Spaces:
Sleeping
Sleeping
initial commit
Browse files
app.py
ADDED
@@ -0,0 +1,110 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import streamlit as st
|
2 |
+
from transformers import AutoTokenizer, AutoModelForSequenceClassification
|
3 |
+
import torch
|
4 |
+
import re
|
5 |
+
# Load the model and tokenizer
|
6 |
+
tokenizer = AutoTokenizer.from_pretrained("nebiyu29/fintunned-v2-roberta_GA")
|
7 |
+
model = AutoModelForSequenceClassification.from_pretrained("nebiyu29/fintunned-v2-roberta_GA")
|
8 |
+
|
9 |
+
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
10 |
+
model = model.to(device)
|
11 |
+
|
12 |
+
|
13 |
+
|
14 |
+
# Define a function to split a text into segments of 512 tokens
|
15 |
+
def split_text(text):
|
16 |
+
text=re.sub(r'[^a-zA-Z\s]','',text)
|
17 |
+
text=str(text)
|
18 |
+
# Tokenize the text
|
19 |
+
tokens = tokenizer.tokenize(text)
|
20 |
+
# Initialize an empty list for segments
|
21 |
+
segments = []
|
22 |
+
# Initialize an empty list for current segment
|
23 |
+
current_segment = []
|
24 |
+
# Initialize a counter for tokens
|
25 |
+
token_count = 0
|
26 |
+
# Loop through the tokens
|
27 |
+
for token in tokens:
|
28 |
+
# Add the token to the current segment
|
29 |
+
current_segment.append(token)
|
30 |
+
# Increment the token count
|
31 |
+
token_count += 1
|
32 |
+
# If the token count reaches 512 or the end of the text, add the current segment to the segments list
|
33 |
+
if token_count == 512 or token == tokens[-1]:
|
34 |
+
# Convert the current segment to a string and add it to the segments list
|
35 |
+
segments.append(tokenizer.convert_tokens_to_string(current_segment))
|
36 |
+
# Reset the current segment and the token count
|
37 |
+
current_segment = []
|
38 |
+
token_count = 0
|
39 |
+
# Return the segments list
|
40 |
+
return segments
|
41 |
+
|
42 |
+
def classify(text):
|
43 |
+
# Define the labels
|
44 |
+
labels = ["depression", "anxiety", "bipolar disorder", "schizophrenia", "PTSD", "OCD", "ADHD", "autism", "eating disorder", "personality disorder", "phobia"]
|
45 |
+
#labels=list(model.config.id2label)
|
46 |
+
# Encode the labels
|
47 |
+
label_encodings = tokenizer(labels, padding=True, return_tensors="pt")
|
48 |
+
# Split the text into segments
|
49 |
+
segments = split_text(text)
|
50 |
+
# Initialize an empty list for logits
|
51 |
+
logits_list = []
|
52 |
+
|
53 |
+
# Loop through the segments
|
54 |
+
for segment in segments:
|
55 |
+
# Encode the segment and the labels
|
56 |
+
inputs = tokenizer([segment] + labels, padding=True, return_tensors="pt")
|
57 |
+
# Get the input ids and attention mask
|
58 |
+
input_ids = inputs["input_ids"]
|
59 |
+
attention_mask = inputs["attention_mask"]
|
60 |
+
# Move the input ids and attention mask to the device
|
61 |
+
input_ids = input_ids.to(device)
|
62 |
+
attention_mask = attention_mask.to(device)
|
63 |
+
# Get the model outputs for each segment
|
64 |
+
|
65 |
+
with torch.no_grad():
|
66 |
+
outputs = model(
|
67 |
+
input_ids,
|
68 |
+
attention_mask=attention_mask,
|
69 |
+
)
|
70 |
+
# Get the logits for each segment and append them to the logits list
|
71 |
+
logits = outputs.logits
|
72 |
+
logits_list.append(logits)
|
73 |
+
# Average the logits across the segments
|
74 |
+
avg_logits = torch.mean(torch.stack(logits_list), dim=0)
|
75 |
+
# Apply softmax to convert logits to probabilities
|
76 |
+
probabilities = torch.softmax(avg_logits, dim=1)
|
77 |
+
# Get the probabilities for each label
|
78 |
+
label_probabilities = probabilities[:, :len(labels)].tolist()
|
79 |
+
|
80 |
+
# Get the top 3 most likely labels and their probabilities
|
81 |
+
# Get the top 3 most likely labels and their probabilities
|
82 |
+
top_labels = []
|
83 |
+
top_probabilities = []
|
84 |
+
label_probabilities = label_probabilities[0] # Extract the list of probabilities for the first (and only) example
|
85 |
+
for _ in range(3):
|
86 |
+
max_prob_index = label_probabilities.index(max(label_probabilities))
|
87 |
+
top_labels.append(labels[max_prob_index])
|
88 |
+
top_probabilities.append(max(label_probabilities))
|
89 |
+
label_probabilities[max_prob_index] = 0 # Set the max probability to 0 to get the next highest probability
|
90 |
+
|
91 |
+
# Create a dictionary to store the results
|
92 |
+
results = {
|
93 |
+
"sequence": text,
|
94 |
+
"top_labels": top_labels,
|
95 |
+
"top_probabilities": top_probabilities
|
96 |
+
}
|
97 |
+
|
98 |
+
return results
|
99 |
+
|
100 |
+
# Streamlit app
|
101 |
+
st.title("Text Classification Demo")
|
102 |
+
st.write("Enter some text, and the model will classify it.")
|
103 |
+
|
104 |
+
text_input = st.text_input("Text Input")
|
105 |
+
if st.button("Classify"):
|
106 |
+
predictions = classify(text_input)
|
107 |
+
for prediction in predictions:
|
108 |
+
# st.write(f"Segment Text: {prediction['segment_text']}")
|
109 |
+
st.write(f"Label: {prediction['top_labels']}")
|
110 |
+
st.write(f"Probability: {prediction['top_probabilities']}")
|