app.py

import evaluate
from evaluate.utils import launch_gradio_widget
import gradio as gr
from transformers import AutoModelForSequenceClassification, pipeline, RobertaForSequenceClassification, RobertaTokenizer, AutoTokenizer
import tempfile

tmp = tempfile.NamedTemporaryFile()

 
# Define the list of available models
available_models = {
    "mskov/roberta-base-toxicity": "Roberta Finetuned Model"
}
 
 
# Create a Gradio interface with audio file and text inputs
def classify_toxicity(audio_file, text_input, selected_model):
    # Transcribe the audio file using Whisper ASR
    if audio_file != None:
        whisper_module = evaluate.load("whisper")
        transcription_results = whisper_module.compute(uploaded=audio_file)
     
        # Extract the transcribed text
        transcribed_text = transcription_results["transcription"]
    else:
        transcribed_text = text_input
 
    # Load the selected toxicity classification model
    toxicity_module = evaluate.load("toxicity", selected_model)
    #toxicity_module = evaluate.load("toxicity", 'DaNLP/da-electra-hatespeech-detection', module_type="measurement")

    toxicity_results = toxicity_module.compute(predictions=[transcribed_text])
 
    toxicity_score = toxicity_results["toxicity"][0]
    print(toxicity_score)
    return toxicity_score, transcribed_text
    # return f"Toxicity Score ({available_models[selected_model]}): {toxicity_score:.4f}"
 
with gr.Blocks() as iface:
    with gr.Column():
        aud_input = gr.Audio(source="upload", type="filepath", label="Upload Audio File"),
    with gr.Row():
        text = gr.Textbox(label="Enter Text", placeholder="Enter text here..."),
        submit = gr.Button(label="Submit", type="submit")
        radio = gr.Radio(available_models, type="value", label="Select Model")
        out_text = gr.Textbox()
    submit.click(fn=classify_toxicity, inputs=[aud_input, text, radio], outputs=out_text)

iface.launch()