Commit Graph

3523 Commits

Author SHA1 Message Date
Carlos Mocholí ec4f592ecf
Sequential `CombinedLoader` to flatten the eval and predict loops (#16726) 2023-02-17 17:37:11 +01:00
Ethan Harris 7f92d5c9d4
[App] Refactor plugins to be a standalone `LightningPlugin` (#16765) 2023-02-17 11:01:38 +00:00
Justus Schock ac5fa03385
Introduce new precision layout in fabric (#16767) 2023-02-17 10:41:18 +00:00
Ethan Harris 3a354acc61
[App] Reserve APP_SERVER_PORT in cloud port allocation (#16782)
Co-authored-by: thomas chaton <thomas@grid.ai>
2023-02-17 09:33:17 +00:00
Adrian Wälchli 91e692c767
Rename the TPUSpawnStrategy to XLAStrategy (#16781)
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
2023-02-17 02:06:24 +00:00
Ethan Harris 6e359dcc86
[App] Fix idle timeout e2e (#16786) 2023-02-17 01:52:46 +00:00
Adrian Wälchli 57c1138525
Trigger colossalai integration test in CI (#16789) 2023-02-16 23:45:09 +00:00
Carlos Mocholí d27881e388
Fix `set_epoch` not getting called for prediction dataloaders (#16785) 2023-02-16 23:08:13 +01:00
Carlos Mocholí 746c734e6a
`SequentialMode` and `dataloader_iter` improvements (#16784) 2023-02-16 23:07:46 +01:00
Carlos Mocholí 51d44f57dd
Prefetch if it's not a sized iterable (#16776) 2023-02-16 14:10:16 +00:00
Carlos Mocholí c9452df005
Run XLA's dataloader validation per dataloader (#16775) 2023-02-16 15:09:23 +01:00
Justus Schock 47c69cd8eb
Remove DP (#16748)
Co-authored-by: Carlos Mocholí <carlossmocholi@gmail.com>
Co-authored-by: Adrian Wälchli <aedu.waelchli@gmail.com>
2023-02-16 05:12:08 +00:00
Carlos Mocholí 52c981267e
Make the trainer a required loop argument (#16771) 2023-02-15 21:40:34 -05:00
Lightning Forever 41dd0d1f85
Remove the QuantizationAwareTraining callback (#16750) 2023-02-15 17:29:49 -05:00
Adrian Wälchli 83f4c83582
Replace ColossalAIStrategy with external implementation (#16757) 2023-02-15 15:11:52 +00:00
Adrian Wälchli 565d3bb8c6
CI: Update colossalai version (#16747)
Co-authored-by: Carlos Mocholí <carlossmocholi@gmail.com>
2023-02-14 14:07:29 -05:00
Carlos Mocholí fbbbbf6ad7
Make dataloader_idx optional for batch start/end hooks (#16753) 2023-02-14 18:31:09 +01:00
thomas chaton 104290efa5
[App] Add rm one level below project level (#16740)
Co-authored-by: Ethan Harris <ethanwharris@gmail.com>
Co-authored-by: Justus Schock <12886177+justusschock@users.noreply.github.com>
Co-authored-by: thomas <thomas@thomass-MacBook-Pro.local>
2023-02-14 14:12:51 +00:00
Adrian Wälchli c4074419b5
Remove the BaguaStrategy (#16746)
* remove bagua

* remove

* remove docker file entry
2023-02-14 08:58:58 -05:00
Adrian Wälchli 39020887d2
Remove Trainer's `track_grad_norm` argument (#16745) 2023-02-14 12:38:17 +00:00
thomas chaton 7e8400d277
[App] Add support for private data (#16738)
Co-authored-by: thomas <thomas@thomass-MacBook-Pro.local>
2023-02-14 12:34:44 +00:00
Carlos Mocholí 6c037a479f
"sequential" mode for `CombinedLoader` (#16743) 2023-02-14 06:51:48 +01:00
Carlos Mocholí a342410e25
Remove the unused `utilities.parsing.flatten_dict` (#16744) 2023-02-14 06:50:11 +01:00
Adrian Wälchli 99cb2cd056
Remove argparse utils (#16708)
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Carlos Mocholí <carlossmocholi@gmail.com>
Co-authored-by: Jirka Borovec <6035284+Borda@users.noreply.github.com>
2023-02-13 20:44:30 +00:00
Adrian Wälchli 67c09e3092
Separate the Gradient Accumulation Scheduler from Trainer (#16729)
Co-authored-by: Carlos Mocholí <carlossmocholi@gmail.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
2023-02-13 20:15:38 +00:00
Jirka Borovec 0be025e8b7
rename docs/source-app & adjust docs links for lightning (#16676)
* update CI

* config / import

* lightning_app imports

* source/ dir

* html

* ci: dirs

* pr

* req dir

* on push

* rename

* drop

* cleaning
2023-02-13 10:59:02 +01:00
Carlos Mocholí d660379a9f
Refactor `CombinedLoader` using pytrees (#16714) 2023-02-11 17:55:33 +01:00
Adrian Wälchli 5340d960b9
Fix min-epochs and early-stopping triggering too many validation runs (#16719)
Co-authored-by: Jirka Borovec <6035284+Borda@users.noreply.github.com>
2023-02-11 03:02:39 +00:00
Carlos Mocholí c35fee92f4
Move `inference_mode` logic to the loops (#16704) 2023-02-10 16:31:21 +01:00
Carlos Mocholí 44c9e385ac
Remove custom log events (#16707) 2023-02-10 14:46:26 +00:00
Carlos Mocholí 5196eaa526
Group `torch.compile` utilities together (#16711) 2023-02-10 13:45:40 +01:00
Adrian Wälchli c4c4793d56
Fix strategy type validation in connectors (#16693) 2023-02-10 10:50:56 +00:00
Carlos Mocholí b7c05d279c
Move `Trainer._log_hyperparams` to an utility (#16712) 2023-02-10 09:22:56 +01:00
Jirka Borovec 4b2cf36e77
move `_HPU_AVAILABLE` (#16713) 2023-02-10 08:56:31 +01:00
Carlos Mocholí 457cd76d1a
Remove the unused `utilities.finite_checks` (#16682) 2023-02-09 21:11:05 +01:00
Adrian Wälchli 18106a8f95
Split train- and val progress into separate bars (#16695) 2023-02-09 19:43:50 +00:00
Sherin Thomas 44557b9de8
[App] Connect and Disconnect node (#16700)
Connect and Disconnect node
2023-02-09 19:21:51 +05:30
thomas chaton 88e089ea4e
[App] Enable to register data connections (#16670)
Co-authored-by: thomas <thomas@thomass-MacBook-Pro.local>
2023-02-08 18:41:05 +00:00
Jirka Borovec 560d6d7956
tests: drop slow flag (#16692)
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
2023-02-08 17:09:59 +00:00
Carlos Mocholí 9f7fbbf264
Move fetchers classes with the loops (#16678) 2023-02-08 16:13:20 +01:00
Carlos Mocholí 338776a388
Move logger utilities with the loggers (#16680) 2023-02-08 16:12:53 +01:00
Adrian Wälchli 74ee699dfd
Add `Strategy.on_exception` (#16646) 2023-02-08 14:00:31 +00:00
Ethan Harris 1288e4ccc4
[App] Initial plugin server (#16523) 2023-02-07 21:38:37 +00:00
Ethan Harris 434ea2560f
[App] Fix e2e CI, use display name in show logs command (#16679) 2023-02-07 19:31:56 +00:00
Carlos Mocholí 71bf4fc581
Flatten fetching abstract interface (#16664) 2023-02-07 18:12:59 +01:00
Carlos Mocholí 2add5d377e
Refactor supporters (#16662) 2023-02-07 18:12:42 +01:00
Carlos Mocholí 35b8543c82
Remove outputs from `on_predict_epoch_end` (#16655) 2023-02-07 17:55:02 +01:00
Adrian Wälchli 923a842e9c
Fix import from torch.distributed when distributed not available (#16658) 2023-02-07 04:51:59 -05:00
Carlos Mocholí 0c1e33e565
Refactor deterministic and benchmark logic (#16653)
Co-authored-by: Adrian Wälchli <aedu.waelchli@gmail.com>
2023-02-07 04:53:26 +01:00
thomas chaton accd2b9e61
[App] Improve Storage Commands (#16645)
* update

* update

* update

* update

---------

Co-authored-by: thomas <thomas@thomass-MacBook-Pro.local>
2023-02-06 15:54:50 -05:00