Hi,
I want to train the model in French, i use the data set from website 'common_voice'.
I wrote commoncoive_fr.py like this:
from concurrent.futures import ProcessPoolExecutor
from functools import partial
import numpy as np
import os
from util import audio
def build_from_path(in_dir, out_dir, num_workers=1, tqdm=lambda x: x):
executor = ProcessPoolExecutor(max_workers=num_workers)
futures = []
index = 1
with open(os.path.join(in_dir, 'train.tsv'), encoding='utf-8') as f:
for line in f:
parts = line.strip().split('\t')
wav_path = os.path.join(in_dir,parts[1])
text = parts[2]
futures.append(executor.submit(partial(_process_utterance, out_dir, index, wav_path, text)))
index += 1
return [future.result() for future in tqdm(futures)]
def _process_utterance(out_dir, index, wav_path, text):
'''Preprocesses a single utterance audio/text pair.
This writes the mel and linear scale spectrograms to disk and returns a tuple to write
to the train.txt file.
Args:
out_dir: The directory to write the spectrograms into
index: The numeric index to use in the spectrogram filenames.
wav_path: Path to the audio file containing the speech input
text: The text spoken in the input audio file
Returns:
A (spectrogram_filename, mel_filename, n_frames, text) tuple to write to train.txt
'''
Load the audio to a numpy array:
wav = audio.load_wav(wav_path)
Compute the linear-scale spectrogram from the wav:
And I modified the preprocess.py like this:
import argparse
import os
from multiprocessing import cpu_count
from tqdm import tqdm
from datasets import amy, blizzard, ljspeech, kusal, mailabs,commonvoice_fr
from datasets import mrs
from hparams import hparams, hparams_debug_string
import sys
def write_metadata(metadata, out_dir):
with open(os.path.join(out_dir, 'train.txt'), 'w', encoding='utf-8') as f:
for m in metadata:
f.write('|'.join([str(x) for x in m]) + '\n')
frames = sum([m[2] for m in metadata])
hours = frames hparams.frame_shift_ms / (3600 1000)
print('Wrote %d utterances, %d frames (%.2f hours)' %
(len(metadata), frames, hours))
print('Max input length: %d' % max(len(m[3]) for m in metadata))
print('Max output length: %d' % max(m[2] for m in metadata))
with open("metadata.txt", 'w') as f:
f.write(
'''
Wrote {} utterances, {} frames, {} hours\n
Max input lengh: {} \n
Max output length: {} \n
'''.format(
len(metadata), frames, hours,
max(len(m[3]) for m in metadata), max(m[2] for m in metadata)
)
)
def main():
parser = argparse.ArgumentParser()
parser.add_argument('--base_dir', default=os.path.expanduser('~/tacotron'))
parser.add_argument('--mrs_dir', required=False)
parser.add_argument('--mrs_username', required=False)
parser.add_argument('--output', default='training')
parser.add_argument(
'--dataset', required=True, choices=['amy', 'blizzard', 'ljspeech',
'kusal', 'mailabs','mrs','commonvoice']
)
parser.add_argument('--mailabs_books_dir',
help='absolute directory to the books for the mlailabs')
parser.add_argument(
'--books',
help='comma-seperated and no space name of books i.e hunter_space,pink_fairy_book,etc.',
)
parser.add_argument('--num_workers', type=int, default=cpu_count())
args = parser.parse_args()
if args.dataset == 'mailabs' and args.books is None:
parser.error("--books required if mailabs is chosen for dataset.")
if args.dataset == 'mailabs' and args.mailabs_books_dir is None:
parser.error(
"--mailabs_books_dir required if mailabs is chosen for dataset.")
But when I preprocces the data by using the commande:
python3 preprocess.py --dataset commonvoice
I got this erros:
Traceback (most recent call last):
File "/usr/lib/python3.5/concurrent/futures/process.py", line 175, in _process_worker
r = call_item.fn(*call_item.args, **call_item.kwargs)
File "/root/mimic2/datasets/commonvoice_fr.py", line 64, in _process_utterance
spectrogram_filename = 'commonvoice_fr-spec.npy' % index
TypeError: not all arguments converted during string formatting
"""
Could you please help me to solve this problem?
Thanks
Hi, I want to train the model in French, i use the data set from website 'common_voice'. I wrote commoncoive_fr.py like this: from concurrent.futures import ProcessPoolExecutor from functools import partial import numpy as np import os from util import audio
def build_from_path(in_dir, out_dir, num_workers=1, tqdm=lambda x: x): executor = ProcessPoolExecutor(max_workers=num_workers) futures = [] index = 1 with open(os.path.join(in_dir, 'train.tsv'), encoding='utf-8') as f: for line in f: parts = line.strip().split('\t') wav_path = os.path.join(in_dir,parts[1]) text = parts[2] futures.append(executor.submit(partial(_process_utterance, out_dir, index, wav_path, text))) index += 1 return [future.result() for future in tqdm(futures)]
def _process_utterance(out_dir, index, wav_path, text): '''Preprocesses a single utterance audio/text pair.
This writes the mel and linear scale spectrograms to disk and returns a tuple to write to the train.txt file.
Args: out_dir: The directory to write the spectrograms into index: The numeric index to use in the spectrogram filenames. wav_path: Path to the audio file containing the speech input text: The text spoken in the input audio file
Returns: A (spectrogram_filename, mel_filename, n_frames, text) tuple to write to train.txt '''
Load the audio to a numpy array:
wav = audio.load_wav(wav_path)
Compute the linear-scale spectrogram from the wav:
spectrogram = audio.spectrogram(wav).astype(np.float32) n_frames = spectrogram.shape[1]
Compute a mel-scale spectrogram from the wav:
mel_spectrogram = audio.melspectrogram(wav).astype(np.float32)
Write the spectrograms to disk:
spectrogram_filename = 'commonvoice_fr-spec.npy' % index mel_filename = 'commonvoice_fr-mel.npy' % index np.save(os.path.join(out_dir, spectrogram_filename), spectrogram.T, allow_pickle=False) np.save(os.path.join(out_dir, mel_filename), mel_spectrogram.T, allow_pickle=False)
Return a tuple describing this training example:
return (spectrogram_filename, mel_filename, n_frames, text)
And I modified the preprocess.py like this: import argparse import os from multiprocessing import cpu_count from tqdm import tqdm from datasets import amy, blizzard, ljspeech, kusal, mailabs,commonvoice_fr from datasets import mrs from hparams import hparams, hparams_debug_string import sys
def preprocess_blizzard(args): in_dir = os.path.join(args.base_dir, 'Blizzard2012') out_dir = os.path.join(args.base_dir, args.output) os.makedirs(out_dir, exist_ok=True) metadata = blizzard.build_from_path( in_dir, out_dir, args.num_workers, tqdm=tqdm) write_metadata(metadata, out_dir)
def preprocess_ljspeech(args): in_dir = os.path.join(args.base_dir, 'LJSpeech-1.1') out_dir = os.path.join(args.base_dir, args.output) os.makedirs(out_dir, exist_ok=True) metadata = ljspeech.build_from_path( in_dir, out_dir, args.num_workers, tqdm=tqdm) write_metadata(metadata, out_dir)
def preprocess_mrs(args): in_dir = args.mrs_dir out_dir = os.path.join(args.base_dir, args.output) username = args.mrs_username os.makedirs(out_dir, exist_ok=True) metadata = mrs.build_from_path( in_dir, out_dir, username, args.num_workers, tqdm=tqdm) write_metadata(metadata, out_dir)
def preprocess_amy(args): in_dir = os.path.join(args.base_dir, 'amy') out_dir = os.path.join(args.base_dir, args.output) os.makedirs(out_dir, exist_ok=True) metadata = amy.build_from_path(in_dir, out_dir, args.num_workers, tqdm=tqdm) write_metadata(metadata, out_dir)
def preprocess_kusal(args): in_dir = os.path.join(args.base_dir, 'kusal') out_dir = os.path.join(args.base_dir, args.output) os.makedirs(out_dir, exist_ok=True) metadata = kusal.build_from_path( in_dir, out_dir, args.num_workers, tqdm=tqdm) write_metadata(metadata, out_dir)
def preprocess_mailabs(args): in_dir = os.path.join(args.mailabs_books_dir) out_dir = os.path.join(args.base_dir, args.output) os.makedirs(out_dir, exist_ok=True) books = args.books metadata = mailabs.build_from_path( in_dir, out_dir, books, args.num_workers, tqdm) write_metadata(metadata, out_dir)
def preprocess_commonvoice(args): in_dir = os.path.join(args.base_dir,'clips') out_dir = os.path.join(args.base_dir,args.output) os.makedirs(out_dir,exist_ok=True) metdata = commonvoice_fr.build_from_path(in_dir,out_dir, args.num_workers,tqdm=tqdm) write_metadata(metadata,out_dir)
def write_metadata(metadata, out_dir): with open(os.path.join(out_dir, 'train.txt'), 'w', encoding='utf-8') as f: for m in metadata: f.write('|'.join([str(x) for x in m]) + '\n') frames = sum([m[2] for m in metadata]) hours = frames hparams.frame_shift_ms / (3600 1000) print('Wrote %d utterances, %d frames (%.2f hours)' % (len(metadata), frames, hours)) print('Max input length: %d' % max(len(m[3]) for m in metadata)) print('Max output length: %d' % max(m[2] for m in metadata)) with open("metadata.txt", 'w') as f: f.write( ''' Wrote {} utterances, {} frames, {} hours\n Max input lengh: {} \n Max output length: {} \n '''.format( len(metadata), frames, hours, max(len(m[3]) for m in metadata), max(m[2] for m in metadata) ) )
def main(): parser = argparse.ArgumentParser() parser.add_argument('--base_dir', default=os.path.expanduser('~/tacotron')) parser.add_argument('--mrs_dir', required=False) parser.add_argument('--mrs_username', required=False) parser.add_argument('--output', default='training') parser.add_argument( '--dataset', required=True, choices=['amy', 'blizzard', 'ljspeech', 'kusal', 'mailabs','mrs','commonvoice'] ) parser.add_argument('--mailabs_books_dir', help='absolute directory to the books for the mlailabs') parser.add_argument( '--books', help='comma-seperated and no space name of books i.e hunter_space,pink_fairy_book,etc.', ) parser.add_argument('--num_workers', type=int, default=cpu_count()) args = parser.parse_args()
if args.dataset == 'mailabs' and args.books is None: parser.error("--books required if mailabs is chosen for dataset.")
if args.dataset == 'mailabs' and args.mailabs_books_dir is None: parser.error( "--mailabs_books_dir required if mailabs is chosen for dataset.")
print(hparams_debug_string())
if args.dataset == 'amy': preprocess_amy(args) elif args.dataset == 'blizzard': preprocess_blizzard(args) elif args.dataset == 'ljspeech': preprocess_ljspeech(args) elif args.dataset == 'kusal': preprocess_kusal(args) elif args.dataset == 'mailabs': preprocess_mailabs(args) elif args.dataset == 'mrs': preprocess_mrs(args) elif args.dataset == 'commonvoice': preprocess_commonvoice(args)
if name == "main": main()
But when I preprocces the data by using the commande: python3 preprocess.py --dataset commonvoice I got this erros: Traceback (most recent call last): File "/usr/lib/python3.5/concurrent/futures/process.py", line 175, in _process_worker r = call_item.fn(*call_item.args, **call_item.kwargs) File "/root/mimic2/datasets/commonvoice_fr.py", line 64, in _process_utterance spectrogram_filename = 'commonvoice_fr-spec.npy' % index TypeError: not all arguments converted during string formatting """
Could you please help me to solve this problem? Thanks