awacke1's picture
Update app.py
24be11a
import streamlit as st
import altair as alt
import torch
from transformers import AlbertTokenizer, AlbertForSequenceClassification
import sentencepiece as spm
import pandas as pd
# Load pre-trained model and tokenizer
model_name = "albert-base-v2"
tokenizer = AlbertTokenizer.from_pretrained(model_name)
model = AlbertForSequenceClassification.from_pretrained(model_name)
# Define function to classify input text
def classify_text(text):
inputs = tokenizer(text, padding=True, truncation=True, return_tensors="pt")
outputs = model(**inputs)
logits = outputs.logits.detach().numpy()[0]
probabilities = torch.softmax(torch.tensor(logits), dim=0).tolist()
return probabilities
# Set up Streamlit app
st.title("ALBERT Text Classification App")
# Create input box for user to enter text
default_text = "Streamlit-Altair: A component that allows the creation of Altair visualizations within Streamlit.\nStreamlit-Bokeh: A component that allows the creation of Bokeh visualizations within Streamlit.\nStreamlit-Plotly: A component that allows the creation of Plotly visualizations within Streamlit.\nStreamlit-Mapbox: A component that allows the creation of Mapbox maps within Streamlit.\nStreamlit-DeckGL: A component that allows the creation of Deck.GL visualizations within Streamlit.\nStreamlit-Wordcloud: A component that allows the creation of word clouds within Streamlit.\nStreamlit-Audio: A component that allows the playing of audio files within Streamlit.\nStreamlit-Video: A component that allows the playing of video files within Streamlit.\nStreamlit-EmbedCode: A component that allows the embedding of code snippets within Streamlit.\nStreamlit-Components: A component that provides a library of custom Streamlit components created by the Streamlit community."
text_input = st.text_area("Enter text to classify", default_text, height=200)
# Classify input text and display results
if st.button("Classify"):
if text_input:
probabilities = classify_text(text_input)
df = pd.DataFrame({
'Label': ['Negative', 'Positive'],
'Probability': probabilities
})
chart = alt.Chart(df).mark_bar().encode(
x='Probability',
y=alt.Y('Label', sort=['Negative', 'Positive'])
)
st.write(chart)
else:
st.write("Please enter some text to classify.")