app.py

import streamlit as st
from datasets import load_dataset

dataset = load_dataset('rwcuffney/pick_a_card_test', batch_size=32, shuffle=True)

from transformers import AutoModelForSequenceClassification

model = AutoModelForSequenceClassification.from_pretrained('rwcuffney/autotrain-pick_a_card-3726099224')

from transformers import AutoTokenizer

tokenizer = AutoTokenizer.from_pretrained('rwcuffney/autotrain-pick_a_card-3726099224')

def preprocess_text(text):
    encoded = tokenizer(text, padding='max_length', truncation=True, max_length=128, return_tensors='pt')
    return encoded

    device = 'cuda' if torch.cuda.is_available() else 'cpu'
model.to(device)
model.eval()

for batch in dataset:
    # Preprocess the text
    text = batch['text']
    inputs = preprocess_text(text)
    inputs = inputs.to(device)

    # Make predictions
    with torch.no_grad():
        outputs = model(**inputs)
        predicted_classes = torch.argmax(outputs.logits, dim=-1)

    # Print the predicted class labels
    predicted_labels = [dataset.features['label'].names[i] for i in predicted_classes]
    st.write(predicted_labels)