app.py

import streamlit as st

from transformers import PerceiverTokenizer, PerceiverForMaskedLM
import transformers

@st.cache(allow_output_mutation=True, show_spinner=False)
def get_pipe():
    model = PerceiverForMaskedLM.from_pretrained("deepmind/language-perceiver")
    tokenizer = PerceiverTokenizer.from_pretrained("deepmind/language-perceiver")
    pipe = transformers.pipeline('text-classification', model=model, tokenizer=tokenizer,
                                 return_all_scores=True, truncation=True)
    return pipe


def sort_predictions(predictions):
    return sorted(predictions, key=lambda x: x['score'], reverse=True)


st.set_page_config(page_title="Emotion Prediction")
st.title("Emotion Prediction")
st.write("Type text into the text box and then press 'Predict' to get the predicted emotion.")

default_text = "I really love using HuggingFace Spaces!"

text = st.text_area('Enter text here:', value=default_text)
submit = st.button('Predict')

with st.spinner("Loading model..."):
    pipe = get_pipe()

if (submit and len(text.strip()) > 0) or len(text.strip()) > 0:

    prediction = pipe(text)[0]
    prediction = sort_predictions(prediction)

    fig, ax = plt.subplots()
    ax.bar(x=[i for i, _ in enumerate(prediction)],
           height=[p['score'] for p in prediction],
           tick_label=[p['label'] for p in prediction])
    ax.tick_params(rotation=90)
    ax.set_ylim(0, 1)

    st.header('Prediction:')
    st.pyplot(fig)

    prediction = dict([(p['label'], p['score']) for p in prediction])
    st.header('Raw values:')
    st.json(prediction)