Main.py

import torch

# To speed-up training process
torch.autograd.set_detect_anomaly(False)
torch.autograd.profiler.profile(False)
torch.autograd.profiler.emit_nvtx(False)

import warnings
import pickle

from Transformer import Transformer
import librosa

import os.path as path
import json
from tqdm import tqdm
import time

import numpy as np

warnings.filterwarnings("ignore")

with open("config.json") as json_data_file:
    data = json.load(json_data_file)

lr = data['learn_rate']
epochs = data['epochs']
batch_size = data['batch_size']
Training = data['Training']
Testing = data['Testing']
main_path = data['MainPath']

device = data['Device']
diag_attn = data['DiagAttn']

BestModelPath = 'Best_GlobalModel_500_0_0.pt'

def pre_process_mfcc(mfcc):
    
    mfcc = mfcc.T

    mean_G = np.mean(mfcc, axis=0)
    std_G = np.std(mfcc, axis=0)
    mfcc = 0.5*(mfcc-mean_G)/std_G
    
    return mfcc

def wav2art(wav):
    rate = 16000
    mfcc = librosa.feature.mfcc(wav, 16000, n_mfcc=13, hop_length=int(0.010*rate), n_fft=int(0.020*rate))
    mfcc = pre_process_mfcc(mfcc)

    mfcc = torch.tensor([mfcc]).float()

    test_model = torch.load(BestModelPath, map_location=torch.device('cpu')).float()
    test_model.eval()

    p = test_model(mfcc, 0, 0, 0)
    p = p[0].detach().numpy()

    return p