Commit Graph

4 Commits

Author SHA1 Message Date
chaton ef8ef12fd0
[feat] pp 2/n (#5026)
* Added changes for RPC plugin

* Add missing kwargs

* Fix code format

* Loading refactors by introducing is_distributed var, fix optimizer step flow

* Add rpc guard

* Added docstrings and typing

* resolve comments

* Add additional rpc hook, refactor name of exit process hook for clarity

* remove annotation

* Modify behaviour to allow optional return, add test for rpc plugin

* resolve tests

* rename is_ddp_based

* update

* update for windows

* update

* resolve test

* code smell

* Added sequential plugin

* resolve bug

* update

* cleanup

* add Exception

* resolve docs

* Remove ddp support

* Revert distributed -> ddp

* Update pl_examples/basic_examples/conv_sequential_example.py

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

* Update pl_examples/basic_examples/conv_sequential_example.py

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

* Update pytorch_lightning/plugins/ddp_sequential_plugin.py

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

* Address code review points

* Update pytorch_lightning/plugins/ddp_sequential_plugin.py

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

* Update pytorch_lightning/plugins/ddp_sequential_plugin.py

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

* Add missing return

* Fix formatting, add datamodule args

* add small comment

* resolve comments

* resolve comments

* update source for fairscale

* update extras

* remove staticmethod

* resolve flake8

* Skip tests that are failing due to bug upstream with multiple optimizers and shard

* update

* update on comments

* clean test

* latest comments

* remove old comments

* add todo

* Update version

* update

* resolve bugs

* resolve bugs

* update test

* remove hanging test

* Update pytorch_lightning/plugins/ddp_sequential_plugin.py

Co-authored-by: Carlos Mocholí <carlossmocholi@gmail.com>

* resolve on comments

* Update pytorch_lightning/plugins/ddp_sequential_plugin.py

Co-authored-by: Carlos Mocholí <carlossmocholi@gmail.com>

* resolve on comments

* Update pytorch_lightning/plugins/ddp_sequential_plugin.py

Co-authored-by: Carlos Mocholí <carlossmocholi@gmail.com>

* Update pytorch_lightning/plugins/ddp_sequential_plugin.py

Co-authored-by: Carlos Mocholí <carlossmocholi@gmail.com>

* Update pytorch_lightning/plugins/ddp_sequential_plugin.py

Co-authored-by: Carlos Mocholí <carlossmocholi@gmail.com>

* Update pytorch_lightning/plugins/ddp_sequential_plugin.py

Co-authored-by: Carlos Mocholí <carlossmocholi@gmail.com>

* remove ImportError

Co-authored-by: SeanNaren <sean@grid.ai>
Co-authored-by: Sean Naren <sean.narenthiran@gmail.com>
Co-authored-by: Jirka Borovec <Borda@users.noreply.github.com>
Co-authored-by: Carlos Mocholí <carlossmocholi@gmail.com>
2020-12-09 12:56:51 +00:00
Sean Naren ee9b3fe574
[feat] pp 1/n (#5016)
* Added changes for RPC plugin

* Add missing kwargs

* Fix code format

* Loading refactors by introducing is_distributed var, fix optimizer step flow

* Add rpc guard

* Added docstrings and typing

* resolve comments

* Add additional rpc hook, refactor name of exit process hook for clarity

* remove annotation

* Modify behaviour to allow optional return, add test for rpc plugin

* resolve tests

* rename is_ddp_based

* update

* update for windows

* update

* resolve test

* code smell

* Revert back to init_ddp_connection for backwards compat

* Swap to explicit name for property

* Add missing speed parity increase for CI variability, fix call counts for child process

Co-authored-by: tchaton <thomas@grid.ai>
2020-12-08 22:02:10 +00:00
chaton 2393474350
[hotfix] ddp + manual_optimisation (#4976)
* Rely on ddp plugin for blocking sync behaviour, and skip if we're using manual optimization

* debug

* Revert "debug"

This reverts commit ccca6b6b

* Expose manual reduce for automatic optimization

* Add input arguments

* Enable parity test

* clean imports

* Expose hook after to ensure we reset

* Fix naming

* add

* fix test

* resolve on comments

* typo

* Update tests/trainer/optimization/test_manual_optimization.py

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

* Update tests/trainer/optimization/test_manual_optimization.py

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

* update on comments

* resolve comments

Co-authored-by: SeanNaren <sean@grid.ai>
Co-authored-by: Sean Naren <sean.narenthiran@gmail.com>
Co-authored-by: Jirka Borovec <Borda@users.noreply.github.com>
2020-12-07 19:31:54 +00:00
chaton 02152c1729
Simplify optimization Logic (#4984)
* Rely on ddp plugin for blocking sync behaviour, and skip if we're using manual optimization

* debug

* Revert "debug"

This reverts commit ccca6b6b

* Expose manual reduce for automatic optimization

* Add input arguments

* Enable parity test

* clean imports

* Expose hook after to ensure we reset

* Fix naming

* add

* fix test

* uniformize optimizer logic

* resolve test

* resovle flake8

* resolve amp bug

* update tests

* remove bug

* remove optimizer_step in accelerators

* typo

* update lightning optimizer

* set doesn't work with ddp_spawn

* resolve flake8

* update threshold

* ignore pyright

* correct codeFactor

* remove useless if

* remove zer_grad function

* simplify step

* remove typo

* resolve bug

* Apply suggestions from code review

* update on comments

* resolve bugs

* remove tests

* Update pytorch_lightning/trainer/configuration_validator.py

Co-authored-by: Rohit Gupta <rohitgr1998@gmail.com>

* simplify testing

* add more tests

Co-authored-by: SeanNaren <sean@grid.ai>
Co-authored-by: Jirka Borovec <Borda@users.noreply.github.com>
Co-authored-by: Rohit Gupta <rohitgr1998@gmail.com>
2020-12-07 12:55:49 +00:00