SwiftML / app.py
arpy8's picture
add app
6d584c3
Raw History Blame Contribute Delete
9.11 kB
import time
import pickle
import base64
import pandas as pd
import streamlit as st
from st_on_hover_tabs import on_hover_tabs
from constants import BACKGROUND, PAGE_ICON, MODEL_DETAILS_TEMPLATE
from train_model import process_dataset, display_response
st.set_page_config(page_title='SwiftML', page_icon=PAGE_ICON, layout='wide')
def set_page_background(png_file):
@st.cache_data()
def get_base64_of_bin_file(bin_file):
with open(bin_file, 'rb') as f:
data = f.read()
return base64.b64encode(data).decode()
bin_str = get_base64_of_bin_file(png_file)
page_bg_img = '''
<style>
{header_css}
.stApp {{
background-image: url("data:image/png;base64,{bin_str}");
}}
</style>
'''.format(
header_css=open('assets/css/styles.css').read(),
bin_str=bin_str
)
st.markdown(page_bg_img, unsafe_allow_html=True)
set_page_background(BACKGROUND)
with st.columns([3,3,1])[2]:
st.write(open('assets/html_components/logo.html', 'r').read(), unsafe_allow_html=True)
with st.sidebar:
st.write('<br>', unsafe_allow_html=True)
selected_task = on_hover_tabs(
tabName=['Home Page', 'Process Dataset', 'Predict Values', 'Download Model', 'Learn ML', 'About Us'],
iconName=['home', 'engineering', 'lightbulb', 'download', 'school', 'contact_support'],
styles = {
'navtab': {'background-color':'#ffffff',
'color': '#ff0290',
'padding': '40px 0px 10px 0px',
'border-radius': '23px',
'font-size': '18px',
'transition': '.5s',
'white-space': 'nowrap',
'text-transform': 'uppercase',
},
'tabOptionsStyle': {':hover :hover': {'color': '#170034',
'cursor': 'pointer'},
},
'iconStyle':{'position':'fixed',
'left':'11.5px'},
'tabStyle' : {'background-color':'rgba(0, 0, 0, 0)',
'list-style-type': 'none',
'margin-bottom': '30px',
},
},
default_choice=1)
if selected_task == 'Home Page':
st.write(open('assets/html_components/home.html', 'r').read(), unsafe_allow_html=True)
elif selected_task == 'Process Dataset':
try:
header_container = st.empty()
header_container.write("""
<div style='text-align:center;'>
<h1 style='text-align:center; font-size: 300%;'>Process Dataset</h1>
<p style=' color: #9c9d9f'>Upload your dataset and watch us do the magic.</p>
<hr>
</div>
""", unsafe_allow_html=True)
uploaded_file = st.file_uploader('Please upload a dataset', type=['csv'])
# uploaded_file = r'misc\Iris.csv'
dataset_option_columns = st.columns([2,2,3,1])
if uploaded_file is not None:
uploaded_data_df = pd.read_csv(uploaded_file)
with st.expander("Preview Data", expanded=False):
with st.columns([1,100,1])[1]:
uploaded_data_container = st.empty()
with dataset_option_columns[0]:
st.write("##### Select Target Column")
target = st.selectbox(
label='Select Target Column',
options=uploaded_data_df.columns[::-1],
label_visibility="hidden",
)
with dataset_option_columns[1]:
st.write("##### Select Target Type")
problem_type = st.selectbox(
label='Select Target Type',
options=["Classification", "Regression"],
label_visibility="hidden",
disabled=False
)
with dataset_option_columns[2]:
st.write("##### Select Extra Columns")
metadata_columns = st.multiselect(
label='Select Extra Columns',
options=[i for i in uploaded_data_df.columns if i!=target],
label_visibility="hidden",
placeholder="Select extra columns to drop"
)
trimmed_uploaded_data = uploaded_data_df[[column for column in uploaded_data_df.columns if column not in metadata_columns]]
with dataset_option_columns[3]:
st.write("<br><br>", unsafe_allow_html=True)
submit_uploaded_file_container = st.empty()
uploaded_data_container.dataframe(trimmed_uploaded_data, hide_index=True, width=2000)
if submit_uploaded_file_container.button(label='Submit', use_container_width=True):
header_container.empty()
submit_uploaded_file_container.empty()
st.write("---")
start_time = time.time()
with st.spinner("Processing data..."):
api_response = process_dataset(trimmed_uploaded_data, target, problem_type.lower())
if api_response:
display_response()
submit_uploaded_file_container.button(label='Submit', use_container_width=True, key='submit_button2')
end_time = time.time()
st.toast(f"Processed finished in : {end_time-start_time:.2f} seconds.")
else:
if "final_model" in st.session_state:
with st.expander("Preview Data", expanded=False):
with st.columns([1,100,1])[1]:
st.dataframe(st.session_state['data'], hide_index=True, width=2000)
display_response()
except UnicodeDecodeError:
st.error('Error reading file: UnicodeDecodeError')
elif selected_task == 'Predict Values':
if "final_model" in st.session_state:
st.write("<center><h1>Predict Values</h1></center><br><br>", unsafe_allow_html=True)
st.caption("We've automatically populated the values from the test dataset. Feel free to edit them.")
final_model = st.session_state['final_model']
X_train_columns = st.session_state['X_train_columns']
pycaret_base = st.session_state['pycaret_base']
prediction_df = pd.DataFrame(
[
st.session_state['X_test'][0],
], columns=X_train_columns
)
edited_df = st.data_editor(prediction_df, hide_index=True, width=2000)
with st.columns(len(X_train_columns))[-1]: predict_button = st.button('Predict', use_container_width=True)
if predict_button:
st.write("---")
prediction = pycaret_base.predict_model(final_model, data=prediction_df)
st.info(f"""
Predicted class/value is: **{prediction['prediction_label'].values[0]}**
""")
else:
st.write("<center><h1>Predict Values</h1></center><br><br>", unsafe_allow_html=True)
st.info("Please train a model first.")
elif selected_task == 'Download Model':
with st.columns([1,4,1])[1]:
st.write("<center><h1>Download Model</h1></center><br><br>", unsafe_allow_html=True)
if "final_model" in st.session_state:
st.write(MODEL_DETAILS_TEMPLATE.format(
model_name=st.session_state['model_name'],
final_model=str(st.session_state['final_model']),
model_definition=st.session_state['model_definition'],
model_info_link=st.session_state['model_info_link']
), unsafe_allow_html=True)
with open('final_model.pkl', 'wb') as f:
pickle.dump(st.session_state['final_model'], f)
st.write("Click the button below to download the trained model.")
st.download_button(label='Download Model', data=open('final_model.pkl', 'rb').read(), file_name='final_model.pkl', mime='application/octet-stream', use_container_width=True)
else:
st.info("Please train a model first.")
elif selected_task == 'Learn ML':
with st.columns([1,4,1])[1]:
st.write(open('assets/html_components/learn.md', 'r').read(), unsafe_allow_html=True)
elif selected_task == 'About Us':
with st.columns([1,4,1])[1]:
st.write(open('assets/html_components/about.html', 'r').read(), unsafe_allow_html=True)
st.write(open('assets/html_components/team.html', 'r').read(), unsafe_allow_html=True)