app.py

import gradio as gr

import torch
import librosa
from transformers import AutoTokenizer, AutoModelForSequenceClassification, TextClassificationPipeline, AutoModelForTokenClassification, TokenClassificationPipeline, Wav2Vec2ForCTC, Wav2Vec2Processor, Wav2Vec2ProcessorWithLM

# ASR
model_name = "jonatasgrosman/wav2vec2-large-xlsr-53-english"
processor_asr = Wav2Vec2Processor.from_pretrained(model_name)
model_asr = Wav2Vec2ForCTC.from_pretrained(model_name)

# Classifier Intent
model_name = 'qanastek/XLMRoberta-Alexa-Intents-Classification'
tokenizer_intent = AutoTokenizer.from_pretrained(model_name)
model_intent = AutoModelForSequenceClassification.from_pretrained(model_name)
classifier_intent = TextClassificationPipeline(model=model_intent, tokenizer=tokenizer_intent, device=0)

# Classifier Language
model_name = 'qanastek/51-languages-classifier'
tokenizer_langs = AutoTokenizer.from_pretrained(model_name)
model_langs = AutoModelForSequenceClassification.from_pretrained(model_name)
classifier_language = TextClassificationPipeline(model=model_langs, tokenizer=tokenizer_langs, device=0)

# NER Extractor
model_name = 'qanastek/XLMRoberta-Alexa-Intents-NER-NLU'
tokenizer_ner = AutoTokenizer.from_pretrained(model_name)
model_ner = AutoModelForTokenClassification.from_pretrained(model_name)
predict_ner = TokenClassificationPipeline(model=model_ner, tokenizer=tokenizer_ner, device=0)

def greet(name):
    return "Hello " + name + "!!"

iface = gr.Interface(fn=greet, inputs="text", outputs="text")
iface.launch()