Training run errors

[planetrocke]

Ubuntu 20.04, nvidia 470 running a 3090.

I get the following errors on training run, halting things at 0.00%

error: the following arguments are required: --user_data_path

AND

ERROR conda.cli.main_run:execute(49): .... etc..... (See above for error)

I am not sure if the logs are in the .db file or somewhere else, but any input on this would be helpful.

Thanks.

[dunky11]

The .db file is a sqlite database file.

The project is still pretty much a work in progress, it's good that you use ubuntu because I currently test on ubuntu only. I know where the bug originates from, its simply that a command line argument is not passed to the training script. Logs are usually written in /home/username/.config/voice-smith/data/training_runs/training_run_id/logs but if the script doesn't even start nothing is written there.

I will test everything on Ubuntu before the next release and come back to you when it's out, shouldn't take longer than 2-3 days.

[dunky11]

v0.1.4 works for me. You may need to start the app using the command line by typing "voice-smith". Otherwise it may currently not detect your conda installation.

[planetrocke]

I'm getting past some of this, now getting

Oops, an error occured, check logs for more info ... ERROR conda.cli.main_run:execute(49): conda run python training_run.py --run_id 2 --training_runs_path /home/vector/.config/voice-smith/data/training_runs --assets_path /usr/lib/voice-smith/resources/assets --db_path /home/vector/.config/voice-smith/db/voice_smith.db --models_path /home/vector/.config/voice-smith/data/models --datasets_path /home/vector/.config/voice-smith/data/datasets --user_data_path /home/vector/.config/voice-smith/data --environment_name __voice_smith__ failed. (See above for error)

When running this from CLI, I get:

Traceback (most recent call last): File "/usr/lib/voice-smith/resources/backend/voice_smith/training_run.py", line 12, in from voice_smith.acoustic_training import train_acoustic ModuleNotFoundError: No module named 'voice_smith'

ERROR conda.cli.main_run:execute(49): conda run python /usr/lib/voice-smith/resources/backend/voice_smith/training_run.py --run_id 2 --training_runs_path /home/vector/.config/voice-smith/data/training_runs --assets_path /usr/lib/voice-smith/resources/assets --db_path /home/vector/.config/voice-smith/db/voice_smith.db --models_path /home/vector/.config/voice-smith/data/models --datasets_path /home/vector/.config/voice-smith/data/datasets --user_data_path /home/vector/.config/voice-smith/data --environment_name __voice_smith__ failed. (See above for error)

[planetrocke]

I ran setup.py on the backend, and was able to get past the previous error, but now it says it can't find voice_smith.config.

[planetrocke]

The actual error in logs is:

Traceback (most recent call last): File "training_run.py", line 873, in continue_training_run( File "training_run.py", line 847, in continue_training_run runner.run( File "/usr/lib/voice-smith/resources/backend/voice_smith/utils/runs.py", line 44, in run finished = stage(**kwargs) File "training_run.py", line 327, in preprocessing_stage extract_data( File "/usr/lib/voice-smith/resources/backend/voice_smith/preprocessing/extract_data.py", line 319, in extract_data pitch_mean, pitch_std = calculate_pitch_stats( File "/usr/lib/voice-smith/resources/backend/voice_smith/preprocessing/extract_data.py", line 438, in calculate_pitch_stats pitch_mean, pitch_std = scaler.get_mean_std() File "/usr/lib/voice-smith/resources/backend/voice_smith/utils/tools.py", line 328, in get_mean_std return self.t_1_m / self.m, torch.sqrt(self.s_1_m / self.m) TypeError: unsupported operand type(s) for /: 'NoneType' and 'int'

[dunky11]

You should try 0.2.0, everything python runs in Docker now, which should fix most issues, hope this one works for you ...