4d06301c18
[FIX] Enable mixed precision in the Fully Sharded Strategy when `precision=16` ( #12965 )
...
* Fix fully sharded mixed precision setter
* Add CHANGELOG.md
2022-05-03 15:39:59 +00:00
f4505ce6b2
Construct the hook kwargs inside each loop ( #12100 )
...
Co-authored-by: Rohit Gupta <rohitgr1998@gmail.com>
2022-05-03 17:08:02 +02:00
cd01856ffc
Add `LightningDataModule.load_from_checkpoint` to load datamodules directly from checkpoint ( #12550 )
...
Co-authored-by: Carlos Mocholí <carlossmocholi@gmail.com>
Co-authored-by: Jirka Borovec <Borda@users.noreply.github.com>
Co-authored-by: otaj <ota@grid.ai>
2022-05-03 12:27:06 +00:00
1c25ab8daf
Support CLI shorthand natively ( #12614 )
...
Co-authored-by: Carlos Mocholí <carlossmocholi@gmail.com>
2022-05-03 14:16:37 +02:00
c278802b64
Prefix log_metrics keys with class name in callbacks ( #12228 )
...
Co-authored-by: Carlos Mocholí <carlossmocholi@gmail.com>
2022-05-03 12:20:24 +02:00
eebba9e632
Enforce eval shuffle warning only for default samplers ( #12653 )
...
Co-authored-by: Adrian Wälchli <aedu.waelchli@gmail.com>
Co-authored-by: Carlos Mocholí <carlossmocholi@gmail.com>
2022-05-02 16:11:09 +00:00
917918ade3
Remove duplicate boring classes ( #12951 )
2022-05-02 17:42:12 +02:00
0e4c4424fd
fix enumerate usage ( #12949 )
2022-05-02 15:29:56 +02:00
0f418a8640
Add missing `super().__init__()` calls ( #12948 )
2022-05-02 15:20:02 +02:00
26acdd6569
Add hook test for reloading with max epochs ( #12932 )
2022-05-02 14:41:28 +02:00
456cc87954
Fuse_modules in a qat-respecting way ( #12891 )
...
* Fuse_modules in a qat-respecting way
* Add compatibility for PyTorch <1.11
In older pytorch versions, `fuse_modules` used the `Module.training`
flag to determine wheter fusion should be QAT-compliant or not, refer
https://github.com/pytorch/pytorch/releases/tag/v1.11.0
* Add CHANGELOG for pull #12891
* Fix conditional import of fuse_modules_qat
`torch.ao.quantization.fuse_modules_qat` was actually added in
torch 1.11.
* Update CHANGELOG.md
Co-authored-by: Akihiro Nitta <nitta@akihironitta.com>
2022-05-02 12:37:08 +00:00
ecd135e939
Update nvidia gpg key to fix nightly docker builds ( #12930 )
...
* Update gpg key
* Use curl instead of wget
* Install key manually
2022-05-02 09:00:44 +02:00
62c73904c0
Override `optimizer_zero_grad` when using the `IPUStrategy` ( #12913 )
...
Co-authored-by: Sean Naren <sean@grid.ai>
Co-authored-by: Carlos Mocholí <carlossmocholi@gmail.com>
Co-authored-by: Rohit Gupta <rohitgr1998@gmail.com>
2022-05-01 22:47:21 +00:00
46c59d04db
Exclude the CHANGELOG from the pre-commit size check ( #12931 )
2022-04-29 13:23:02 +00:00
61e44103ea
Fix pickling of KFoldLoop ( #12441 )
...
* allow pickling of KFoldLoop
* Update pl_examples/loop_examples/kfold.py
Co-authored-by: Adrian Wälchli <aedu.waelchli@gmail.com>
Co-authored-by: Nicolas Berger <nicolas.berger@inait.ai>
Co-authored-by: Kushashwa Ravi Shrimali <kushashwaravishrimali@gmail.com>
2022-04-29 11:52:42 +00:00
00a3321d64
Remove use of jsonargparse internals ( #12918 )
2022-04-29 11:02:09 +00:00
c461854fa7
Versioning of last checkpoins ( #12902 )
...
* last checkpoint versioning
* changelog
* Simplify test
* Update CHANGELOG.md
Co-authored-by: Carlos Mocholí <carlossmocholi@gmail.com>
* Update CHANGELOG.md
Co-authored-by: Carlos Mocholí <carlossmocholi@gmail.com>
Co-authored-by: Carlos Mocholí <carlossmocholi@gmail.com>
2022-04-29 14:13:50 +09:00
74d46d655d
Threading support for legacy loading of checkpoints ( #12814 )
2022-04-28 20:37:58 +00:00
6896f00a16
Invoke parent DDP configuration for torch>1.10.2 ( #12912 )
...
Co-authored-by: Carlos Mocholí <carlossmocholi@gmail.com>
2022-04-28 19:31:05 +00:00
55b3bc3e36
Print ragged dict of metrics in `EvaluationLoop._print_results` properly ( #12857 )
...
* first fix
* full bugfix + tests
* Apply Adrian's suggestion
* Add test with tensor(0)
* Minor code simplification
* change sorting to make the comment correct
Co-authored-by: Carlos Mocholí <carlossmocholi@gmail.com>
2022-04-28 16:05:24 +00:00
bcbd9c359e
ShardedGradScaler should only be set for FP16 ( #12915 )
...
Co-authored-by: Carlos Mocholí <carlossmocholi@gmail.com>
2022-04-28 17:44:31 +02:00
0fe2ae74e9
Remove reference to genindex and search in index.rst ( #12919 )
2022-04-28 10:48:31 -04:00
7271095b7b
Merge pull request #12920 from PyTorchLightning/rename/lightning_extension
...
Update docs conf with renamed lightning extension
2022-04-28 10:48:09 -04:00
a62c227932
Support automatic seeding of the LightningCLI ( #12822 )
...
Co-authored-by: Carlos Mocholí <carlossmocholi@gmail.com>
2022-04-28 14:24:32 +00:00
f29d0ab6f8
Fix typo in predict_step docs ( #12911 )
2022-04-28 12:26:57 +00:00
98b206e836
Use cmake installed with apt ( #12907 )
2022-04-28 07:44:52 +00:00
9ff1b39981
Update CHANGELOG after the 1.6.2 release ( #12904 )
2022-04-28 08:53:54 +02:00
0b87bfc65c
Update the link to jsonargparse's link_arguments ( #12898 )
2022-04-28 04:27:27 +09:00
a95bbe6953
Update TPU Accelerator docs ( #12850 )
...
Co-authored-by: Rohit Gupta <rohitgr1998@gmail.com>
2022-04-27 22:38:26 +05:30
c781a9ea59
clean up side menu ( #12892 )
...
Co-authored-by: Akihiro Nitta <nitta@akihironitta.com>
Co-authored-by: Kaushik B <45285388+kaushikb11@users.noreply.github.com>
Co-authored-by: Kaushik B <kaushikbokka@gmail.com>
2022-04-27 22:38:13 +05:30
f3e746c145
Fix tests related to DDP communication hooks ( #12878 )
...
* Fix ddp_comm_hook tests
* Refactor ddp_comm_hook tests
Co-authored-by: Akihiro Nitta <akihiro@pytorchlightning.ai>
2022-04-27 22:37:19 +05:30
10c7a7c84f
Fix `trainer.logger` deprecation message ( #12671 )
2022-04-27 16:11:34 +02:00
70754bea83
Fix to ensure the checkpoint states are saved in a common filepath with deepspeed ( #12887 )
...
Co-authored-by: Akihiro Nitta <nitta@akihironitta.com>
2022-04-27 14:41:51 +02:00
6490996b39
Support deterministic="warn" in Trainer for Pytorch 1.11+ ( #12588 )
...
Co-authored-by: carmocca <carlossmocholi@gmail.com>
Co-authored-by: Akihiro Nitta <nitta@akihironitta.com>
2022-04-27 12:05:26 +00:00
a41486245a
Use a single instance of `rich.console.Console` throughout the codebase ( #12886 )
2022-04-27 01:47:43 +00:00
cac02a04ba
Update `jsonargparse` to unblock master ( #12884 )
...
Co-authored-by: Mauricio Villegas <mauricio_ville@yahoo.com>
2022-04-26 16:55:07 +00:00
f541481da6
Fix reference to basic level doc ( #12848 )
2022-04-26 11:59:29 -04:00
ab60cdbdcb
Raise better error when calling `Trainer.save_checkpoint` without a model attached ( #12772 )
...
* add error message
* add test
* changelog
Co-authored-by: Carlos Mocholí <carlossmocholi@gmail.com>
2022-04-26 11:16:41 +01:00
bb81802bff
Update `deepspeed` and `fairscale` versions ( #12860 )
...
* Fix deepspeed installation
* Adapt to deepspeed>=0.5.9
* Fix fairscale installation
Co-authored-by: Akihiro Nitta <akihiro@pytorchlightning.ai>
2022-04-26 01:40:25 +02:00
2a60161d49
Clarify `fast_dev_run` docs ( #12751 )
...
Co-authored-by: Rohit Gupta <rohitgr1998@gmail.com>
2022-04-25 16:41:23 +00:00
958310a3fc
Fixed encoding issues on terminals that do not support unicode characters ( #12828 )
...
Co-authored-by: Carlos Mocholí <carlossmocholi@gmail.com>
2022-04-25 12:24:30 +00:00
afd2025943
Fix false positive deprecation warning from `register_ddp_comm_hook` ( #12846 )
...
* Use new rank_zero_debug
* Fix and move import statement to the top
2022-04-24 23:46:18 +02:00
8034919c44
Remove deprecated `TestTubeLogger` ( #12859 )
...
* remove deprecated test_tube logger
* remove testube from logger __init__
* remove relevant testtube tests
* update CHANGELOG with removal of deprecated `TestTubeLogger`
2022-04-24 20:05:48 +02:00
70309d898f
Add missing `f` prefix to f-strings ( #12869 )
2022-04-24 19:39:35 +09:00
7eda0bf0c2
Use an uniform call hook style in the loops ( #12742 )
...
Co-authored-by: Carlos Mocholí <carlossmocholi@gmail.com>
2022-04-24 02:42:44 +09:00
f931e27373
Remove the deprecated get_progress_bar_dict ( #12839 )
...
Co-authored-by: Raymond G Schireman <raymond.schireman@uvm.edu>
Co-authored-by: Carlos Mocholí <carlossmocholi@gmail.com>
2022-04-22 22:22:26 +00:00
c4bb078c94
Refactor `PredictionLoop.on_run_start` for consistency ( #12732 )
...
Co-authored-by: Carlos Mocholí <carlossmocholi@gmail.com>
2022-04-22 22:10:26 +00:00
f4f70a8a08
Add required for positional arguments in argparse logic ( #12504 )
...
Co-authored-by: Adrian Wälchli <aedu.waelchli@gmail.com>
2022-04-22 17:39:39 +02:00
5b511da4ad
Call `on_predict_model_eval` hook through method ( #12741 )
...
Co-authored-by: carmocca <carlossmocholi@gmail.com>
2022-04-22 11:27:12 +00:00
63ad92efda
Fix formatting issue in Trainer docs ( #12777 )
2022-04-22 11:45:20 +01:00
Sean Naren