SumanthKarnati commited on
Commit
c2d79c4
·
1 Parent(s): 5071afd

Create app.py

Browse files
Files changed (1) hide show
  1. app.py +75 -0
app.py ADDED
@@ -0,0 +1,75 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Import necessary libraries
2
+ import streamlit as st
3
+ from transformers import AutoModelForSequenceClassification, AutoTokenizer, AutoModelForSeq2SeqLM
4
+ import torch
5
+
6
+ # Define function to load models
7
+ @st.cache(allow_output_mutation=True)
8
+ def load_models():
9
+ classification_model_name = 'distilbert-base-uncased-finetuned-sst-2-english'
10
+ classification_model = AutoModelForSequenceClassification.from_pretrained(classification_model_name)
11
+ classification_tokenizer = AutoTokenizer.from_pretrained(classification_model_name, model_max_length=512)
12
+
13
+ summarization_model_name = 't5-base'
14
+ summarization_model = AutoModelForSeq2SeqLM.from_pretrained(summarization_model_name)
15
+ summarization_tokenizer = AutoTokenizer.from_pretrained(summarization_model_name, model_max_length=512)
16
+
17
+ return classification_model, classification_tokenizer, summarization_model, summarization_tokenizer
18
+
19
+ classification_model, classification_tokenizer, summarization_model, summarization_tokenizer = load_models()
20
+
21
+ # Title of the app
22
+ st.title('Text Classification and Summarization with Hugging Face')
23
+
24
+ # Take user input
25
+ text = st.text_area("Enter text:", "")
26
+ submit_button = st.button("Analyze Text")
27
+
28
+ # Predict function for sentiment analysis
29
+ def predict_sentiment(text):
30
+ tokenized_text = classification_tokenizer.tokenize(text)
31
+ results = []
32
+
33
+ # Break text into chunks of max_model_length tokens
34
+ for i in range(0, len(tokenized_text), classification_tokenizer.model_max_length):
35
+ chunk = tokenized_text[i:i+classification_tokenizer.model_max_length]
36
+ chunk = classification_tokenizer.convert_tokens_to_string(chunk)
37
+
38
+ inputs = classification_tokenizer(chunk, return_tensors="pt", truncation=True, padding='max_length')
39
+ outputs = classification_model(**inputs)
40
+ probs = torch.nn.functional.softmax(outputs[0], dim=-1)
41
+ results.append(probs.detach().numpy())
42
+ return results
43
+
44
+ # Predict function for text summarization
45
+ def summarize_text(text):
46
+ tokenized_text = summarization_tokenizer.tokenize(text)
47
+ summaries = []
48
+
49
+ # Break text into chunks of max_model_length tokens
50
+ for i in range(0, len(tokenized_text), summarization_tokenizer.model_max_length):
51
+ chunk = tokenized_text[i:i+summarization_tokenizer.model_max_length]
52
+ chunk = summarization_tokenizer.convert_tokens_to_string(chunk)
53
+
54
+ inputs = summarization_tokenizer.encode("summarize: " + chunk, return_tensors="pt", truncation=True, padding='max_length')
55
+ outputs = summarization_model.generate(inputs, max_length=150, min_length=40, length_penalty=2.0, num_beams=4, early_stopping=True)
56
+ summary = summarization_tokenizer.decode(outputs[0]).replace('<pad>', '').replace('</s>', '')
57
+ summaries.append(summary)
58
+ return summaries
59
+
60
+ if submit_button:
61
+ if text:
62
+ with st.spinner("Analyzing..."):
63
+ # Sentiment analysis
64
+ results = predict_sentiment(text)
65
+ for i, probs in enumerate(results):
66
+ st.markdown(f"**Result {i+1}:**")
67
+ st.markdown(f"**Positive sentiment:** `{probs[0][1]:.2f}`")
68
+ st.markdown(f"**Negative sentiment:** `{probs[0][0]:.2f}`")
69
+
70
+ # Text summarization
71
+ summaries = summarize_text(text)
72
+ for i, summary in enumerate(summaries):
73
+ st.markdown(f"**Summary {i+1}:** `{summary}`")
74
+ else:
75
+ st.warning("Please enter text to analyze.")