Giters

RussellSB / tt-vae-gan

Timbre transfer with variational autoencoding and cycle-consistent adversarial networks. Able to transfer the timbre of an audio source to that of another.

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Two small issues in the script for training the WaveNet vocoder (run.sh)

rohitgupta3 opened this issue · comments

commented

Hi, two fairly minor questions.

  1. Within the WaveNet vocoder training section, for the preprocessing step (step 2.2), it gives an example
    spk="[name]_[id]" ./run.sh --stage 1 --stop-stage 1

Do I need to also pass hparams (à la step 2.3)? I seem to get an unbound variable if not, and I assume this is the reason.

  1. The first time I run the training for the WaveNet vocoder, it appears that I get a FileNotFoundError along the lines of:
    FileNotFoundError: [Errno 2] No such file or directory: 'exp/flickr_1_train_no_dev_flickr/checkpoint_latest.pth'

Is this because wavenet_vocoder/egs/gaussian/run.sh passes the --checkpoint=${expdir}/checkpoint_latest.pth argument to train.py even though (if this is a fresh model run) there wouldn't be any latest checkpoint saved? If I edit out that arg from that line, the training at least starts.

commented

The parameter inside run.sh is passed how you describe it.

The tutorial assumes you follow it using pre-trained models (which is not compulsory of course, just as a showcase).
It refers to checkpoint latest as that is the name of the pretrained model weights file I uploaded.

It sounds correct that removing line 78 is the way to go for training things yourself. I'll update the readme with clarifications

commented

Hi, can you please advise on how to train the Wavenet vocoder? I'm struggling with step 2.2.
In step 2.1, I used --tag test, which created data/test_1/ and data/test_2/ for the two speakers, containing their respective data, so in step 2.2 I run:
spk="test_1" ./run.sh --stage 1 --stop-stage 1
And this results in:
./run.sh: line 38: hparams: unbound variable

I have then tried to run
spk="test_1" hparams=conf/urmp.json ./run.sh --stage 1 --stop-stage 1
But this results in

stage 1: Feature Generation
./run.sh: line 55: python: command not found

The path to VOC_DIR seems correct, as running only python $VOC_DIR/preprocess.py is possible, but I'm struggling with understanding the rest. And the same issue will arise when running stage 2 later.