app.py

import pandas as pd
import streamlit as st
from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch
from datasets import load_dataset
import time

# Check if GPU is available
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

# Load the model and tokenizer
model_name = "SamLowe/roberta-base-go_emotions"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSequenceClassification.from_pretrained(model_name).to(device)

# Define the emotion labels (based on the GoEmotions dataset)
emotion_labels = ["admiration", "amusement", "anger", "annoyance", "approval",
                  "caring", "confusion", "curiosity", "desire", "disappointment",
                  "disapproval", "disgust", "embarrassment", "excitement", "fear",
                  "gratitude", "grief", "joy", "love", "nervousness", "optimism",
                  "pride", "realization", "relief", "remorse", "sadness", "surprise",
                  "neutral"]

# Function to classify emotions in batches and save results after each batch
def classify_emotions_in_batches(texts, batch_size=64, output_file="enron_emails_with_emotions.csv"):
    results = []
    start_time = time.time()

    # DataFrame to store the results
    result_df = pd.DataFrame(columns=['From', 'To', 'body', 'emotion'])

    for i in range(0, len(texts), batch_size):
        batch = texts[i:i+batch_size]
        inputs = tokenizer(batch, return_tensors="pt", truncation=True, padding=True).to(device)
        with torch.no_grad():
            outputs = model(**inputs)
            logits = outputs.logits
            predicted_class_ids = torch.argmax(logits, dim=-1).tolist()
            results.extend(predicted_class_ids)

        # Save the batch results
        batch_results = {
            'From': enron_data['From'][i:i+batch_size],
            'To': enron_data['To'][i:i+batch_size],
            'body': batch,
            'emotion': [emotion_labels[idx] for idx in predicted_class_ids]
        }
        batch_df = pd.DataFrame(batch_results)
        result_df = pd.concat([result_df, batch_df])

        # Save to CSV
        result_df.to_csv(output_file, index=False)

        # Log progress
        batch_time = time.time() - start_time
        st.write(f"Processed batch {i//batch_size + 1} of {len(texts)//batch_size + 1} in {batch_time:.2f} seconds")
        start_time = time.time()

    return result_df

# Streamlit interface
st.title("Enron Emails Emotion Analysis")

# Button to run the inference script
if st.button("Run Inference"):
    # Load the Enron dataset
    with st.spinner('Loading dataset...'):
        dataset = load_dataset("Hellisotherpeople/enron_emails_parsed")
        enron_data = pd.DataFrame(dataset['train'])

    # Apply emotion classification to the email content
    with st.spinner('Running inference...'):
        email_texts = enron_data['body'].tolist()
        classify_emotions_in_batches(email_texts, batch_size=64)

    st.success("Inference completed and results saved!")