5 Commits

Author SHA1 Message Date
pkufool
3060c5a556 Add grad_clip and weight-decay, small fix of dataloader and masking 2021-08-19 17:34:28 +08:00
Fangjun Kuang
5a0b9bcb23
Refactoring (#4)
* Fix an error in TDNN-LSTM training.

* WIP: Refactoring

* Refactor transformer.py

* Remove unused code.

* Minor fixes.
2021-08-04 14:53:02 +08:00
Fangjun Kuang
398ed80d7a Minor fixes to support DDP training. 2021-07-31 15:26:57 +08:00
Fangjun Kuang
b94d97da37 Disable gradient computation in evaluation mode. 2021-07-29 20:37:31 +08:00
Fangjun Kuang
acc63a9172 WIP: Add BPE training code. 2021-07-29 20:23:52 +08:00