search

onejiin / CycleGAN-VC2

CycleGAN-VC2: Improved CycleGAN-based Non-parallel Voice Conversion
MIT License
41 stars 7 forks source link

ERROR For preprocessing #1

Closed yuguochencuc closed 5 years ago

yuguochencuc commented 5 years ago

excuse me.could you please tell me why there is an error when i preprocess vcc2016 data. when i run the code, "python train.py --train_A_dir ./data/vcc2016_training/SF1 --train_B_dir ./data/vcc2016_training/TM1 --model_dir ./model/sf1_tm1 --model_name sf1_tm1.ckpt --random_seed 0 --gen_model CycleGAN-VC2 --validation_A_dir ./data/evaluation_all/SF1 --validation_B_dir ./data/evaluation_all/TM1 --output_dir ./validation_output --tensorboard_log_dir ./log --MCEPs_dim 24".i don't know why there is a traceback.Traceback (most recent call last): File "train.py", line 265, in hdf5_path=[hdf5A_path, hdf5B_path], lambda_list=[lambda_cycle, lambda_identity]) File "train.py", line 57, in train log_f0s_mean_A, log_f0s_std_A = logf0_statistics(f0s_A) File "/home/yuguochen/CycleGAN-VC2/preprocess.py", line 162, in logf0_statistics log_f0s_concatenated = np.ma.log(np.concatenate(f0s)) ValueError: need at least one array to concatenate i want know what is the "hdf5A_path"? do i need put the data into this path?

yuguochencuc commented 5 years ago

thanks a lot for answering my question.

onejiin commented 5 years ago

Hi, This is an error to HDF5 directory existence. If you want to train a short utterance like vcc2016, you do not need to write data db as HDF5. Because I modified code, you can run hdf5A_path=None (Default is None) Please update code by using git clone. Best,