app.py

import os 
import sys 
os.system("pip install transformers==4.27.0")
os.system("pip install numpy==1.23")
from transformers import pipeline, WhisperModel, WhisperTokenizer, WhisperFeatureExtractor, AutoFeatureExtractor, AutoProcessor, WhisperConfig, WhisperProcessor, WhisperForConditionalGeneration
os.system("pip install jiwer")
from jiwer import wer
os.system("pip install datasets[audio]")
from evaluate import evaluator, load
from transformers import AutoModelForSequenceClassification, pipeline, BertTokenizer, AutoTokenizer, GPT2Model
from datasets import load_dataset, Audio, disable_caching, set_caching_enabled
import gradio as gr
import torch
from datasets import load_dataset
from transformers import WhisperForConditionalGeneration, WhisperProcessor


processor = WhisperProcessor.from_pretrained("mskov/whisper-small-esc50")
model = WhisperForConditionalGeneration.from_pretrained("mskov/whisper-small-esc50").to("cuda")

def map_to_pred(batch):
    audio = batch["audio"]
    input_features = processor(audio["array"], sampling_rate=16000, return_tensors="pt").input_features
    batch["reference"] = processor.tokenizer._normalize(batch['category'])

    with torch.no_grad():
        predicted_ids = model.generate(input_features.to("cuda"))[0]
    transcription = processor.decode(predicted_ids)
    batch["prediction"] = processor.tokenizer._normalize(transcription)
    print(batch["prediction"])
    return batch

result = librispeech_test_clean.map(map_to_pred)

wer = load("wer")
print(100 * wer.compute(references=result["reference"], predictions=result["prediction"]))


    
def transcribe(audio):
    text = pipe(audio)["text"]
    return text, test

iface = gr.Interface(
    fn=transcribe, 
    inputs=gr.Audio(source="microphone", type="filepath"), 
    outputs="text",
    title="Whisper Small ESC50 Test",
)

iface.launch()


'''
print("check check")
print(inputs)
input_features = inputs.input_features
decoder_input_ids = torch.tensor([[1, 1]]) * model.config.decoder_start_token_id
last_hidden_state = model(input_features, decoder_input_ids=decoder_input_ids).last_hidden_state
list(last_hidden_state.shape)
print(list(last_hidden_state.shape))
'''