RUN DOCKER

nvidia-docker run -ti -p 6006:6006 \ -v ~/labs/wavenet_vocoder:/wavenet_vocoder \ -v ~/labs/liepa_dataset/MII_LIEPA_SYN_V1:/MII_LIEPA_SYN_V1\ wavenet_vocoder_liepa:gpu bash

PREPROCESS

python preprocess.py liepa /MII_LIEPA_SYN_V1 /MII_LIEPA_SYN_V1/proc/ –preset=presets/liepa_mixture.json

Tensorboard

pkill -f “tensorboard”

tensorboard –logdir=./log &

TRAIN

python train.py –data-root=/MII_LIEPA_SYN_V1/proc_4/ –preset=presets/liepa_mixture.json

RESUME

python train.py –data-root=/MII_LIEPA_SYN_V1/proc_4/ –preset=presets/liepa_mixture.json \ –checkpoint-dir=/wavenet_vocoder/checkpoints_fine \ –log-event-path=./log/run-test2018-09- –checkpoint=/wavenet_vocoder/checkpoints/checkpoint_step000000835.pth

FINETUNE

python train.py –data-root=/MII_LIEPA_SYN_V1/proc_4/ \ –preset=presets/liepa_mixture.json \ –checkpoint-dir=/wavenet_vocoder/checkpoints_fine \ –log-event-path=./log/run-test2018-08-31_09:00:00.000000 \ –restore-parts=/wavenet_vocoder/checkpoints/checkpoint_step00

EVALUATE

python evaluate.py checkpoints_fine/checkpoint_step000210000_ema.pth \ ./eval/checkpoint_step000210000_ema/speaker_0 \ –data-root=/MII_LIEPA_SYN_V1/proc_4/ \ –preset=./presets/liepa_mixture.json \ –output-html –num-utterances=3 –speaker-id=0

python evaluate.py checkpoints_fine/checkpoint_step000210000_ema.pth \ ./eval/checkpoint_step000210000_ema/speaker_1 \ –data-root=/MII_LIEPA_SYN_V1/proc_4/ \ –preset=./presets/liepa_mixture.json \ –output-html –num-utterances=3 –speaker-id=1

python evaluate.py checkpoints_fine/checkpoint_step000210000_ema.pth \ ./eval/checkpoint_step000210000_ema/speaker_2 \ –data-root=/MII_LIEPA_SYN_V1/proc_4/ \ –preset=./presets/liepa_mixture.json \ –output-html –num-utterances=3 –speaker-id=2

python evaluate.py checkpoints_fine/checkpoint_step000210000_ema.pth \ ./eval/checkpoint_step000210000_ema/speaker_3 \ –data-root=/MII_LIEPA_SYN_V1/proc_4/ \ –preset=./presets/liepa_mixture.json \ –output-html –num-utterances=3 –speaker-id=3