Issues when train the Tacotron model on LJSpeech dataset.

[thomasvonl]

I have chang the last two cell of TTS_example.ipynb to
!python /content/TTS/TTS/bin/compute_statistics.py /content/TTS/recipes/ljspeech/tacotron2-DDC/tacotron2-DDC.json /content/TTS/scale_stats.npy --data_path /content/LJSpeech-1.1/wavs/' '!CUDA_VISIBLE_DEVICES="0" python /content/TTS/TTS/bin/train_tacotron.py --config_path /content/TTS/recipes/ljspeech/tacotron2-DDC/tacotron2-DDC.json \ --coqpit.output_path ./Results \ --coqpit.datasets.0.path /content/LJSpeech-1.1/ \ --coqpit.audio.stats_path /content/TTS/scale_stats.npy \

and I have get the following error message
Using CUDA: True Number of GPUs: 1 Mixed precision mode is ON fatal: not a git repository (or any of the parent directories): .git Git Hash: 0000000 Experiment folder: DEFINE THIS/ljspeech-ddc-June-20-2021_02+22PM-0000000 fatal: not a git repository (or any of the parent directories): .git Traceback (most recent call last): File "/content/TTS/TTS/bin/train_tacotron.py", line 737, in args, config, OUT_PATH, AUDIO_PATH, c_logger, tb_logger = init_training(sys.argv) File "/content/TTS/TTS/utils/arguments.py", line 182, in init_training config, OUT_PATH, AUDIO_PATH, c_logger, tb_logger = process_args(args) File "/content/TTS/TTS/utils/arguments.py", line 168, in process_args copy_model_files(config, experiment_path, new_fields) File "/content/TTS/TTS/utils/io.py", line 42, in copy_model_files copy_stats_path, File "/usr/lib/python3.7/shutil.py", line 120, in copyfile with open(src, 'rb') as fsrc: FileNotFoundError: [Errno 2] No such file or directory: 'scale_stats.npy'

The file scale_stats.npy exists and the file path is correct. Is the -- coqpit.addio.stats_path argument set the state file path? And I track the error to the "utils/io.py" file and test the value of "config.audio.stats_path", I assume it should be "/content/TTS/scale_stats.npy " but it is "scale_stats.npy". And I run the ipynb file on the Colab, so it clone the latest repo.
@Sadam1195

[Sadam1195]

Have you tried running it with

'!CUDA_VISIBLE_DEVICES="0" python /content/TTS/TTS/bin/train_tacotron.py  \
--config_path /content/TTS/recipes/ljspeech/tacotron2-DDC/tacotron2-DDC.json \ 
--coqpit.output_path ./Results \ 
--coqpit.datasets.0.path /content/LJSpeech-1.1/ \ 

try it without adding \ --coqpit.audio.stats_path /content/TTS/scale_stats.npy \ to the command

[thomasvonl]

the config file is /content/TTS/recipes/ljspeech/tacotron2-DDC/tacotron2-DDC.json right? because I have found nothing about stats_path there

[Sadam1195]

Yes there is, see the line 32.

[ragireddydpriya]

Hello @Sadam1195,

Iam also facing similar issue and i've followed up your conversations and came till this point.
After modifying line 32 in /content/TTS/recipes/ljspeech/tacotron2-DDC/tacotron2-DDC.json, I am facing this issue.

command:

!CUDA_VISIBLE_DEVICES="0" python /content/TTS/TTS/bin/train_tacotron.py --config_path /content/TTS/recipes/ljspeech/tacotron2-DDC/tacotron2-DDC.json \ --coqpit.output_path ./Results \ --coqpit.datasets.0.path ./content/LJSpeech-1.1/

output of cell:

Using CUDA: True
Number of GPUs: 1
Mixed precision mode is ON
Git Hash: d245b5d
Experiment folder: DEFINE THIS/ljspeech-ddc-July-01-2021_02+44PM-d245b5d4
Setting up Audio Processor...
| > sample_rate:22050
| > resample:False
| > num_mels:80
| > min_level_db:-100
| > frame_shift_ms:None
| > frame_length_ms:None
| > ref_level_db:20
| > fft_size:1024
| > power:1.5
| > preemphasis:0.0
| > griffin_lim_iters:60
| > signal_norm:True
| > symmetric_norm:True
| > mel_fmin:50.0
| > mel_fmax:7600.0
| > spec_gain:1.0
| > stft_pad_mode:reflect
| > max_norm:4.0
| > clip_norm:True
| > do_trim_silence:True
| > trim_db:60
| > do_sound_norm:False
| > stats_path:None
| > base:10
| > hop_length:256
| > win_length:1024
! Run is removed from DEFINE THIS/ljspeech-ddc-July-01-2021_02+44PM-d245b5d4
Traceback (most recent call last):
File "/content/TTS/TTS/bin/train_tacotron.py", line 739, in
main(args)
File "/content/TTS/TTS/bin/train_tacotron.py", line 603, in main
meta_data_train, meta_data_eval = load_meta_data(config.datasets)
File "/content/TTS/TTS/tts/datasets/preprocess.py", line 51, in load_meta_data
meta_data_train = preprocessor(root_path, meta_file_train)
File "/content/TTS/TTS/tts/datasets/preprocess.py", line 183, in ljspeech
with open(txt_file, "r", encoding="utf-8") as ttf:
FileNotFoundError: [Errno 2] No such file or directory: 'DEFINE THIS/metadata.csv'

Please do needful to proceed further

Thank you

[erogol]

👁️ FileNotFoundError: [Errno 2] No such file or directory: 'DEFINE THIS/metadata.csv'

[ragireddydpriya]

Hello erogol,

I changed the path of metadata.csv in /content/TTS/recipes/ljspeech/tacotron2-DDC/tacotron2-DDC.json.After modifying fields {audio.stats_path, output_path, phoneme_cache_path, and datasets.path} in .json to specified path , Iam getting below error.

Using CUDA: True
Number of GPUs: 1
Traceback (most recent call last):
File "/content/TTS/TTS/bin/train_tacotron.py", line 737, in
args, config, OUT_PATH, AUDIO_PATH, c_logger, tb_logger = init_training(sys.argv)
File "/content/TTS/TTS/utils/arguments.py", line 182, in init_training
config, OUT_PATH, AUDIO_PATH, c_logger, tb_logger = process_args(args)
File "/content/TTS/TTS/utils/arguments.py", line 145, in process_args
config = load_config(args.config_path)
File "/content/TTS/TTS/config/init.py", line 62, in load_config
with open(config_path, "r", encoding="utf-8") as f:
FileNotFoundError: [Errno 2] No such file or directory: './content/TTS/recipes/ljspeech/tacotron2-DDC/tacotron2-DDC.json'

Please kindly suggest ,do i need to modify any fields in config file inorder to train the model.[attaching modified tacotron2-DDC.json
tacotron2-DDC.zip
]