Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,170 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
|
2 |
+
import streamlit as st
|
3 |
+
import numpy as np
|
4 |
+
import pandas as pd
|
5 |
+
import json
|
6 |
+
import base64
|
7 |
+
import uuid
|
8 |
+
from pandas import DataFrame
|
9 |
+
import time
|
10 |
+
import re
|
11 |
+
|
12 |
+
def download_button(object_to_download, download_filename, button_text):
|
13 |
+
|
14 |
+
if isinstance(object_to_download, bytes):
|
15 |
+
pass
|
16 |
+
|
17 |
+
elif isinstance(object_to_download, pd.DataFrame):
|
18 |
+
object_to_download = object_to_download.to_csv(index=False)
|
19 |
+
# Try JSON encode for everything else
|
20 |
+
else:
|
21 |
+
object_to_download = json.dumps(object_to_download)
|
22 |
+
|
23 |
+
try:
|
24 |
+
# some strings <-> bytes conversions necessary here
|
25 |
+
b64 = base64.b64encode(object_to_download.encode()).decode()
|
26 |
+
except AttributeError as e:
|
27 |
+
b64 = base64.b64encode(object_to_download).decode()
|
28 |
+
|
29 |
+
button_uuid = str(uuid.uuid4()).replace("-", "")
|
30 |
+
button_id = re.sub("\d+", "", button_uuid)
|
31 |
+
|
32 |
+
custom_css = f"""
|
33 |
+
<style>
|
34 |
+
#{button_id} {{
|
35 |
+
display: inline-flex;
|
36 |
+
align-items: center;
|
37 |
+
justify-content: center;
|
38 |
+
background-color: rgb(255, 255, 255);
|
39 |
+
color: rgb(38, 39, 48);
|
40 |
+
padding: .25rem .75rem;
|
41 |
+
position: relative;
|
42 |
+
text-decoration: none;
|
43 |
+
border-radius: 4px;
|
44 |
+
border-width: 1px;
|
45 |
+
border-style: solid;
|
46 |
+
border-color: rgb(230, 234, 241);
|
47 |
+
border-image: initial;
|
48 |
+
}}
|
49 |
+
#{button_id}:hover {{
|
50 |
+
border-color: rgb(246, 51, 102);
|
51 |
+
color: rgb(246, 51, 102);
|
52 |
+
}}
|
53 |
+
#{button_id}:active {{
|
54 |
+
box-shadow: none;
|
55 |
+
background-color: rgb(246, 51, 102);
|
56 |
+
color: white;
|
57 |
+
}}
|
58 |
+
</style> """
|
59 |
+
|
60 |
+
dl_link = (
|
61 |
+
custom_css
|
62 |
+
+ f'<a download="{download_filename}" id="{button_id}" href="data:file/txt;base64,{b64}">{button_text}</a><br><br>'
|
63 |
+
)
|
64 |
+
# dl_link = f'<a download="{download_filename}" id="{button_id}" href="data:file/txt;base64,{b64}"><input type="button" kind="primary" value="{button_text}"></a><br></br>'
|
65 |
+
|
66 |
+
st.markdown(dl_link, unsafe_allow_html=True)
|
67 |
+
|
68 |
+
|
69 |
+
|
70 |
+
class c_model:
|
71 |
+
def __init__(self):
|
72 |
+
# st.write('my model')
|
73 |
+
pass
|
74 |
+
|
75 |
+
@st.cache
|
76 |
+
def load_model(self, name_or_path):
|
77 |
+
time.sleep(3)
|
78 |
+
return None
|
79 |
+
|
80 |
+
def predict(self, texts):
|
81 |
+
return np.random.randint(2), np.random.rand()
|
82 |
+
|
83 |
+
|
84 |
+
st.title('Sentiment Analysis')
|
85 |
+
|
86 |
+
|
87 |
+
# Load classification model
|
88 |
+
with st.spinner('Loading classification model...'):
|
89 |
+
from transformers import pipeline
|
90 |
+
|
91 |
+
checkpoint = "amir7d0/distilbert-base-uncased-finetuned-amazon-reviews"
|
92 |
+
classifier = pipeline("text-classification", model=checkpoint)
|
93 |
+
|
94 |
+
|
95 |
+
tab1, tab2 = st.tabs(["Single Comment", "Multiple Comment"])
|
96 |
+
|
97 |
+
with tab1:
|
98 |
+
st.subheader('Single comment classification')
|
99 |
+
text_input = st.text_area(label='Paste your text below (max 256 words)',
|
100 |
+
value='Hiiiiiiiii')
|
101 |
+
MAX_WORDS = 256
|
102 |
+
res = len(re.findall(r"\w+", text_input))
|
103 |
+
if res > MAX_WORDS:
|
104 |
+
st.warning(
|
105 |
+
"β οΈ Your text contains "
|
106 |
+
+ str(res)
|
107 |
+
+ " words."
|
108 |
+
+ " Only the first 256 words will be reviewed! π"
|
109 |
+
)
|
110 |
+
text_input = text_input[:MAX_WORDS]
|
111 |
+
|
112 |
+
submit_button = st.button(label='Submit comment')
|
113 |
+
if submit_button:
|
114 |
+
with st.spinner('Predicting ...'):
|
115 |
+
start_time = time.time()
|
116 |
+
time.sleep(2)
|
117 |
+
preds = classifier([text_input])[0]
|
118 |
+
end_time = time.time()
|
119 |
+
p_time = round(end_time-start_time, 2)
|
120 |
+
st.success(f'Prediction finished in {p_time}s!')
|
121 |
+
|
122 |
+
st.write(f'Label: {preds["label"]}, with certainty: {preds["score"]}')
|
123 |
+
|
124 |
+
|
125 |
+
with tab2:
|
126 |
+
st.subheader('Multiple comment classification')
|
127 |
+
file_input = st.file_uploader(label='Choose a file:', type='csv')
|
128 |
+
if file_input:
|
129 |
+
try:
|
130 |
+
df = pd.read_csv(file_input)
|
131 |
+
texts = df['text'].to_list()
|
132 |
+
except:
|
133 |
+
st.write('Bad File Error...')
|
134 |
+
|
135 |
+
st.write(f"First 5 rows of {file_input.name} texts")
|
136 |
+
st.write(texts[:5])
|
137 |
+
|
138 |
+
submit_button = st.button(label='Submit file')
|
139 |
+
if submit_button:
|
140 |
+
with st.spinner('Predicting ...'):
|
141 |
+
start_time = time.time()
|
142 |
+
time.sleep(2)
|
143 |
+
preds = classifier(texts)
|
144 |
+
end_time = time.time()
|
145 |
+
p_time = round(end_time-start_time, 2)
|
146 |
+
st.success(f'Prediction finished in {p_time}s!')
|
147 |
+
|
148 |
+
c1, c2 = st.columns([3, 1])
|
149 |
+
with c1:
|
150 |
+
st.subheader("π Check & download results")
|
151 |
+
with c2:
|
152 |
+
CSVButton2 = download_button(results, "Data.csv", "π₯ Download (.csv)")
|
153 |
+
|
154 |
+
st.header("")
|
155 |
+
|
156 |
+
for text, pred in zip(texts, preds):
|
157 |
+
pred['text'] = text
|
158 |
+
|
159 |
+
df = pd.DataFrame(preds, columns=['text', 'label', 'score'])
|
160 |
+
|
161 |
+
import seaborn as sns
|
162 |
+
# Add styling
|
163 |
+
cmGreen = sns.light_palette("green", as_cmap=True)
|
164 |
+
cmRed = sns.light_palette("red", as_cmap=True)
|
165 |
+
df = df.style.background_gradient(
|
166 |
+
cmap=cmGreen,
|
167 |
+
subset=["score"],
|
168 |
+
)
|
169 |
+
|
170 |
+
st.table(df)
|