deepvoice3_pytorch
deepvoice3_pytorch copied to clipboard
Error in `python3.5': corrupted size vs. prev_size when running python3.5 synthesis.py
I was able to run the "python3.5 synthesis.py ....". But it generated an error at the end each time.
.............................. Finished! Check out ./output for generated audio samples.
*** Error in `python3.5': corrupted size vs. prev_size: 0x0000000001682bd0 *** Aborted (core dumped)
By commenting out the following code section, the error went way. Not sure if it is related to the setting "backend : Agg" I set for matplot.
dst_alignment_path = join(
dst_dir, "{}_{}{}_alignment.png".format(idx, checkpoint_name,
file_name_suffix))
plot_alignment(alignment.T, dst_alignment_path,
info="{}, {}".format(hparams.builder, basename(checkpoint_path)))
There is another bug when running on Python 3.5 ubuntu 14.04. Sometime the 1st run causes segment fault. It usually runs fine 2nd time. See below for example run: #1st run failed python3.5 synthesis.py --hparams="builder=deepvoice3,preset=deepvoice3_ljspeech" models/20171213_deepvoice3_checkpoint_step000210000.pth ./text_list.txt ./output
Segmentation fault (core dumped)
#2nd run, succeeded python3.5 synthesis.py --hparams="builder=deepvoice3,preset=deepvoice3_ljspeech" models/20171213_deepvoice3_checkpoint_step000210000.pth ./text_list.txt ./output
Command line args: {'--checkpoint-postnet': None, '--checkpoint-seq2seq': None, '--file-name-suffix': '', '--help': False,
Occasionally also got this exception at the end of a run:
Exception ignored in: <function WeakValueDictionary.init.