Skip to content

[pull] master from tensorflow:master#263

Merged
pull[bot] merged 1 commit intoseewpx:masterfrom
tensorflow:master
Mar 6, 2020
Merged

[pull] master from tensorflow:master#263
pull[bot] merged 1 commit intoseewpx:masterfrom
tensorflow:master

Conversation

@pull
Copy link
Copy Markdown

@pull pull bot commented Mar 6, 2020

See Commits and Changes for more details.


Created by pull[bot]. Want to support this open source service? Please star it : )

In BERT SQuAD, disable explicit allreduce for now to keep the original clip_by_global_norm math. With explicit allreduce, the gradients before allreduce are scaled so even if we move clip_by_global_norm before allreduce (as in TF1 and pre-TF 2.2) it will operate on scaled gradients, the math will be changed. So with explicit allreduce, it is better to move clip_by_global_norm to after allreduce.

PiperOrigin-RevId: 299278082
@pull pull bot added the ⤵️ pull label Mar 6, 2020
@pull pull bot merged commit 11ccb99 into seewpx:master Mar 6, 2020
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

0 participants