Spaces:
Sleeping
Sleeping
TahaRasouli
commited on
Commit
•
2427679
1
Parent(s):
9f1df33
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,82 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
#importing the necessary libraries
|
2 |
+
import gradio as gr
|
3 |
+
import numpy as np
|
4 |
+
import pandas as pd
|
5 |
+
import re
|
6 |
+
import torch
|
7 |
+
|
8 |
+
from transformers import AutoTokenizer, AutoModelForSequenceClassification
|
9 |
+
from topic_labels import labels
|
10 |
+
|
11 |
+
#Defining the models and tokenuzer
|
12 |
+
model_name = "valurank/distilroberta-topic-classification"
|
13 |
+
model = AutoModelForSequenceClassification.from_pretrained(model_name)
|
14 |
+
#model.to(device)
|
15 |
+
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
16 |
+
|
17 |
+
|
18 |
+
def clean_text(raw_text):
|
19 |
+
text = raw_text.encode("ascii", errors="ignore").decode(
|
20 |
+
"ascii"
|
21 |
+
) # remove non-ascii, Chinese characters
|
22 |
+
|
23 |
+
text = re.sub(r"\n", " ", text)
|
24 |
+
text = re.sub(r"\n\n", " ", text)
|
25 |
+
text = re.sub(r"\t", " ", text)
|
26 |
+
text = text.strip(" ")
|
27 |
+
text = re.sub(
|
28 |
+
" +", " ", text
|
29 |
+
).strip() # get rid of multiple spaces and replace with a single
|
30 |
+
|
31 |
+
text = re.sub(r"Date\s\d{1,2}\/\d{1,2}\/\d{4}", "", text) #remove date
|
32 |
+
text = re.sub(r"\d{1,2}:\d{2}\s[A-Z]+\s[A-Z]+", "", text) #remove time
|
33 |
+
|
34 |
+
return text
|
35 |
+
|
36 |
+
|
37 |
+
def find_two_highest_indices(arr):
|
38 |
+
if len(arr) < 2:
|
39 |
+
raise ValueError("Array must have at least two elements")
|
40 |
+
|
41 |
+
# Initialize the indices of the two highest values
|
42 |
+
max_idx = second_max_idx = None
|
43 |
+
|
44 |
+
for i, value in enumerate(arr):
|
45 |
+
if max_idx is None or value > arr[max_idx]:
|
46 |
+
second_max_idx = max_idx
|
47 |
+
max_idx = i
|
48 |
+
elif second_max_idx is None or value > arr[second_max_idx]:
|
49 |
+
second_max_idx = i
|
50 |
+
|
51 |
+
return max_idx, second_max_idx
|
52 |
+
|
53 |
+
|
54 |
+
def predict_topic(text):
|
55 |
+
text = clean_text(text)
|
56 |
+
dict_topic = {}
|
57 |
+
|
58 |
+
input_tensor = tokenizer.encode(text, return_tensors="pt", truncation=True)
|
59 |
+
logits = model(input_tensor).logits
|
60 |
+
|
61 |
+
softmax = torch.nn.Softmax(dim=1)
|
62 |
+
probs = softmax(logits)[0]
|
63 |
+
probs = probs.cpu().detach().numpy()
|
64 |
+
|
65 |
+
max_index = find_two_highest_indices(probs)
|
66 |
+
emotion_1, emotion_2 = labels[max_index[0]], labels[max_index[1]]
|
67 |
+
probs_1, probs_2 = probs[max_index[0]], probs[max_index[1]]
|
68 |
+
dict_topic[emotion_1] = round((probs_1), 2)
|
69 |
+
|
70 |
+
#if probs_2 > 0.01:
|
71 |
+
dict_topic[emotion_2] = round((probs_2), 2)
|
72 |
+
|
73 |
+
return dict_topic
|
74 |
+
|
75 |
+
|
76 |
+
#Creating the interface for the radio appdemo = gr.Interface(multi_label_emotions, inputs=gr.Textbox(),
|
77 |
+
demo = gr.Interface(predict_topic, inputs=gr.Textbox(),
|
78 |
+
outputs = gr.Label(num_top_classes=2),
|
79 |
+
title="Topic Classification")
|
80 |
+
|
81 |
+
if __name__ == "__main__":
|
82 |
+
demo.launch(debug=True)
|