sh-lee-prml
commented
2 years ago I will update inference code soon.
import os
import torch
import argparse
from glob import glob
import tqdm
import numpy as np
from scipy.io.wavfile import write
import utils
from mel_processing import mel_spectrogram_torch
from models_bigvgan import Generator
h = None
device = None
def get_mel(x):
return mel_spectrogram_torch(x, h.n_fft, h.num_mels, h.sampling_rate, h.hop_size, h.win_size, h.fmin, h.fmax)
def load_checkpoint(filepath, device):
assert os.path.isfile(filepath)
print("Loading '{}'".format(filepath))
checkpoint_dict = torch.load(filepath, map_location=device)
print("Complete.")
return checkpoint_dict
def scan_checkpoint(cp_dir, prefix):
pattern = os.path.join(cp_dir, prefix + '*')
cp_list = glob.glob(pattern)
if len(cp_list) == 0:
return ''
return sorted(cp_list)[-1]
def inference(a):
os.makedirs(a.output_dir, exist_ok=True)
generator = Generator(h).to(device)
state_dict_g = load_checkpoint(a.checkpoint_file, device)
generator.load_state_dict(state_dict_g['generator'])
generator.eval()
generator.remove_weight_norm()
npz_path = glob(os.path.join(a.input_dir, os.path.join("test", "*.npz")))
print("data len: ", len(npz_path))
i = 0
for path in tqdm.tqdm(npz_path, desc="synthesizing each utterance"):
files = np.load(path)
file_name = os.path.splitext(os.path.basename(path))[0]
with torch.no_grad():
audio = torch.FloatTensor(files['audio'])
audio = audio.to(device)
audio = audio / 32768
mel = get_mel(audio.unsqueeze(0))
audio= generator(mel)
audio = audio.squeeze()
audio = audio / (torch.abs(audio).max()) * 0.999 * 32768.0
audio = audio.cpu().numpy().astype('int16')
file_name = "generated_{}.wav".format(file_name)
output_file = os.path.join(a.output_dir, file_name)
write(output_file, 22050, audio)
def main():
print('Initializing Inference Process..')
parser = argparse.ArgumentParser()
parser.add_argument('--input_dir', default='')
parser.add_argument('--output_dir', default='inference/')
parser.add_argument('--checkpoint_file', default='logs/checkpoint')
a = parser.parse_args()
config_file = os.path.join(os.path.split(a.checkpoint_file)[0], 'config.json')
global h
h = utils.get_hparams_from_file(config_file)
torch.manual_seed(1234)
global device
if torch.cuda.is_available():
torch.cuda.manual_seed(1234)
device = torch.device('cuda')
else:
device = torch.device('cpu')
inference(a)
if __name__ == '__main__':
main()
skol101
Could you provide inference example?