Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -1,9 +1,9 @@
|
|
1 |
import streamlit as st
|
2 |
import pandas as pd
|
3 |
-
import os
|
4 |
import torch
|
5 |
from transformers import DistilBertTokenizerFast
|
6 |
from transformers import DistilBertForSequenceClassification
|
|
|
7 |
|
8 |
|
9 |
def getTop95(predictions):
|
@@ -11,6 +11,26 @@ def getTop95(predictions):
|
|
11 |
vals, ids = torch.topk(predictions, i)
|
12 |
if torch.sum(vals).item() >= 0.95:
|
13 |
return ids
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
14 |
|
15 |
|
16 |
st.set_page_config(
|
@@ -20,7 +40,7 @@ st.set_page_config(
|
|
20 |
|
21 |
st.header("Theme classification of ArXiv articles")
|
22 |
st.markdown("""
|
23 |
-
|
24 |
""")
|
25 |
|
26 |
with st.form(key='input_form'):
|
@@ -28,29 +48,18 @@ with st.form(key='input_form'):
|
|
28 |
summary = st.text_area("Enter summary of the article here")
|
29 |
submit = st.form_submit_button(label='Analyze')
|
30 |
|
31 |
-
if submit
|
32 |
-
|
33 |
-
|
34 |
-
|
35 |
-
|
36 |
-
|
37 |
-
|
38 |
-
|
39 |
-
|
40 |
-
|
41 |
-
|
42 |
-
|
43 |
-
model.eval()
|
44 |
-
logits = model(tokens, mask)[0][0]
|
45 |
-
softmax = torch.nn.Softmax()
|
46 |
-
predictions = softmax(logits)
|
47 |
-
ids = getTop95(predictions)
|
48 |
-
st.markdown("Most likely it is:")
|
49 |
-
for tag in classes.to_numpy()[ids[:5]]:
|
50 |
-
st.markdown(f"- {tag[1]}")
|
51 |
-
st.markdown("Other possible variants:")
|
52 |
-
st.write(', '.join(classes.tag.to_numpy()[ids[5:]]))
|
53 |
-
st.balloons()
|
54 |
|
55 |
hide_streamlit_style = """
|
56 |
<style>
|
|
|
1 |
import streamlit as st
|
2 |
import pandas as pd
|
|
|
3 |
import torch
|
4 |
from transformers import DistilBertTokenizerFast
|
5 |
from transformers import DistilBertForSequenceClassification
|
6 |
+
import os
|
7 |
|
8 |
|
9 |
def getTop95(predictions):
|
|
|
11 |
vals, ids = torch.topk(predictions, i)
|
12 |
if torch.sum(vals).item() >= 0.95:
|
13 |
return ids
|
14 |
+
|
15 |
+
|
16 |
+
@st.cache(show_spinner=False)
|
17 |
+
def predict(text):
|
18 |
+
classes = pd.read_csv('classes.csv')
|
19 |
+
tokenizer = DistilBertTokenizerFast.from_pretrained("distilbert-base-cased")
|
20 |
+
to_predict = title + '|' + summary
|
21 |
+
X = tokenizer(to_predict, truncation=True, padding=True)
|
22 |
+
tokens = torch.tensor(X['input_ids']).unsqueeze(0)
|
23 |
+
mask = torch.tensor(X['attention_mask']).unsqueeze(0)
|
24 |
+
model = DistilBertForSequenceClassification.from_pretrained(
|
25 |
+
os.getcwd(),
|
26 |
+
num_labels=len(classes)
|
27 |
+
)
|
28 |
+
model.eval()
|
29 |
+
logits = model(tokens, mask)[0][0]
|
30 |
+
softmax = torch.nn.Softmax()
|
31 |
+
predictions = softmax(logits)
|
32 |
+
ids = getTop95(predictions)
|
33 |
+
return classes.tag.to_numpy()[ids]
|
34 |
|
35 |
|
36 |
st.set_page_config(
|
|
|
40 |
|
41 |
st.header("Theme classification of ArXiv articles")
|
42 |
st.markdown("""
|
43 |
+
Please enter title and summary (at least one is required) and oracul will predict classes of the arcticle according to taxonometry of ArXiv.
|
44 |
""")
|
45 |
|
46 |
with st.form(key='input_form'):
|
|
|
48 |
summary = st.text_area("Enter summary of the article here")
|
49 |
submit = st.form_submit_button(label='Analyze')
|
50 |
|
51 |
+
if submit:
|
52 |
+
if not title and not summary:
|
53 |
+
st.markdown('Please enter at least one: title or summary')
|
54 |
+
else:
|
55 |
+
with st.spinner(text='Oracul thinks, please wait for his wise prediction'):
|
56 |
+
prediction = predict(title + '|' + summary)
|
57 |
+
st.markdown("Most likely it is:")
|
58 |
+
for tag in prediction[:5]:
|
59 |
+
st.markdown(f"- {tag}")
|
60 |
+
st.markdown("Other possible variants:")
|
61 |
+
st.write(', '.join(prediction[5:]))
|
62 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
63 |
|
64 |
hide_streamlit_style = """
|
65 |
<style>
|