logasanjeev's picture
Update app.py
f4dcd5d verified
raw
history blame contribute delete
5.7 kB
import gradio as gr
import torch
import numpy as np
from transformers import BertForSequenceClassification, BertTokenizer
import requests
import json
import plotly.express as px
import pandas as pd
# Load model and tokenizer from Hugging Face Hub
repo_id = "logasanjeev/goemotions-bert"
model = BertForSequenceClassification.from_pretrained(repo_id)
tokenizer = BertTokenizer.from_pretrained(repo_id)
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
model.to(device)
if torch.cuda.device_count() > 1:
model = nn.DataParallel(model)
model.eval()
# Load optimized thresholds from Hugging Face Hub
thresholds_url = f"https://huggingface.co/{repo_id}/raw/main/thresholds.json"
response = requests.get(thresholds_url)
thresholds_data = json.loads(response.text)
emotion_labels = thresholds_data["emotion_labels"]
default_thresholds = thresholds_data["thresholds"]
# Prediction function
def predict_emotions(text, confidence_threshold=0.0):
encodings = tokenizer(
text,
padding='max_length',
truncation=True,
max_length=128,
return_tensors='pt'
)
input_ids = encodings['input_ids'].to(device)
attention_mask = encodings['attention_mask'].to(device)
with torch.no_grad():
outputs = model(input_ids, attention_mask=attention_mask)
logits = torch.sigmoid(outputs.logits).cpu().numpy()[0]
# Apply thresholds with user-defined confidence boost
predictions = []
for i, (logit, thresh) in enumerate(zip(logits, default_thresholds)):
adjusted_thresh = max(thresh, confidence_threshold)
if logit >= adjusted_thresh:
predictions.append((emotion_labels[i], logit))
predictions.sort(key=lambda x: x[1], reverse=True)
if not predictions:
return "No emotions predicted above thresholds.", None
# Format output
text_output = "\n".join([f"{emotion}: {confidence:.4f}" for emotion, confidence in predictions])
# Create bar chart
df = pd.DataFrame(predictions, columns=["Emotion", "Confidence"])
fig = px.bar(
df,
x="Emotion",
y="Confidence",
color="Emotion",
text="Confidence",
title="Emotion Confidence Levels",
height=400
)
fig.update_traces(texttemplate='%{text:.2f}', textposition='auto')
fig.update_layout(showlegend=False, margin=dict(t=40, b=40))
return text_output, fig
# Custom CSS for modern UI
custom_css = """
body {
font-family: 'Segoe UI', Arial, sans-serif;
}
.gr-panel {
border-radius: 12px;
box-shadow: 0 4px 12px rgba(0,0,0,0.1);
background: linear-gradient(145deg, #ffffff, #f0f4f8);
}
.gr-button {
border-radius: 8px;
background: #007bff;
color: white;
padding: 10px 20px;
transition: background 0.3s;
}
.gr-button:hover {
background: #0056b3;
}
#title {
font-size: 2.5em;
color: #1a3c6e;
text-align: center;
margin-bottom: 20px;
}
#description {
font-size: 1.1em;
color: #333;
text-align: center;
max-width: 700px;
margin: 0 auto;
}
#theme-toggle {
position: absolute;
top: 20px;
right: 20px;
}
.dark-mode {
background: #1a1a1a;
color: #e0e0e0;
}
.dark-mode .gr-panel {
background: linear-gradient(145deg, #2a2a2a, #3a3a3a);
}
.dark-mode #title {
color: #66b3ff;
}
.dark-mode #description {
color: #b0b0b0;
}
"""
# JavaScript for theme toggle
theme_js = """
function toggleTheme() {
document.body.classList.toggle('dark-mode');
}
"""
# Gradio Blocks UI
with gr.Blocks(css=custom_css) as demo:
# Header
gr.Markdown("<div id='title'>GoEmotions BERT Classifier</div>", elem_id="title")
gr.Markdown(
"""
<div id='description'>
Predict emotions from text using a fine-tuned BERT-base model.
Explore 28 emotions with optimized thresholds (Micro F1: 0.6025).
Try examples or enter your own text!
</div>
""",
elem_id="description"
)
# Theme toggle button
with gr.Row():
gr.HTML(
"""
<button id='theme-toggle' onclick='toggleTheme()'>Toggle Dark Mode</button>
<script>{}</script>
""".format(theme_js)
)
# Main input and output
with gr.Row():
with gr.Column(scale=1):
text_input = gr.Textbox(
label="Enter Your Text",
placeholder="Type something like 'I’m just chilling today'...",
lines=3
)
confidence_slider = gr.Slider(
minimum=0.0,
maximum=0.9,
value=0.0,
step=0.05,
label="Minimum Confidence Threshold",
info="Adjust to filter low-confidence predictions"
)
submit_btn = gr.Button("Predict Emotions", variant="primary")
with gr.Column(scale=1):
output_text = gr.Textbox(label="Predicted Emotions", lines=5)
output_plot = gr.Plot(label="Emotion Confidence Chart")
# Example carousel
examples = gr.Examples(
examples=[
"I’m just chilling today.",
"Thank you for saving my life!",
"I’m nervous about my exam tomorrow.",
"I love my new puppy so much!",
"I’m so relieved the storm passed."
],
inputs=text_input,
label="Try These Examples"
)
# Bind prediction
submit_btn.click(
fn=predict_emotions,
inputs=[text_input, confidence_slider],
outputs=[output_text, output_plot]
)
# Launch
if __name__ == "__main__":
demo.launch()