# train model # if you wish to resume from an exists model, uncomment --checkpoint_path and --fastspeech_step
export CUDA_VISIBLE_DEVICES=0
python -u train.py \ --batch_size=32 \ --epochs=10000 \ --lr=0.001 \ --save_step=500 \ --use_gpu=1 \ --use_data_parallel=0 \ --data_path='../../dataset/LJSpeech-1.1' \ --transtts_path='../transformer_tts/checkpoint' \ --transformer_step=160000 \ --save_path='./checkpoint' \ --log_dir='./log' \ --config_path='config/fastspeech.yaml' \ #--checkpoint_path='./checkpoint' \ #--fastspeech_step=97000 \ if [ $? -ne 0 ]; then echo "Failed in training!" exit 1 fi exit 0
