Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Update learning rate on each backward pass instead of each forward pass. #1477

Merged
merged 8 commits into from
Apr 20, 2020
Merged

Update learning rate on each backward pass instead of each forward pass. #1477

Show file tree
Hide file tree
Changes from 7 commits
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
d1a05ae
change lr scheduler step interval to update every backwards pass inst…
rmrao Apr 13, 2020
548a363
update CHANGELOG
rmrao Apr 13, 2020
1ead46c
fix spacing
rmrao Apr 13, 2020
390095a
Merge branch 'master' into master
rmrao Apr 15, 2020
2f1be61
Add TODO to lr schedule update
rmrao Apr 15, 2020
8ce73a4
remove trailing whitespace
rmrao Apr 15, 2020
9d29f57
Merge branch 'master' into master
williamFalcon Apr 16, 2020
0c583d3
Merge branch 'master' into master
williamFalcon Apr 20, 2020
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
11 changes: 5 additions & 6 deletions CHANGELOG.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -17,29 +17,28 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).

- Changed the default behaviour to no longer include a NaN check with each training iteration. ([#1475](https://github.com/PyTorchLightning/pytorch-lightning/pull/1475))

- Changed lr schedule step interval behavior to update every backwards pass instead of every forwards pass ([#1476](https://github.com/PyTorchLightning/pytorch-lightning/issues/1476))

- Updated semantic segmentation example with custom u-net and logging ([#1371](https://github.com/PyTorchLightning/pytorch-lightning/pull/1371))

-

### Deprecated

-
-


### Removed

-
-

-
-

### Fixed

- Fixed loggers - flushing last logged metrics even before continue, e.g. `trainer.test()` results ([#1459](https://github.com/PyTorchLightning/pytorch-lightning/pull/1459))

- Added a missing call to the `on_before_zero_grad` model hook ([#1493](https://github.com/PyTorchLightning/pytorch-lightning/pull/1493)).

-


## [0.7.3] - 2020-04-09

Expand Down
7 changes: 5 additions & 2 deletions pytorch_lightning/trainer/training_loop.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -450,8 +450,11 @@ def run_training_epoch(self):
# when returning -1 from train_step, we end epoch early
early_stop_epoch = batch_result == -1

# update lr
self.update_learning_rates(interval='step')
# TODO: consolidate all actions that need to take place only after
# self.accumulate_grad_batches steps (optimizer step, lr update, global step increment)
if (self.batch_idx + 1) % self.accumulate_grad_batches == 0:
# update lr
self.update_learning_rates(interval='step')

# ---------------
# RUN VAL STEP
Expand Down