Spaces:
Running
Running
File size: 8,002 Bytes
1825fc1 d4ac1ba 75cfb71 d4ac1ba 6d68f91 d4ac1ba 1825fc1 d4ac1ba 1825fc1 d4ac1ba 75cfb71 d4ac1ba 75cfb71 d4ac1ba 1825fc1 d4ac1ba |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 |
import os
import toml
#from dotenv import load_dotenv
#load_dotenv()
# Load secrets from TOML file
#os.environ["GOOGLE_APPLICATION_CREDENTIALS"] = ".streamlit/secrets.toml"
""" google_creds = os.getenv('GOOGLE_APPLICATION_CREDENTIALS_JSON')
if google_creds:
with open('credentials.json', 'w') as f:
f.write(google_creds)
os.environ["GOOGLE_APPLICATION_CREDENTIALS"] = "credentials.json" """
import streamlit as st
from datetime import datetime
import pandas as pd
from lime.lime_text import LimeTextExplainer
from test import predict_hoax, predict_proba_for_lime
import streamlit.components.v1 as components
from load_model import load_model
from st_aggrid import AgGrid, GridOptionsBuilder, GridUpdateMode
from styles import COMMON_CSS
from google.cloud import storage
from io import StringIO
import pytz
def save_corrections_to_gcs(bucket_name, file_name, correction_data):
try:
client = storage.Client() # Uses the credentials set by the environment variable
bucket = client.bucket("dashboardindohoax-bucket")
blob = bucket.blob("koreksi_pengguna.csv")
# Check if the blob (file) exists
if blob.exists():
# Download existing CSV from GCS
existing_data = blob.download_as_string().decode('utf-8')
existing_df = pd.read_csv(StringIO(existing_data))
else:
# Create a new DataFrame if the file does not exist
existing_df = pd.DataFrame(columns=['Timestamp', 'Title', 'Content', 'Prediction', 'Correction'])
# Append the new data to the existing data
new_data_df = pd.DataFrame(correction_data)
updated_df = pd.concat([existing_df, new_data_df], ignore_index=True)
# Convert the DataFrame back to CSV and upload
updated_csv_data = updated_df.to_csv(index=False)
blob.upload_from_string(updated_csv_data, content_type='text/csv')
except Exception as e:
# If GCS fails, save locally as fallback
local_file = "koreksi_pengguna.csv"
if os.path.exists(local_file):
existing_df = pd.read_csv(local_file)
else:
existing_df = pd.DataFrame(columns=['Timestamp', 'Title', 'Content', 'Prediction', 'Correction'])
new_data_df = pd.DataFrame(correction_data)
updated_df = pd.concat([existing_df, new_data_df], ignore_index=True)
updated_df.to_csv(local_file, index=False)
st.warning(f"GCS save failed ({e}), saved locally to {local_file}.")
def show_deteksi_konten():
st.markdown(COMMON_CSS, unsafe_allow_html=True)
if 'correction' not in st.session_state:
st.session_state.correction = None
if 'detection_result' not in st.session_state:
st.session_state.detection_result = None
if 'lime_explanation' not in st.session_state:
st.session_state.lime_explanation = None
if 'headline' not in st.session_state:
st.session_state.headline = ""
if 'content' not in st.session_state:
st.session_state.content = ""
if 'is_correct' not in st.session_state:
st.session_state.is_correct = None
# Dropdown for selecting a model
st.markdown("<h6 style='font-size: 14px; margin-bottom: 0;'>Pilih Model</h6>", unsafe_allow_html=True)
selected_model = st.selectbox(
"",
[
"cahya/bert-base-indonesian-522M",
"indobenchmark/indobert-base-p2",
"indolem/indobert-base-uncased",
"mdhugol/indonesia-bert-sentiment-classification"
],
key="model_selector_content"
)
# Load the selected model
tokenizer, model = load_model(selected_model)
st.markdown("<h6 style='font-size: 14px; margin-bottom: 0;'>Masukkan Judul Berita :</h6>", unsafe_allow_html=True)
st.session_state.headline = st.text_input("", value=st.session_state.headline)
st.markdown("<h6 style='font-size: 14px; margin-bottom: 0;'>Masukkan Konten Berita :</h6>", unsafe_allow_html=True)
st.session_state.content = st.text_area("", value=st.session_state.content)
# Detection button
if st.button("Deteksi", key="detect_content"):
st.session_state.detection_result = predict_hoax(st.session_state.headline, st.session_state.content)
st.success(f"Prediksi: {st.session_state.detection_result}")
# Prepare the text for LIME
lime_texts = [f"{st.session_state.headline} [SEP] {st.session_state.content}"]
# Add a spinner and progress bar to indicate processing
with st.spinner("Sedang memproses LIME, harap tunggu..."):
# Explain the prediction
explainer = LimeTextExplainer(class_names=['NON-HOAX', 'HOAX'])
explanation = explainer.explain_instance(lime_texts[0], predict_proba_for_lime, num_features=10, num_samples=500)
# Save the LIME explanation in session state
st.session_state.lime_explanation = explanation.as_html()
# Display the detection result and LIME explanation if available
if st.session_state.lime_explanation:
lime_html = st.session_state.lime_explanation
# Inject CSS for font size adjustment
lime_html = f"""
<style>
.lime-text-explanation, .lime-highlight, .lime-classification,
.lime-text-explanation * {{
font-size: 14px !important;
}}
</style>
<div class="lime-text-explanation">
{lime_html}
</div>
"""
components.html(lime_html, height=200, scrolling=True)
# Display a radio button asking if the detection result is correct
if st.session_state.detection_result is not None:
st.markdown("<h6 style='font-size: 16px; margin-bottom: -150px;'>Apakah hasil deteksi sudah benar?</h6>", unsafe_allow_html=True)
st.session_state.is_correct = st.radio("", ("Ya", "Tidak"))
if st.session_state.is_correct == "Ya":
st.success("Deteksi sudah benar.")
else:
# Determine the correction based on the prediction
st.session_state.correction = "HOAX" if st.session_state.detection_result == "NON-HOAX" else "NON-HOAX"
wib = pytz.timezone('Asia/Jakarta')
# Display the correction DataFrame
correction_data = [{
'Title': st.session_state.headline,
'Content': st.session_state.content,
'Prediction': st.session_state.detection_result,
'Correction': st.session_state.correction,
'Timestamp': datetime.now(wib).strftime("%Y-%m-%d %H:%M:%S")
}]
# Save button
if st.button("Simpan"):
# Save the correction data to GCS
save_corrections_to_gcs("dashboardindohoax-bucket", "koreksi_pengguna.csv", correction_data)
# Create a formatted string with CSS for alignment and multi-line content handling
formatted_text = f"""
<div style='font-size: 14px;'>
<p style='margin: 0;'><span style='display: inline-block; width: 120px; font-weight: bold;'>Title</span> : <span style='white-space: pre-wrap;'>{st.session_state.headline}</span></p>
<p style='margin: 0;'><span style='display: inline-block; width: 120px; font-weight: bold;'>Content</span> : <span style='white-space: pre-wrap;'>{st.session_state.content}</span></p>
<p style='margin: 0;'><span style='display: inline-block; width: 120px; font-weight: bold;'>Prediction</span> : {st.session_state.detection_result}</p>
<p style='margin: 0;'><span style='display: inline-block; width: 120px; font-weight: bold;'>Correction</span> : {st.session_state.correction}</p>
</div>
"""
# Display the correction as text
st.markdown(formatted_text, unsafe_allow_html=True)
st.success("Koreksi telah disimpan.")
|