Commit Graph

2486 Commits

Author SHA1 Message Date
Carlos Mocholí 6309a59c3c
Do not prefetch when possible (#12101) 2022-02-28 18:31:18 +00:00
Kushashwa Ravi Shrimali 02ccd874b9
Stop loading a few properties if checkpoint's `dirpath` has changed (#12045)
Co-authored-by: Rohit Gupta <rohitgr1998@gmail.com>
Co-authored-by: Carlos Mocholí <carlossmocholi@gmail.com>
2022-02-28 16:42:09 +00:00
Kaushik B a52a6ea030
Add support for pluggable Accelerators (#12030)
Co-authored-by: Jirka Borovec <Borda@users.noreply.github.com>
Co-authored-by: Rohit Gupta <rohitgr1998@gmail.com>
Co-authored-by: Carlos Mocholí <carlossmocholi@gmail.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Carlos Mocholi <carlossmocholi@gmail.com>
2022-02-28 21:36:23 +05:30
Carlos Mocholí a9024ce870
[CLI] Fix `SaveConfigCallback` with DDP spawn (#12011) 2022-02-28 13:27:42 +00:00
Cai Q.T 01c31ae434
Fix `LightningModule.{un,}toggle_model` when only 1 optimizer is used (#12088)
Co-authored-by: Carlos Mocholí <carlossmocholi@gmail.com>
2022-02-28 12:41:51 +00:00
Rohit Gupta 17bb815d01
Add `estimated_stepping_batches` property to `Trainer` (#11599) 2022-02-28 12:40:48 +00:00
Rohit Gupta 5b342f14a6
fix to avoid common hook warning if no hook is overridden (#12131) 2022-02-28 18:07:05 +05:30
Carlos Mocholí db1c709519
Clean loop fetching usage (#12103) 2022-02-28 10:51:33 +00:00
Carlos Mocholí 5f920dc088
Refactor Horovod NCCL check (#11948) 2022-02-28 10:45:32 +00:00
Mauricio Villegas 54b9a85227
Unit test for CLI with subcommands and a common default config file (#12061)
Co-authored-by: Carlos Mocholí <carlossmocholi@gmail.com>
2022-02-28 10:17:49 +00:00
DuYicong515 c9af112801
Remove `AcceleratorConnector.num_nodes` (#12107)
Co-authored-by: Aki Nitta <nitta@akihironitta.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
2022-02-28 09:53:38 +00:00
Carlos Mocholí 8fd17f2edf
[IPU] Support manually instantiating the `poptorch.DataLoader` (#12116) 2022-02-28 09:36:26 +00:00
DuYicong515 0b677ecf2b
Remove `AcceleratorConnector.has_tpu` (#12109) 2022-02-27 14:16:03 +00:00
DuYicong515 b2932337bc
Remove `AcceleratorConnector.has_ipu` (#12111)
Co-authored-by: Adrian Wälchli <aedu.waelchli@gmail.com>
2022-02-27 13:36:36 +00:00
Adrian Wälchli dc4c3171fc
add parity test for sync batchnorm (#12021) 2022-02-26 03:51:57 +00:00
Akash Kwatra 7e2f9fbad5
Refactor codebase to use `trainer.loggers` over `trainer.logger` when needed (#11920) 2022-02-25 16:01:04 -08:00
Kushashwa Ravi Shrimali 4f68c635a1
Refactor tests, skip if sklearn not available (#12093)
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
2022-02-25 19:31:06 +00:00
Akash Kwatra f5304897ce
Deprecate `BaseProfiler.profile_iterable` (#12102) 2022-02-25 16:26:20 +01:00
Carlos Mocholí 61dd5e4d5e
[IPU] Do not use `DistributedSampler` (#12114) 2022-02-25 14:29:51 +00:00
Rohit Gupta 5d2d9b09df
Avoid patching common `DataHooks` to the `LightningModule` (#10603) 2022-02-25 09:26:59 +01:00
Krishna Kalyan 29d5afbda6
Deprecate callback hooks `on_pretrain_routine_{start,end}` (#11794)
Co-authored-by: rohitgr7 <rohitgr1998@gmail.com>
2022-02-24 19:32:48 +00:00
Kushashwa Ravi Shrimali 00211c1de3
Default `benchmark` based on `deterministic` flag (#11944)
Co-authored-by: Carlos Mocholi <carlossmocholi@gmail.com>
Co-authored-by: Adrian Wälchli <aedu.waelchli@gmail.com>
Co-authored-by: Rohit Gupta <rohitgr1998@gmail.com>
2022-02-24 19:06:03 +00:00
ananthsub 5d040e7fc6
Add support for `dataloader_iter` to validate and test steps (#11546)
Co-authored-by: Carlos Mocholi <carlossmocholi@gmail.com>
2022-02-24 18:01:35 +00:00
Jan Stratil c877d54c04
Fix passing _ddp_params_and_buffers_to_ignore (#11949) 2022-02-24 17:22:48 +00:00
Carlos Mocholí 1fa0639bca
Set the last global step saved only when actually saving (#12057) 2022-02-24 09:43:50 +01:00
Kaushik B 49420548ce
Update warnings for available accelerators not being used (#11909)
Co-authored-by: Carlos Mocholí <carlossmocholi@gmail.com>
2022-02-24 09:12:02 +01:00
Carlos Mocholí 2d043857ee
Add callback to manage fault-tolerance checkpoints (#11862)
Co-authored-by: Danielle Pintz <38207072+daniellepintz@users.noreply.github.com>
2022-02-22 13:55:29 +00:00
Adrian Wälchli d0f54609de
Fix `is_interactive_compatible` logic after AcceleratorConnector rewrite (#12008)
* fix is_interactive_compatible

* improve tests

* update message

* address review

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

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

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
2022-02-22 19:20:54 +05:30
Carlos Mocholí 4c4b9d540f
Remove `Trainer._device_type` (#11992) 2022-02-22 13:02:13 +00:00
Rohit Gupta 5ea811b1d9
Avoid loading dataloaders if `limit_batches=0` (#11576) 2022-02-22 11:33:53 +00:00
Kaushik B 6ff38d4c8e
Update RichProgressBarTheme after detecting light theme on colab (#10993) 2022-02-22 05:32:27 +00:00
Kaushik B dcad2ea4bc
Move strategy tests from accelerators to strategies directory (#11329) 2022-02-22 05:14:18 +00:00
Akash Kwatra e50653da08
Deprecate `agg_key_funcs`, `agg_default_func`, and `update_agg_funcs` from `LightningLoggerBase` (#11871)
Co-authored-by: Danielle Pintz <38207072+daniellepintz@users.noreply.github.com>
2022-02-21 22:35:59 +00:00
Jv Kyle Eclarin 9c067c2a3e
Update `tests/plugins/*.py` to use `devices` instead of `gpus` or `ipus` (#11872) 2022-02-21 22:57:21 +01:00
Carlos Mocholí 3579a3083f
Describe the behavior with `limit_*_batches=1|1.0` (#11950) 2022-02-21 15:12:03 +01:00
Carlos Mocholí 08384c45ea
Support manual optimization step profiling without a trainer reference (#11883) 2022-02-21 13:43:12 +00:00
Carlos Mocholí 0771a553de
Remove `Trainer._strategy_type` (#11990) 2022-02-21 13:39:30 +00:00
four4fish e15a66412c
Add back deterministic support in accelerator_connector (#11999)
Co-authored-by: ananthsub <ananth.subramaniam@gmail.com>
Co-authored-by: Justus Schock <12886177+justusschock@users.noreply.github.com>
2022-02-20 18:11:47 +00:00
Carlos Mocholí 6bc0e1da63
Restore test after #11448 (#11986) 2022-02-19 21:39:41 +00:00
Kushashwa Ravi Shrimali 0374fe65db
Support gradient accumulation using Horovod's `backward_passes_per_step` (#11911)
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Rohit Gupta <rohitgr1998@gmail.com>
Co-authored-by: ananthsub <ananth.subramaniam@gmail.com>
Co-authored-by: Carlos Mocholí <carlossmocholi@gmail.com>
2022-02-19 02:54:04 +01:00
ananthsub cf64f34434
Refactor `Strategy._move_optimizer_states` as utility functions (#11758)
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: thomas chaton <thomas@grid.ai>
2022-02-18 08:36:07 +00:00
Akash Kwatra d61371922b
Deprecate and remove calls to agg_and_log_metrics (#11832)
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Adrian Wälchli <aedu.waelchli@gmail.com>
2022-02-18 02:54:33 +00:00
four4fish 6e14209185
Rewrite accelerator_connector (#11448) 2022-02-17 23:38:39 +00:00
Carlos Mocholí a0ca8d076f
Refactor early stopping test (#11866) 2022-02-17 23:20:39 +00:00
Rohit Gupta 25b505508d
Add process launchers (#11643)
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Adrian Wälchli <aedu.waelchli@gmail.com>
Co-authored-by: ananthsub <ananth.subramaniam@gmail.com>
Co-authored-by: Carlos Mocholí <carlossmocholi@gmail.com>
2022-02-17 21:16:51 +00:00
Konstantin Dobler d902b59886
Fix environment variable order for global rank determination (#11406)
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Adrian Wälchli <aedu.waelchli@gmail.com>
2022-02-17 02:09:05 +00:00
edward-io 87bd54aedf
fix typos (#11937) 2022-02-16 17:27:51 -08:00
Carlos Mocholí 4cc05b2cf9
Support optimizer step progress tracking with manual optimization (#11848) 2022-02-16 21:27:27 +00:00
Jirka Borovec 7bc87015ea
Unblock GPU CI (#11934)
Co-authored-by: Carlos Mocholi <carlossmocholi@gmail.com>
2022-02-16 21:15:44 +01:00
jjenniferdai d69b33f1f0
Introduce `Stateful` PrecisionPlugin (#11638)
Co-authored-by: Rohit Gupta <rohitgr1998@gmail.com>
Co-authored-by: ananthsub <ananth.subramaniam@gmail.com>
2022-02-14 15:56:09 +05:30