cpu-demo / appStore /netzero.py
ppsingh's picture
Updating
6d737a4
# set path
import glob, os, sys;
sys.path.append('../utils')
#import needed libraries
import seaborn as sns
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import streamlit as st
from utils.netzero_classifier import load_netzeroClassifier, netzero_classification
import logging
logger = logging.getLogger(__name__)
from utils.config import get_classifier_params
from io import BytesIO
import xlsxwriter
import plotly.express as px
# Declare all the necessary variables
classifier_identifier = 'netzero'
params = get_classifier_params(classifier_identifier)
# Labels dictionary ###
_lab_dict = {
'NEGATIVE':'NO NETZERO TARGET',
'NA':'NOT APPLICABLE',
'NETZERO':'NETZERO TARGET',
}
@st.cache_data
def to_excel(df):
len_df = len(df)
output = BytesIO()
writer = pd.ExcelWriter(output, engine='xlsxwriter')
df.to_excel(writer, index=False, sheet_name='Sheet1')
workbook = writer.book
worksheet = writer.sheets['Sheet1']
worksheet.data_validation('E2:E{}'.format(len_df),
{'validate': 'list',
'source': ['No', 'Yes', 'Discard']})
writer.save()
processed_data = output.getvalue()
return processed_data
def app():
### Main app code ###
with st.container():
if 'key1' in st.session_state:
df = st.session_state.key1
# Load the classifier model
classifier = load_netzeroClassifier(classifier_name=params['model_name'])
st.session_state['{}_classifier'.format(classifier_identifier)] = classifier
if sum(df['Target Label'] == 'TARGET') > 100:
warning_msg = ": This might take sometime, please sit back and relax."
else:
warning_msg = ""
df = netzero_classification(haystack_doc=df,
threshold= params['threshold'])
st.session_state.key1 = df
def netzero_display():
if 'key1' in st.session_state:
df = st.session_state.key2
hits = df[df['Netzero Label'] == 'NETZERO']
range_val = min(5,len(hits))
if range_val !=0:
count_df = df['Netzero Label'].value_counts()
count_df = count_df.rename('count')
count_df = count_df.rename_axis('Netzero Label').reset_index()
count_df['Label_def'] = count_df['Netzero Label'].apply(lambda x: _lab_dict[x])
fig = px.bar(count_df, y="Label_def", x="count", orientation='h', height =200)
c1, c2 = st.columns([1,1])
with c1:
st.plotly_chart(fig,use_container_width= True)
hits = hits.sort_values(by=['Netzero Score'], ascending=False)
st.write("")
st.markdown("###### Top few NetZero Target Classified paragraph/text results ######")
range_val = min(5,len(hits))
for i in range(range_val):
# the page number reflects the page that contains the main paragraph
# according to split limit, the overlapping part can be on a separate page
st.write('**Result {}** `page {}` (Relevancy Score: {:.2f})'.format(i+1,hits.iloc[i]['page'],hits.iloc[i]['Netzero Score']))
st.write("\t Text: \t{}".format(hits.iloc[i]['text']))
else:
st.info("πŸ€” No Netzero target found")