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.

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

Make gradients available for all_gather on TPU #15003

Merged
merged 19 commits into from
Dec 8, 2022
Merged

Make gradients available for all_gather on TPU #15003

merged 19 commits into from
Dec 8, 2022

Conversation

stekiri
Copy link
Contributor

@stekiri stekiri commented Oct 5, 2022

What does this PR do?

Support autograd for all_gather on TPU using torch_xla.core.functions.all_gather

Fixes #6295

Does your PR introduce any breaking changes? If yes, please list them.

None

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 list all the breaking changes introduced by this pull request?
  • Did you update the CHANGELOG? (not for typos, docs, test updates, or minor internal changes/refactors)

PR review

Anyone in the community is welcome to review the PR.
Before you start reviewing, make sure you have read the 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 🙃

@github-actions github-actions bot added the pl Generic label for PyTorch Lightning package label Oct 5, 2022
@stekiri stekiri marked this pull request as ready for review October 13, 2022 11:53
@awaelchli awaelchli self-assigned this Oct 13, 2022
@awaelchli awaelchli added accelerator: tpu Tensor Processing Unit fabric lightning.fabric.Fabric labels Oct 13, 2022
@carmocca carmocca added feature Is an improvement or enhancement community This PR is from the community and removed fabric lightning.fabric.Fabric labels Oct 13, 2022
src/pytorch_lightning/strategies/tpu_spawn.py Outdated Show resolved Hide resolved
src/lightning_lite/strategies/xla.py Outdated Show resolved Hide resolved
tests/tests_pytorch/accelerators/test_tpu.py Outdated Show resolved Hide resolved
Copy link
Contributor

@carmocca carmocca left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

#15349 needs to be merged first

@stekiri
Copy link
Contributor Author

stekiri commented Nov 2, 2022

@carmocca, I ran my new tests successfully on Colab using a TPU runtime. However, the test-on-tpus run in this PR keeps failing, I seem to hit the 100 minutes timeout and there seem to be a problem with the logs: error: You must be logged in to the server (Unauthorized).

Copy link
Contributor

@carmocca carmocca left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You can ignore the "Unauthorized" error. If the job timed out it's because there's no hardware availability at this moment.

tests/tests_pytorch/accelerators/test_tpu.py Outdated Show resolved Hide resolved
@mergify mergify bot added the ready PRs ready to be merged label Nov 3, 2022
@carmocca carmocca assigned carmocca and unassigned awaelchli Nov 9, 2022
@Borda Borda requested review from awaelchli and removed request for otaj, rohitgr7 and kaushikb11 November 22, 2022 08:37
@carmocca carmocca added this to the v1.9 milestone Nov 26, 2022
@awaelchli
Copy link
Contributor

@stekiri We haven't forgotten this PR. Some more patience is required, sorry about this. We have some issue getting the TPU CI to work on GitHub here #15788.

@Borda Borda enabled auto-merge (squash) December 8, 2022 04:08
@Borda Borda merged commit 0d822e4 into Lightning-AI:master Dec 8, 2022
@stekiri stekiri deleted the tpu-all-gather-with-grads branch December 8, 2022 13:12
justusschock added a commit that referenced this pull request Dec 9, 2022
* Simplify enabling CPU offload in FSDP (#15832)


Co-authored-by: Jirka Borovec <[email protected]>

* [App] Enable running with spawn context (#15923)

* Fix compiler support test (#15927)

* Enable back inference mode support with hpu & update links (#15918)

* Enable back inference mode support with hpu
* Remove unused
* Update document link and address comment

Signed-off-by: Jerome <[email protected]>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

* [App] Introduce auto scaler (#15769)

* Exlucde __pycache__ in setuptools

* Add load balancer example

* wip

* Update example

* rename

* remove prints

* _LoadBalancer -> LoadBalancer

* AutoScaler(work)

* change var name

* remove locust

* Update docs

* include autoscaler in api ref

* docs typo

* docs typo

* docs typo

* docs typo

* remove unused loadtest

* remove unused device_type

* clean up

* clean up

* clean up

* Add docstring

* type

* env vars to args

* expose an API for users to override to customise autoscaling logic

* update example

* comment

* udpate var name

* fix scale mechanism and clean up

* Update exampl

* ignore mypy

* Add test file

* .

* update impl and update tests

* Update changlog

* .

* revert docs

* update test

* update state to keep calling 'flow.run()'

Co-authored-by: Aniket Maurya <[email protected]>

* Add aiohttp to base requirements

* Update docs

Co-authored-by: Luca Antiga <[email protected]>

* Use deserializer utility

* fake trigger

* wip: protect /system/* with basic auth

* read password at runtime

* Change env var name

* import torch as optional

* Don't overcreate works

* simplify imports

* Update example

* aiohttp

* Add work_args work_kwargs

* More docs

* remove FIXME

* Apply Jirka's suggestions

Co-authored-by: Jirka Borovec <[email protected]>

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* clean example device

* add comment on init threshold value

* bad merge

* nit: logging format

* {in,out}put_schema -> {in,out}put_type

* lowercase

* docs on seconds

* process_time -> processing_time

* Dont modify work state from flow

* Update tests

* worker_url -> endpoint

* fix exampl

* Fix default scale logic

* Fix default scale logic

* Fix num_pending_works

* Update num_pending_works

* Fix bug creating too many works

* Remove up/downscale_threshold args

* Update example

* Add typing

* Fix example in docstring

* Fix default scale logic

* Update src/lightning_app/components/auto_scaler.py

Co-authored-by: Noha Alon <[email protected]>

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* rename method

* rename locvar

* Add todo

* docs ci

* docs ci

* asdfafsdasdf pls docs

* Apply suggestions from code review

Co-authored-by: Ethan Harris <[email protected]>

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* .

* doc

* Update src/lightning_app/components/auto_scaler.py

Co-authored-by: Noha Alon <[email protected]>

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Revert "[pre-commit.ci] auto fixes from pre-commit.com hooks"

This reverts commit 24983a0.

* Revert "Update src/lightning_app/components/auto_scaler.py"

This reverts commit 56ea78b.

* Remove redefinition

* Remove load balancer run blocker

* raise RuntimeError

* remove has_sent

* lower the default timeout_batching from 10 to 1

* remove debug

* update the default timeout_batching

* .

* tighten condition

* fix endpoint

* typo in runtimeerror cond

* async lock update severs

* add a test

* {in,out}put_type typing

* Update examples/app_server_with_auto_scaler/app.py

Co-authored-by: Jirka Borovec <[email protected]>

* Update .actions/setup_tools.py

Co-authored-by: Aniket Maurya <[email protected]>
Co-authored-by: Luca Antiga <[email protected]>
Co-authored-by: Jirka Borovec <[email protected]>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Noha Alon <[email protected]>
Co-authored-by: Ethan Harris <[email protected]>
Co-authored-by: Akihiro Nitta <[email protected]>
Co-authored-by: thomas chaton <[email protected]>

* ENG-627: Docs for CloudCompute Mount Argument (#15182)

fixed conflicts

* Fix LRScheduler import for PyTorch 2.0 (#15940)

* Fix LRScheduler import for PyTorch 2.0
* Add comment for posterity

* CI: fix pypi flow (#15944)

* CI: fixing pypi syntax (#15943)
* connect
* input

* [App] Remove `SingleProcessRuntime` (#15933)

* Remove SingleProcessRuntime
* Remove unused queues
* Docs

* [App] Fix bug when using structures with works (#15911)

* Fix bug when using structures with works
* Add test
* Update CHANGELOG.md

* [App] Wait for full file to be transferred in Path / Payload (#15934)

* Wait for full file to be transferred in Path / Payload
* Fixes

* [docs] Include all components in the API reference (#15805)

* Update docs

Co-authored-by: Jirka Borovec <[email protected]>

* Bump playwright from 1.27.1 to 1.28.0 in /requirements (#15903)

* Bump playwright from 1.27.1 to 1.28.0 in /requirements

Bumps [playwright](https://github.com/Microsoft/playwright-python) from 1.27.1 to 1.28.0.
- [Release notes](https://github.com/Microsoft/playwright-python/releases)
- [Commits](microsoft/playwright-python@v1.27.1...v1.28.0)

---
updated-dependencies:
- dependency-name: playwright
  dependency-type: direct:production
  update-type: version-update:semver-minor
...

Signed-off-by: dependabot[bot] <[email protected]>

* 1.28

Signed-off-by: dependabot[bot] <[email protected]>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: Jirka <[email protected]>

* [App] Add `configure_layout` method for works (#15926)

* Add `configure_layout` method for works
* Check for api access availability
* Updates from review
* Update CHANGELOG.md
* Apply suggestions from code review

Co-authored-by: Sherin Thomas <[email protected]>

* Make gradients available for all_gather on TPU (#15003)

* Make gradients available for all_gather on TPU
* Modify switch and tests
* Apply suggestions from code review
* Modify tests
* Fix test
* Drop test

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Jirka Borovec <[email protected]>
Co-authored-by: Adrian Wälchli <[email protected]>
Co-authored-by: Carlos Mocholí <[email protected]>
Co-authored-by: Jirka Borovec <[email protected]>

* Don't try to aggregate `requirements/__pycache__/base.txt` in setuptools (#15775)

Exlucde __pycache__ in setuptools

* [App] Multiprocessing-safe work pickling (#15836)

* Upgrade to HPU release 1.7.1 (#15956)

* Upgrade to HPU release 1.7.1
Update torch version check for hpu

Signed-off-by: Jerome <[email protected]>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

* Multinode on MPS (#15748)

* Fix restarting attribute for lr finder
* update lite executor
* update trainer executor
* update spawn executor
* add multinode component tests
* add testing helpers
* add lite tests
* add trainer tests
* update changelog
* update trainer
* update workflow
* update tests
* debug
* add reason for skipif
* Apply suggestions from code review
* switch skipif

Co-authored-by: Jirka <[email protected]>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Carlos Mocholí <[email protected]>
Co-authored-by: Adrian Wälchli <[email protected]>
Co-authored-by: Jirka Borovec <[email protected]>

* [App] Resolve PythonServer on M1 (#15949)


Co-authored-by: thomas <[email protected]>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

* Lite: Fix DataLoader shuffling when using DistributedSampler (#15931)


Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

* [App] Temporarily disable ready (#15958)

* Fix restarting attribute for lr finder (#15620)

* [App] Improve pdb for multiprocessing (#15950)


Co-authored-by: thomas <[email protected]>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

* [App] Improve debug triggering (#15951)

* [App] Add automatic conversion to structures (#15961)

* Make LightningModule torch.jit.script-able again (#15947)

* Make LightningModule torch.jit.script-able again
* remove skip

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

* refactor: simplify Tensor import (#15959)

* Fix ImportErrors on Multinode if package not present (#15963)

* Fix typo in definition of world size in docs (#15954)

* [App] Enable running an app from the Gallery (#15941)


Co-authored-by: thomas <[email protected]>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Ethan Harris <[email protected]>
Co-authored-by: Jirka <[email protected]>

* Apply dynamo to training_step, validation_step, test_step, predict_step (#15957)

* Apply dynamo to training_step, validation_step, test_step, predict_step

* Add entry to CHANGELOG.md

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix merge conflict

* rename tpu workflow

Signed-off-by: Jerome <[email protected]>
Signed-off-by: dependabot[bot] <[email protected]>
Co-authored-by: Jirka Borovec <[email protected]>
Co-authored-by: thomas chaton <[email protected]>
Co-authored-by: Luca Antiga <[email protected]>
Co-authored-by: Jerome Anand <[email protected]>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Akihiro Nitta <[email protected]>
Co-authored-by: Aniket Maurya <[email protected]>
Co-authored-by: Noha Alon <[email protected]>
Co-authored-by: Ethan Harris <[email protected]>
Co-authored-by: Akihiro Nitta <[email protected]>
Co-authored-by: Rick Izzo <[email protected]>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: Jirka <[email protected]>
Co-authored-by: Sherin Thomas <[email protected]>
Co-authored-by: stekiri <[email protected]>
Co-authored-by: Jirka Borovec <[email protected]>
Co-authored-by: Carlos Mocholí <[email protected]>
Co-authored-by: Justus Schock <[email protected]>
Co-authored-by: thomas <[email protected]>
carmocca added a commit that referenced this pull request Jan 4, 2023
* Simplify enabling CPU offload in FSDP (#15832)

Co-authored-by: Jirka Borovec <[email protected]>

* [App] Enable running with spawn context (#15923)

* Fix compiler support test (#15927)

* Enable back inference mode support with hpu & update links (#15918)

* Enable back inference mode support with hpu
* Remove unused
* Update document link and address comment

Signed-off-by: Jerome <[email protected]>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

* [App] Introduce auto scaler (#15769)

* Exlucde __pycache__ in setuptools

* Add load balancer example

* wip

* Update example

* rename

* remove prints

* _LoadBalancer -> LoadBalancer

* AutoScaler(work)

* change var name

* remove locust

* Update docs

* include autoscaler in api ref

* docs typo

* docs typo

* docs typo

* docs typo

* remove unused loadtest

* remove unused device_type

* clean up

* clean up

* clean up

* Add docstring

* type

* env vars to args

* expose an API for users to override to customise autoscaling logic

* update example

* comment

* udpate var name

* fix scale mechanism and clean up

* Update exampl

* ignore mypy

* Add test file

* .

* update impl and update tests

* Update changlog

* .

* revert docs

* update test

* update state to keep calling 'flow.run()'

Co-authored-by: Aniket Maurya <[email protected]>

* Add aiohttp to base requirements

* Update docs

Co-authored-by: Luca Antiga <[email protected]>

* Use deserializer utility

* fake trigger

* wip: protect /system/* with basic auth

* read password at runtime

* Change env var name

* import torch as optional

* Don't overcreate works

* simplify imports

* Update example

* aiohttp

* Add work_args work_kwargs

* More docs

* remove FIXME

* Apply Jirka's suggestions

Co-authored-by: Jirka Borovec <[email protected]>

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* clean example device

* add comment on init threshold value

* bad merge

* nit: logging format

* {in,out}put_schema -> {in,out}put_type

* lowercase

* docs on seconds

* process_time -> processing_time

* Dont modify work state from flow

* Update tests

* worker_url -> endpoint

* fix exampl

* Fix default scale logic

* Fix default scale logic

* Fix num_pending_works

* Update num_pending_works

* Fix bug creating too many works

* Remove up/downscale_threshold args

* Update example

* Add typing

* Fix example in docstring

* Fix default scale logic

* Update src/lightning_app/components/auto_scaler.py

Co-authored-by: Noha Alon <[email protected]>

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* rename method

* rename locvar

* Add todo

* docs ci

* docs ci

* asdfafsdasdf pls docs

* Apply suggestions from code review

Co-authored-by: Ethan Harris <[email protected]>

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* .

* doc

* Update src/lightning_app/components/auto_scaler.py

Co-authored-by: Noha Alon <[email protected]>

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Revert "[pre-commit.ci] auto fixes from pre-commit.com hooks"

This reverts commit 24983a0.

* Revert "Update src/lightning_app/components/auto_scaler.py"

This reverts commit 56ea78b.

* Remove redefinition

* Remove load balancer run blocker

* raise RuntimeError

* remove has_sent

* lower the default timeout_batching from 10 to 1

* remove debug

* update the default timeout_batching

* .

* tighten condition

* fix endpoint

* typo in runtimeerror cond

* async lock update severs

* add a test

* {in,out}put_type typing

* Update examples/app_server_with_auto_scaler/app.py

Co-authored-by: Jirka Borovec <[email protected]>

* Update .actions/setup_tools.py

Co-authored-by: Aniket Maurya <[email protected]>
Co-authored-by: Luca Antiga <[email protected]>
Co-authored-by: Jirka Borovec <[email protected]>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Noha Alon <[email protected]>
Co-authored-by: Ethan Harris <[email protected]>
Co-authored-by: Akihiro Nitta <[email protected]>
Co-authored-by: thomas chaton <[email protected]>

* ENG-627: Docs for CloudCompute Mount Argument (#15182)

fixed conflicts

* Fix LRScheduler import for PyTorch 2.0 (#15940)

* Fix LRScheduler import for PyTorch 2.0
* Add comment for posterity

* CI: fix pypi flow (#15944)

* CI: fixing pypi syntax (#15943)
* connect
* input

* [App] Remove `SingleProcessRuntime` (#15933)

* Remove SingleProcessRuntime
* Remove unused queues
* Docs

* [App] Fix bug when using structures with works (#15911)

* Fix bug when using structures with works
* Add test
* Update CHANGELOG.md

* [App] Wait for full file to be transferred in Path / Payload (#15934)

* Wait for full file to be transferred in Path / Payload
* Fixes

* [docs] Include all components in the API reference (#15805)

* Update docs

Co-authored-by: Jirka Borovec <[email protected]>

* Bump playwright from 1.27.1 to 1.28.0 in /requirements (#15903)

* Bump playwright from 1.27.1 to 1.28.0 in /requirements

Bumps [playwright](https://github.com/Microsoft/playwright-python) from 1.27.1 to 1.28.0.
- [Release notes](https://github.com/Microsoft/playwright-python/releases)
- [Commits](microsoft/playwright-python@v1.27.1...v1.28.0)

---
updated-dependencies:
- dependency-name: playwright
  dependency-type: direct:production
  update-type: version-update:semver-minor
...

Signed-off-by: dependabot[bot] <[email protected]>

* 1.28

Signed-off-by: dependabot[bot] <[email protected]>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: Jirka <[email protected]>

* [App] Add `configure_layout` method for works (#15926)

* Add `configure_layout` method for works
* Check for api access availability
* Updates from review
* Update CHANGELOG.md
* Apply suggestions from code review

Co-authored-by: Sherin Thomas <[email protected]>

* Make gradients available for all_gather on TPU (#15003)

* Make gradients available for all_gather on TPU
* Modify switch and tests
* Apply suggestions from code review
* Modify tests
* Fix test
* Drop test

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Jirka Borovec <[email protected]>
Co-authored-by: Adrian Wälchli <[email protected]>
Co-authored-by: Carlos Mocholí <[email protected]>
Co-authored-by: Jirka Borovec <[email protected]>

* Don't try to aggregate `requirements/__pycache__/base.txt` in setuptools (#15775)

Exlucde __pycache__ in setuptools

* [App] Multiprocessing-safe work pickling (#15836)

* Upgrade to HPU release 1.7.1 (#15956)

* Upgrade to HPU release 1.7.1
Update torch version check for hpu

Signed-off-by: Jerome <[email protected]>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

* Multinode on MPS (#15748)

* Fix restarting attribute for lr finder
* update lite executor
* update trainer executor
* update spawn executor
* add multinode component tests
* add testing helpers
* add lite tests
* add trainer tests
* update changelog
* update trainer
* update workflow
* update tests
* debug
* add reason for skipif
* Apply suggestions from code review
* switch skipif

Co-authored-by: Jirka <[email protected]>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Carlos Mocholí <[email protected]>
Co-authored-by: Adrian Wälchli <[email protected]>
Co-authored-by: Jirka Borovec <[email protected]>

* [App] Resolve PythonServer on M1 (#15949)

Co-authored-by: thomas <[email protected]>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

* Lite: Fix DataLoader shuffling when using DistributedSampler (#15931)

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

* [App] Temporarily disable ready (#15958)

* Fix restarting attribute for lr finder (#15620)

* [App] Improve pdb for multiprocessing (#15950)

Co-authored-by: thomas <[email protected]>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

* [App] Improve debug triggering (#15951)

* [App] Add automatic conversion to structures (#15961)

* Make LightningModule torch.jit.script-able again (#15947)

* Make LightningModule torch.jit.script-able again
* remove skip

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

* refactor: simplify Tensor import (#15959)

* Fix ImportErrors on Multinode if package not present (#15963)

* Fix typo in definition of world size in docs (#15954)

* [App] Enable running an app from the Gallery (#15941)

Co-authored-by: thomas <[email protected]>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Ethan Harris <[email protected]>
Co-authored-by: Jirka <[email protected]>

* Apply dynamo to training_step, validation_step, test_step, predict_step (#15957)

* Apply dynamo to training_step, validation_step, test_step, predict_step

* Add entry to CHANGELOG.md

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix merge conflict

* rename tpu workflow

Signed-off-by: Jerome <[email protected]>
Signed-off-by: dependabot[bot] <[email protected]>
Co-authored-by: Jirka Borovec <[email protected]>
Co-authored-by: thomas chaton <[email protected]>
Co-authored-by: Luca Antiga <[email protected]>
Co-authored-by: Jerome Anand <[email protected]>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Akihiro Nitta <[email protected]>
Co-authored-by: Aniket Maurya <[email protected]>
Co-authored-by: Noha Alon <[email protected]>
Co-authored-by: Ethan Harris <[email protected]>
Co-authored-by: Akihiro Nitta <[email protected]>
Co-authored-by: Rick Izzo <[email protected]>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: Jirka <[email protected]>
Co-authored-by: Sherin Thomas <[email protected]>
Co-authored-by: stekiri <[email protected]>
Co-authored-by: Jirka Borovec <[email protected]>
Co-authored-by: Carlos Mocholí <[email protected]>
Co-authored-by: Justus Schock <[email protected]>
Co-authored-by: thomas <[email protected]>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
accelerator: tpu Tensor Processing Unit community This PR is from the community feature Is an improvement or enhancement pl Generic label for PyTorch Lightning package ready PRs ready to be merged
Projects
None yet
Development

Successfully merging this pull request may close these issues.

all_gather for TPU doesn't support backward gradients.
5 participants