.uuid4()}.json"
|
24 |
-
# log_folder = log_file.parent
|
25 |
-
|
26 |
-
# scheduler = CommitScheduler(
|
27 |
-
# repo_id="insurance-charge-mlops-logs", # provide a name "insurance-charge-mlops-logs" for the repo_id
|
28 |
-
# repo_type="dataset",
|
29 |
-
# folder_path=log_folder,
|
30 |
-
# path_in_repo="data",
|
31 |
-
# every=2
|
32 |
-
# )
|
33 |
-
|
34 |
-
# Define the predict function which will take features, convert to dataframe and make predictions using the saved model
|
35 |
-
# the functions runs when 'Submit' is clicked or when a API request is made
|
36 |
-
def predict_charge(age, sex, bmi, children, somker, region):
|
37 |
-
smaple = {'age': age,
|
38 |
-
'bmi': bmi,
|
39 |
-
'children': children,
|
40 |
-
'sex': sex,
|
41 |
-
'smoker': smoker,
|
42 |
-
'region': region,
|
43 |
-
'prediction': prediction[0]
|
44 |
-
}
|
45 |
-
data_point = pd.DataFrame([sample])
|
46 |
-
prediction = insurance_charge_predicter.predict(data_point).tolist()
|
47 |
|
48 |
-
|
49 |
-
|
50 |
-
# access
|
51 |
-
|
52 |
-
# with scheduler.lock:
|
53 |
-
# with log_file.open("a") as f:
|
54 |
-
# f.write(json.dumps(
|
55 |
-
# {
|
56 |
-
# 'age': age,
|
57 |
-
# 'bmi': bmi,
|
58 |
-
# 'children': children,
|
59 |
-
# 'sex': sex,
|
60 |
-
# 'smoker': smoker,
|
61 |
-
# 'region': region,
|
62 |
-
# 'prediction': prediction[0]
|
63 |
-
# }
|
64 |
-
# ))
|
65 |
-
# f.write("\n")
|
66 |
|
67 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
68 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
69 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
70 |
|
71 |
# Set up UI components for input and output
|
72 |
|
73 |
-
age_input = gr.Number(label
|
74 |
-
bmi_input = gr.Number(label
|
75 |
-
children_input = gr.Number(label
|
76 |
-
sex_input = gr.Dropdown([
|
77 |
-
smoker_input = gr.Dropdown([
|
78 |
-
region_input = gr.Dropdown(
|
79 |
-
|
|
|
|
|
80 |
|
81 |
-
|
|
|
|
|
82 |
demo = gr.Interface(
|
83 |
fn=predict_insurance_charge,
|
84 |
-
inputs=[age_input, bmi_input, children_input,
|
85 |
-
outputs
|
86 |
-
title
|
87 |
-
description
|
88 |
-
examples
|
89 |
-
|
90 |
-
|
91 |
-
concurrency_limit
|
92 |
-
|
93 |
)
|
94 |
|
95 |
# Launch with a load balancer
|
96 |
demo.queue()
|
97 |
demo.launch(share=False)
|
|
|
|
1 |
# Import the libraries
|
2 |
|
3 |
+
import os
|
4 |
import uuid
|
5 |
import joblib
|
6 |
import json
|
7 |
|
8 |
import gradio as gr
|
9 |
+
import pandas as pd
|
10 |
+
|
11 |
+
from huggingface_hub import CommitScheduler
|
12 |
+
from pathlib import Path
|
13 |
|
14 |
# Run the training script placed in the same directory as app.py
|
15 |
+
# The training script will train and persist a logistic regression
|
16 |
# model with the filename 'model.joblib'
|
17 |
|
18 |
os.system("python train.py")
|
19 |
|
|
|
20 |
# Load the freshly trained model from disk
|
21 |
|
22 |
+
insurance_charge_predictor = joblib.load('model.joblib')
|
23 |
|
24 |
# Prepare the logging functionality
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
25 |
|
26 |
+
log_file = Path("logs/") / f"data_{uuid.uuid4()}.json"
|
27 |
+
log_folder = log_file.parent
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
28 |
|
29 |
+
scheduler = CommitScheduler(
|
30 |
+
repo_id="insurance-charge-logs",
|
31 |
+
repo_type="dataset",
|
32 |
+
folder_path=log_folder,
|
33 |
+
path_in_repo="data",
|
34 |
+
every=2
|
35 |
+
)
|
36 |
|
37 |
+
# Define the predict function that runs when 'Submit' is clicked or when a API request is made
|
38 |
+
def predict_insurance_charge(age, bmi, children,sex, smoker, region):
|
39 |
+
sample = {
|
40 |
+
'age': age,
|
41 |
+
'bmi': bmi,
|
42 |
+
'children': children,
|
43 |
+
'sex': sex,
|
44 |
+
'smoker': smoker,
|
45 |
+
'region': region
|
46 |
+
}
|
47 |
+
|
48 |
+
data_point = pd.DataFrame([sample])
|
49 |
+
prediction = insurance_charge_predictor.predict(data_point).tolist()
|
50 |
|
51 |
+
# While the prediction is made, log both the inputs and outputs to a local log file
|
52 |
+
# While writing to the log file, ensure that the commit scheduler is locked to avoid parallel
|
53 |
+
# access
|
54 |
+
|
55 |
+
with scheduler.lock:
|
56 |
+
with log_file.open("a") as f:
|
57 |
+
f.write(json.dumps(
|
58 |
+
{
|
59 |
+
'age': age,
|
60 |
+
'bmi': bmi,
|
61 |
+
'children': children,
|
62 |
+
'sex': sex,
|
63 |
+
'smoker': smoker,
|
64 |
+
'region': region,
|
65 |
+
'prediction': prediction[0]
|
66 |
+
}
|
67 |
+
))
|
68 |
+
f.write("\n")
|
69 |
+
|
70 |
+
return round(prediction[0],2)
|
71 |
|
72 |
# Set up UI components for input and output
|
73 |
|
74 |
+
age_input = gr.Number(label='age')
|
75 |
+
bmi_input = gr.Number(label='bmi')
|
76 |
+
children_input = gr.Number(label='children')
|
77 |
+
sex_input = gr.Dropdown(['female','male'],label='sex')
|
78 |
+
smoker_input = gr.Dropdown(['yes','no'],label='smoker')
|
79 |
+
region_input = gr.Dropdown(
|
80 |
+
['southeast', 'southwest', 'northwest', 'northeast'],
|
81 |
+
label='region'
|
82 |
+
)
|
83 |
|
84 |
+
model_output = gr.Label(label="Insurance Charges")
|
85 |
+
|
86 |
+
# Create the interface
|
87 |
demo = gr.Interface(
|
88 |
fn=predict_insurance_charge,
|
89 |
+
inputs=[age_input, bmi_input, children_input,sex_input, smoker_input, region_input],
|
90 |
+
outputs=model_output,
|
91 |
+
title="HealthyLife Insurance Charge Prediction",
|
92 |
+
description="This API allows you to predict the estimating insurance charges based on customer attributes",
|
93 |
+
examples=[[33,33.44,5,'male','no','southeast'],
|
94 |
+
[58,25.175,0,'male','no','northeast'],
|
95 |
+
[52,38.380,2,'female','no','northeast']],
|
96 |
+
concurrency_limit=16
|
|
|
97 |
)
|
98 |
|
99 |
# Launch with a load balancer
|
100 |
demo.queue()
|
101 |
demo.launch(share=False)
|
102 |
+
|