app.py

import streamlit as st
from transformers import pipeline
from io import StringIO

unmasker = pipeline('fill-mask', model='dsfsi/zabantu-bantu-250m')
st.stop_rerun = True
st.set_page_config(layout="wide")

def fill_mask(sentences):
    results = {}
    warnings = []
    for language, sentence in sentences.items():
        if "<mask>" in sentence:
            masked_sentence = sentence.replace('<mask>', unmasker.tokenizer.mask_token)
            unmasked = unmasker(masked_sentence)
            results[language] = unmasked
        else:
            warnings.append(f"Warning: No <mask> token found in sentence: {sentence}")
    return results, warnings

def replace_mask(sentence, predicted_word):
    return sentence.replace("<mask>", f"**{predicted_word}**")

# Set up title and description
st.title("Fill Mask | Zabantu-XLM-Roberta")
st.markdown("Zabantu-XLMR refers to a fleet of models trained on South African Bantu languages...")

# Initialize session state
if 'warnings' not in st.session_state:
    st.session_state['warnings'] = []
if 'results' not in st.session_state:
    st.session_state['results'] = {}

# Define layout
col1, col2 = st.columns(2)

with col1:
    st.markdown("### Input :clipboard:")

    select_options = ['Choose option', 'Enter text input', 'Upload a file(csv/txt)']
    sample_sentence = {
        'tshivenda': "Rabulasi wa <mask> u khou bvelela nga u lima.",
        "tsonga": "N'wana wa xisati u <mask> ku tsaka."
    }

    option_selected = st.selectbox("Select an input option:", select_options, index=0)
    input_sentences = {}

    if option_selected == 'Enter text input':
        st.session_state['warnings'].clear()  # Clear warnings before new input
        language_options = ['Choose language', 'Zulu', 'Tshivenda', 'Sepedi', 'Tswana', 'Tsonga']
        
        for i in range(5):
            language = st.selectbox(f"Select language for input {i+1}:", language_options, key=f'language_{i}')
            sentence = st.text_input(f"Enter sentence for input {i+1} (with <mask>):", key=f'sentence_{i}')
            
            # Only process filled language and sentence pairs
            if language != 'Choose language' and sentence:
                input_sentences[language.lower()] = sentence

        if st.button("Submit"):
            if input_sentences:
                results, warnings = fill_mask(input_sentences)
                st.session_state['results'] = results
                st.session_state['warnings'] = warnings
            else:
                st.warning("Please fill at least one language and sentence.")

    elif option_selected == 'Upload a file(csv/txt)':
        uploaded_file = st.file_uploader("Choose a file (one sentence per line)")
        if uploaded_file:
            stringio = StringIO(uploaded_file.getvalue().decode("utf-8"))
            sentences = stringio.read().splitlines()

            for i, sentence in enumerate(sentences):
                # Here, you might need to define how to assign a language to each sentence
                # Assuming all sentences are in the same language for simplicity
                input_sentences[f'input_{i+1}'] = sentence

            if st.button("Submit"):
                results, warnings = fill_mask(input_sentences)
                st.session_state['results'] = results
                st.session_state['warnings'] = warnings

    st.markdown("### Example")
    st.code(sample_sentence)
    if st.button("Test Example"):
        result, warnings = fill_mask(sample_sentence)
        st.session_state['results'] = result
        st.session_state['warnings'] = warnings

with col2:
    st.markdown("### Output :bar_chart:")
    
    if st.session_state['results']:
        # Use st.fragment for dynamic content
        with st.container():
            for language, predictions in st.session_state['results'].items():
                if predictions:
                    top_prediction = predictions[0]
                    predicted_word = top_prediction['token_str']
                    score = top_prediction['score'] * 100

                    # Displaying the prediction with fragment
                    st.markdown(f"**{language.capitalize()} Prediction:** {predicted_word} ({score:.2f}%)")
                    st.markdown(f"<div class='bar'><div class='bar-fill' style='width:{score}%;'></div></div>", unsafe_allow_html=True)

    if st.session_state['warnings']:
        for warning in st.session_state['warnings']:
            st.warning(warning)

# CSS for styling
css = """
<style>
footer {display:none !important;}
.bar {width: 70%; background-color: #e6e6e6; border-radius: 12px; height: 5px;}
.bar-fill {background-color: #17152e; height: 100%; border-radius: 12px;}
.container {display: flex; justify-content: space-between; align-items: center; margin-bottom: 5px;}
</style>
"""
st.markdown(css, unsafe_allow_html=True)