Riha1992

I did that for Farsi, which also can be counted as a low resource language using 25 hours of speech from ebooks and the results suggest the need for a...

You can simply run the following command: python train.py --restore_step = xxxx

Have you tried Tensorflow 1.5?