Add `Trainer(gradient_clip_algorithm='value'|'norm')` #6123

dhkim0225 · 2021-02-22T12:37:01Z

What does this PR do?

Add "clip_gradients_by_value"
Fixes #5460
fixes #5456
fixes #4927
Previous PR: #5477
@tchaton Thanks for giving me a lot of advice.

Before submitting

Was this discussed/approved via a GitHub issue? (not for typos and docs)
Did you read the contributor guideline, Pull Request section?
Did you make sure your PR does only one thing, instead of bundling different changes together?
Did you make sure to update the documentation with your changes? (if necessary)
Did you write any new necessary tests? (not for typos and docs)
Did you verify new and existing tests pass locally with your changes?
Did you update the CHANGELOG? (not for typos, docs, test updates, or internal minor changes/refactorings)

PR review

Anyone in the community is free to review the PR once the tests have passed.
Before you start reviewing make sure you have read Review guidelines. In short, see the following bullet-list:

Is this pull request ready for review? (if not, please submit in draft mode)
Check that all items from Before submitting are resolved
Make sure the title is self-explanatory and the description concisely explains the PR
Add labels and milestones (and optionally projects) to the PR so it can be classified

Did you have fun?

Make sure you had fun coding 🙃

codecov · 2021-02-22T12:39:07Z

Codecov Report

Merging #6123 (4d5e05f) into master (6dc1078) will decrease coverage by 6%.
The diff coverage is 94%.

@@           Coverage Diff           @@
##           master   #6123    +/-   ##
=======================================
- Coverage      91%     86%    -6%     
=======================================
  Files         192     192            
  Lines       12192   12513   +321     
=======================================
- Hits        11149   10721   -428     
- Misses       1043    1792   +749

Borda · 2021-02-22T17:32:58Z

Related Issues: #5460, #5456, #4927

what does it mean that it solves these issues?

CHANGELOG.md

pytorch_lightning/accelerators/accelerator.py

pytorch_lightning/plugins/precision/deepspeed_precision.py

docs/source/advanced/training_tricks.rst

pytorch_lightning/utilities/enums.py

Co-authored-by: Jirka Borovec <Borda@users.noreply.github.com>

dhkim0225 · 2021-02-23T03:04:03Z

@Borda Thank you for your comments.
I changed codes based on comments.

Related Issues: #5460, #5456, #4927

what does it mean that it solves these issues?

It means, this PR adds a clip_gradient by value option.

pytorch_lightning/plugins/precision/precision_plugin.py

dhkim0225 · 2021-03-25T16:20:38Z

Bugs in the test codes clip gradients by value with horovod are fixed.

pytorch_lightning/plugins/precision/tpu_bfloat.py

pytorch_lightning/plugins/precision/mixed.py

…lightning into feat/clip_grad_by_value

tests/models/test_horovod.py

dhkim0225 · 2021-04-06T05:00:10Z

@carmocca Honestly, I'm tired of the slow progress.
The first time this PR was written was on January 12th. (#5477)
Merge is continuously postponed and new conflicts continue to occur, although It is a simple feature.
Please clearly guide me if there is anything I can do to make this feature merge quickly.

Borda · 2021-04-06T09:41:18Z

pytorch_lightning/plugins/precision/precision_plugin.py

+        clip_grad_func = self.clip_grad_funcs[gradient_clip_algorithm]
+        clip_grad_func(optimizer, clip_val)
+
+    def clip_grad_by_value(self, optimizer: 'Optimizer', clip_val: Union[int, float]) -> None:


or protected :]

carmocca · 2021-04-06T09:46:27Z

Hey @dhkim0225! Thanks a lot for your patience. This PR is just waiting for the final approval so I will just push it internally 👍

No point in testing a signature

williamFalcon · 2021-04-06T13:28:18Z

merged!
sorry about that! Looks like this one fell through the cracks.

Thanks for the contribution!

…ter) to github/third-party/PyTorchLightning/pytorch-lightning Summary: ### New commit log messages ## [UnReleased] - 2021-MM-DD ### Added - Added more explicit exception message when trying to execute `trainer.test()` or `trainer.validate()` with `fast_dev_run=True` ([#6667](Lightning-AI/pytorch-lightning#6667)) - Added `LightningCLI` class to provide simple reproducibility with minimum boilerplate training cli. ([#4492](Lightning-AI/pytorch-lightning#4492)) - Trigger warning when non-metric logged value with multi processes hasn't been reduced ([#6417](Lightning-AI/pytorch-lightning#6417)) - Added `gradient_clip_algorithm` argument to Trainer for gradient clipping by value ([#6123](Lightning-AI/pytorch-lightning#6123)). - Added a way to print to terminal without breaking up the progress bar ([#5470](Lightning-AI/pytorch-lightning#5470)) - Added support to checkpoint after training steps in `ModelCheckpoint` callback ([#6146](Lightning-AI/pytorch-lightning#6146)) - Added `checkpoint` parameter to callback's `on_save_checkpoint` hook ([#6072](Lightning-AI/pytorch-lightning#6072)) - Added `RunningStage.SANITY_CHECKING` ([#4945](Lightning-AI/pytorch-lightning#4945)) - Added `TrainerState.{FITTING,VALIDATING,TESTING,PREDICTING,TUNING}` ([#4945](Lightning-AI/pytorch-lightning#4945)) - Added `Trainer.validate()` method to perform one evaluation epoch over the validation set ([#4948](Lightning-AI/pytorch-lightning#4948)) - Added `LightningEnvironment` for Lightning-specific DDP ([#5915](Lightning-AI/pytorch-lightning#5915)) - Added `teardown()` hook to LightningDataModule ([#4673](Lightning-AI/pytorch-lightning#4673)) - Added `auto_insert_metric_name` parameter to `ModelCheckpoint` ([#6277](Lightning-AI/pytorch-lightning#6277)) - Added arg to `self.log` that enables users to give custom names when dealing with multiple dataloaders ([#6274](Lightning-AI/pytorch-lightning#6274)) - Added `teardown` method to `BaseProfiler` to enable subclasses defining post-profiling steps outside of `__del__` ([#6370](Lightning-AI/pytorch-lightning#6370)) - Added `setup` method to `BaseProfiler` to enable subclasses defining pre-profiling steps for every process ([#6633](Lightning-AI/pytorch-lightning#6633)) - Added no return warning to predict ([#6139](Lightning-AI/pytorch-lightning#6139)) - Added `Trainer.predict` config validation ([#6543](Lightning-AI/pytorch-lightning#6543)) - Added `AbstractProfiler` interface ([#6621](Lightning-AI/pytorch-lightning#6621)) - Added support for including module names for forward in the autograd trace of `PyTorchProfiler` ([#6349](Lightning-AI/pytorch-lightning#6349)) - Added support for the PyTorch 1.8.1 autograd profiler ([#6618](Lightning-AI/pytorch-lightning#6618)) - Added `outputs` parameter to callback's `on_validation_epoch_end` & `on_test_epoch_end` hooks ([#6120](Lightning-AI/pytorch-lightning#6120)) - Added `configure_sharded_model` hook ([#6679](Lightning-AI/pytorch-lightning#6679)) - Added support for `precision=64`, enabling training with double precision ([#6595](Lightning-AI/pytorch-lightning#6595)) - Added support for DDP communication hooks ([#6736](Lightning-AI/pytorch-lightning#6736)) - Added `artifact_location` argument to `MLFlowLogger` which will be passed to the `MlflowClient.create_experiment` call ([#6677](Lightning-AI/pytorch-lightning#6677)) - Added `model` parameter to precision plugins' `clip_gradients` signature ([#6764](Lightning-AI/pytorch-lightning#6764)) ### Changed - Renamed `pytorch_lightning.callbacks.swa` to `pytorch_lightning.callbacks.stochastic_weight_avg` ([#6259](Lightning-AI/pytorch-lightning#6259)) - Refactor `RunningStage` and `TrainerState` usage ([#4945](Lightning-AI/pytorch-lightning#4945)) - Changed `trainer.evaluating` to return `True` if validating or testing ([#4945](Lightning-AI/pytorch-lightning#4945)) - Changed `setup()` and `teardown()` stage argument to take any of `{fit,validate,test,predict}` ([#6386](Lightning-AI/pytorch-lightning#6386)) - Changed profilers to save separate report files per state and rank ([#6621](Lightning-AI/pytorch-lightning#6621)) - Changed `PyTorchProfiler` to use `torch.autograd.profiler.record_function` to record functions ([#6349](Lightning-AI/pytorch-lightning#6349)) ### Deprecated - `period` has been deprecated in favor of `every_n_val_epochs` in the `ModelCheckpoint` callback ([#6146](Lightning-AI/pytorch-lightning#6146)) - Deprecated `trainer.running_sanity_check` in favor of `trainer.sanity_checking` ([#4945](Lightning-AI/pytorch-lightning#4945)) - Deprecated `Profiler(output_filename)` in favor of `dirpath` and `filename` ([#6621](Lightning-AI/pytorch-lightning#6621)) - Deprecated `PytorchProfiler(profiled_functions)` in favor of `record_functions` ([#6349](Lightning-AI/pytorch-lightning#6349)) - Deprecated metrics in favor of `torchmetrics` ([#6505](Lightning-AI/pytorch-lightning#6505), [#6530](Lightning-AI/pytorch-lightning#6530), [#6540](Lightning-AI/pytorch-lightning#6540), [#6547](Lightning-AI/pytorch-lightning#6547), [#6515](Lightning-AI/pytorch-lightning#6515), [#6572](Lightning-AI/pytorch-lightning#6572), [#6573](Lightning-AI/pytorch-lightning#6573), [#6584](Lightning-AI/pytorch-lightning#6584), [#6636](Lightning-AI/pytorch-lightning#6636), [#6637](Lightning-AI/pytorch-lightning#6637), [#6649](Lightning-AI/pytorch-lightning#6649), [#6659](Lightning-AI/pytorch-lightning#6659), ) ### Removed - Removed support for passing a bool value to `profiler` argument of Trainer ([#6164](Lightning-AI/pytorch-lightning#6164)) - Removed no return warning from val/test step ([#6139](Lightning-AI/pytorch-lightning#6139)) - Removed passing a `ModelCheckpoint` instance to `Trainer(checkpoint_callback)` ([#6166](Lightning-AI/pytorch-lightning#6166)) - Removed deprecated Trainer argument `enable_pl_optimizer` and `automatic_optimization` ([#6163](Lightning-AI/pytorch-lightning#6163)) - Removed deprecated metrics ([#6161](Lightning-AI/pytorch-lightning#6161)) * from `pytorch_lightning.metrics.functional.classification` removed `to_onehot`, `to_categorical`, `get_num_classes`, `roc`, `multiclass_roc`, `average_precision`, `precision_recall_curve`, `multiclass_precision_recall_curve` * from `pytorch_lightning.metrics.functional.reduction` removed `reduce`, `class_reduce` - Removed deprecated `ModelCheckpoint` arguments `prefix`, `mode="auto"` ([#6162](Lightning-AI/pytorch-lightning#6162)) - Removed `mode='auto'` from `EarlyStopping` ([#6167](Lightning-AI/pytorch-lightning#6167)) - Removed legacy references for magic keys in the `Result` object ([#6016](Lightning-AI/pytorch-lightning#6016)) - Removed deprecated `LightningModule` `hparams` setter ([#6207](Lightning-AI/pytorch-lightning#6207)) - Removed legacy code to log or include metrics in the progress bar by returning them in a dict with the `"log"/"progress_bar"` magic keys. Use `self.log` instead ([#6734](Lightning-AI/pytorch-lightning#6734)) - Removed `optimizer_idx` argument from `training_step` in manual optimization ([#6093](Lightning-AI/pytorch-lightning#6093)) ### Fixed - Set better defaults for `rank_zero_only.rank` when training is launched with SLURM and torchelastic ([#6802](Lightning-AI/pytorch-lightning#6802)) - Made the `Plugin.reduce` method more consistent across all Plugins to reflect a mean-reduction by default ([#6011](Lightning-AI/pytorch-lightning#6011)) - Move lightning module to correct device type when using LightningDistributedWrapper ([#6070](Lightning-AI/pytorch-lightning#6070)) - Do not print top-k verbose log with `ModelCheckpoint(monitor=None)` ([#6109](Lightning-AI/pytorch-lightning#6109)) - Fixed csv extension check ([#6436](Lightning-AI/pytorch-lightning#6436)) - Fixed `ModelCheckpoint(monitor=None, save_last=True)` not saving checkpoints ([#6136](Lightning-AI/pytorch-lightning#6136)) - Fixed `ModelCheckpoint(save_top_k=0, save_last=True)` not saving the `last` checkpoint ([#6136](Lightning-AI/pytorch-lightning#6136)) - Fixed `.teardown(stage='fit')` getting called during `trainer.test` ([#6386](Lightning-AI/pytorch-lightning#6386)) - Fixed `.on_fit_{start,end}()` getting called during `trainer.test` ([#6386](Lightning-AI/pytorch-lightning#6386)) - Fixed LightningModule `all_gather` on cpu tensors ([#6416](Lightning-AI/pytorch-lightning#6416)) - Fixed torch distributed not available in setup hook for DDP ([#6506](Lightning-AI/pytorch-lightning#6506)) - Fixed `EarlyStopping` logic when `min_epochs` or `min_steps` requirement is not met ([#6705](Lightning-AI/pytorch-lightning#6705)) ## [1.2.7] - 2021-04-06 ### Fixed - Fixed resolve a bug with omegaconf and xm.save ([#6741](Lightning-AI/pytorch-lightning#6741)) - Fixed an issue with IterableDataset when __len__ is not defined ([#6828](Lightning-AI/pytorch-lightning#6828)) - Sanitize None params during pruning ([#6836](Lightning-AI/pytorch-lightning#6836)) - Enforce an epoch scheduler interval when using SWA ([#6588](Lightning-AI/pytorch-lightning#6588)) - Fixed TPU Colab hang issue, post training ([#6816](Lightning-AI/pytorch-lightning#6816)) - Fixed a bug where `TensorBoardLogger` would give a warning and not log correctly to a symbolic link `save_dir` ([#6730](Lightning-AI/pytorch-lightning#6730)) ## [1.2.6] - 2021-03-30 ### Changed - Changed the behavior of `on_epoch_start` to run at the beginning of validation & test epoch ([#6498](Lightning-AI/pytorch-lightning#6498)) ### Removed - Removed legacy code to include `step` dictionary returns in `callback_metrics`. Use `self.log_dict` instead. ([#6682](Lightning-AI/pytorch-lightning#6682)) ### Fixed - Fixed `DummyLogger.log_hyperparams` raising a `TypeError` when running with `fast_dev_run=True` ([#6398](Lightning-AI/pytorch-lightning#6398)) - Fixed error on TPUs when there was no `ModelCheckpoint` ([#6654](Lightning-AI/pytorch-lightning#6654)) - Fixed `trainer.test` freeze on TPUs ([#6654](Lightning-AI/pytorch-lightning#6654)) - Fixed a bug where gradients were disabled after calling `Trainer.predict` ([#6657](Lightning-AI/pytorch-lightning#6657)) - Fixed bug where no TPUs were detected in a TPU pod env ([#6719](Lightning-AI/pytorch-lightning#6719)) ## [1.2.5] - 2021-03-23 ### Changed - Update Gradient Clipping for the TPU Accelerator ([#6576](Lightning-AI/pytorch-lightning#6576)) - Refactored setup for typing friendly ([#6590](Lightning-AI/pytorch-lightning#6590)) ### Fixed - Fixed a bug where `all_gather` would not work correctly with `tpu_cores=8` ([#6587](Lightning-AI/pytorch-lightning#6587)) - Fixed comparing required versions ([#6434](Lightning-AI/pytorch-lightning#6434)) - Fixed duplicate logs appearing in console when using the python logging module ([#6275](Lightning-AI/pytorch-lightning#6275)) - Added Autocast in validation, test and predict modes for Native AMP ([#6565](Lightning-AI/pytorch-lightning#6565)) Reviewed By: shuyingsunshine21 Differential Revision: D27528929 fbshipit-source-id: 311c88f71461c2c79bbf185e28d7a6d683ccc26f

add changelog

09ea112

dhkim0225 requested review from awaelchli, Borda, carmocca, justusschock, SeanNaren, tchaton and williamFalcon as code owners February 22, 2021 12:37

dhkim0225 marked this pull request as draft February 22, 2021 12:37

dhkim0225 changed the title ~~add changelog~~ add clip_grad_by_value feature Feb 22, 2021

dhkim0225 added 5 commits February 22, 2021 23:16

add clip by value

c0e8064

fix bug in training tricks.rst

ca8e6fd

fix bug in trainer.rst

87f12c1

Update trainer.rst

8e43b8a

Update trainer.rst

2bb9924

dhkim0225 marked this pull request as ready for review February 22, 2021 15:37

dhkim0225 requested a review from edenlightning as a code owner February 22, 2021 15:37

Borda added the feature Is an improvement or enhancement label Feb 22, 2021

Borda added this to the 1.3 milestone Feb 22, 2021

Borda reviewed Feb 22, 2021

View reviewed changes

dhkim0225 and others added 5 commits February 23, 2021 10:20

Update CHANGELOG.md

5b83f0d

Co-authored-by: Jirka Borovec <Borda@users.noreply.github.com>

Update pytorch_lightning/plugins/precision/deepspeed_precision.py

caafdf2

Co-authored-by: Jirka Borovec <Borda@users.noreply.github.com>

Update pytorch_lightning/utilities/enums.py

ca774b6

Co-authored-by: Jirka Borovec <Borda@users.noreply.github.com>

yapf formatting

5a741e2

update training tricks

0568e3a

dhkim0225 requested a review from Borda February 23, 2021 03:04

ananthsub reviewed Feb 23, 2021

View reviewed changes

pytorch_lightning/plugins/precision/precision_plugin.py Outdated Show resolved Hide resolved

carmocca added ready PRs ready to be merged Will labels Mar 25, 2021

carmocca reviewed Mar 25, 2021

View reviewed changes

pytorch_lightning/plugins/precision/tpu_bfloat.py Outdated Show resolved Hide resolved

pytorch_lightning/plugins/precision/mixed.py Outdated Show resolved Hide resolved

dhkim0225 added 3 commits March 26, 2021 12:45

revert some changes

e7e3b47

revert some changes

2997536

Merge branch 'master' of https://github.com/PyTorchLightning/pytorch-…

fb34e84

…lightning into feat/clip_grad_by_value

mergify bot added the has conflicts label Mar 26, 2021

Merge branch 'master' into feat/clip_grad_by_value

8e665ec

mergify bot removed the has conflicts label Mar 27, 2021

carmocca reviewed Mar 27, 2021

View reviewed changes

tests/models/test_horovod.py Outdated Show resolved Hide resolved

Update tests/models/test_horovod.py

9575774

mergify bot added the has conflicts label Mar 29, 2021

Merge branch 'master' into feat/clip_grad_by_value

7c16f6a

mergify bot added has conflicts and removed has conflicts labels Mar 29, 2021

merge master

fec189a

mergify bot removed the has conflicts label Apr 6, 2021

merge master

1e80304

Borda approved these changes Apr 6, 2021

View reviewed changes

Delete signature test

4d5e05f

No point in testing a signature

williamFalcon approved these changes Apr 6, 2021

View reviewed changes

williamFalcon merged commit 7f6154f into Lightning-AI:master Apr 6, 2021

dhkim0225 deleted the feat/clip_grad_by_value branch April 6, 2021 13:30

ceshine mentioned this pull request Apr 9, 2021

Trainer(gradient_clip_algorithm='value') has no effect (from #6123) #6920

Closed

ceshine mentioned this pull request Apr 15, 2021

Fix the gradient_clip_algorithm has no effect issue. #6928

Merged

11 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add `Trainer(gradient_clip_algorithm='value'|'norm')` #6123

Add `Trainer(gradient_clip_algorithm='value'|'norm')` #6123

dhkim0225 commented Feb 22, 2021 •

edited by Borda

Loading

codecov bot commented Feb 22, 2021 •

edited

Loading

Borda commented Feb 22, 2021

dhkim0225 commented Feb 23, 2021

dhkim0225 commented Mar 25, 2021 •

edited

Loading

dhkim0225 commented Apr 6, 2021

Borda Apr 6, 2021

carmocca commented Apr 6, 2021

williamFalcon commented Apr 6, 2021

Add Trainer(gradient_clip_algorithm='value'|'norm') #6123

Add Trainer(gradient_clip_algorithm='value'|'norm') #6123

Conversation

dhkim0225 commented Feb 22, 2021 • edited by Borda Loading

What does this PR do?

Before submitting

PR review

Did you have fun?

codecov bot commented Feb 22, 2021 • edited Loading

Codecov Report

Borda commented Feb 22, 2021

dhkim0225 commented Feb 23, 2021

dhkim0225 commented Mar 25, 2021 • edited Loading

dhkim0225 commented Apr 6, 2021

Borda Apr 6, 2021

Choose a reason for hiding this comment

carmocca commented Apr 6, 2021

williamFalcon commented Apr 6, 2021

Add `Trainer(gradient_clip_algorithm='value'|'norm')` #6123

Add `Trainer(gradient_clip_algorithm='value'|'norm')` #6123

dhkim0225 commented Feb 22, 2021 •

edited by Borda

Loading

codecov bot commented Feb 22, 2021 •

edited

Loading

dhkim0225 commented Mar 25, 2021 •

edited

Loading