Devashish ShankarandJirka Borovec ade3f36b7a Raise an error when lightning replaces an existing sampler (#2020) ...

* Raise an error when lightning replaces an existing sampler

Currently, Trainer replaces the existing sampler with DistributedSampler
if running distributing training and `replace_sampler_ddp=True` (default
behaviour). If a user has configured an existing sampler, this would
lead to widely different results if running a distributed vs
non-distributed training.

This PR fixes this by raising an Error if user has configured a sampler
and uses `replace_sampler_ddp=True`. The recommended behavior from now
on is to either remove the sampler or set `replace_sampler_ddp=False`

* Fix tests

* Simpler fix

* Fix tests

* Make inner method protected

* Apply suggestions from code review

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

2020-06-02 18:52:04 -04:00

..

__init__.py

split trainer tests (#956)

2020-02-26 20:31:40 -05:00

test_checks.py

replace Hparams by init args (#1896)

2020-05-24 18:59:08 -04:00

test_dataloaders.py

Raise an error when lightning replaces an existing sampler (#2020)

2020-06-02 18:52:04 -04:00

test_lr_finder.py

Replaces ddp .spawn with subprocess (#2029)

2020-06-01 11:00:32 -04:00

test_optimizers.py

replace Hparams by init args (#1896)

2020-05-24 18:59:08 -04:00

test_trainer_cli.py

handle unknown args passed to Trainer.from_argparse_args (#1932)

2020-05-25 16:01:29 -04:00

test_trainer_tricks.py

unify tests (#1940)

2020-05-27 22:45:23 -04:00

test_trainer.py

Replaces ddp .spawn with subprocess (#2029)

2020-06-01 11:00:32 -04:00