wassname/pytorch-lightning
1
0
Fork 0
mirror of https://github.com/wassname/pytorch-lightning.git synced 2026-09-09 11:32:07 +08:00
Code Issues Packages Projects Releases Wiki Activity

Simplify progress bar args (#1108)

Browse Source 
* show progress bar dependent on refresh_rate

* test progress_bar_refresh control show bar

* remove show_progress_bar from other tests

* borda fixes

* flake8 fix

* changelog update prog bar refresh rate

* move show_progress_bar to deprecated 0.9 api

* rm show_progress_bar references, test deprecated

* Update pytorch_lightning/trainer/__init__.py

* fix test

* changelog

* minor CHANGELOG.md format

* Update pytorch_lightning/trainer/__init__.py

* Update pytorch_lightning/trainer/trainer.py

Co-authored-by: Gerard Bentley <gbkh2015@mymail.pomona.edu>
Co-authored-by: William Falcon <waf2107@columbia.edu>
Co-authored-by: Jirka Borovec <Borda@users.noreply.github.com>
Co-authored-by: J. Borovec <jirka.borovec@seznam.cz>
This commit is contained in:
Gerard Bentley
2020-04-03 00:53:00 +02:00
committed by GitHub
co-authored by Gerard Bentley William Falcon Jirka Borovec J. Borovec
parent 724b787cd1
commit f33b5a8d99
15 changed files with 79 additions and 55 deletions
Download Patch File Download Diff File Expand all files Collapse all files
CHANGELOG.md
+10 -11
View File
@@ -24,6 +24,7 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
### Changed
- Changed `progress_bar_refresh_rate` trainer flag to disable progress bar when set to 0. ([#1108](https://github.com/PyTorchLightning/pytorch-lightning/pull/1108))
- Enhanced `load_from_checkpoint` to also forward params to the model ([#1307](https://github.com/PyTorchLightning/pytorch-lightning/pull/1307))
- Updated references to self.forward() to instead use the `__call__` interface. ([#1211](https://github.com/PyTorchLightning/pytorch-lightning/pull/1211))
- Added option to run without an optimizer by returning `None` from `configure_optimizers`. ([#1279](https://github.com/PyTorchLightning/pytorch-lightning/pull/1279))
@@ -42,6 +43,7 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
### Deprecated
- Deprecated Trainer argument `print_nan_grads` ([#1097](https://github.com/PyTorchLightning/pytorch-lightning/pull/1097))
- Deprecated Trainer argument `show_progress_bar` ([#1108](https://github.com/PyTorchLightning/pytorch-lightning/pull/1108))
### Removed
@@ -70,9 +72,9 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
### Added
- Added automatic sampler setup. Depending on DDP or TPU, lightning configures the sampler correctly (user needs to do nothing) ([#926](https://github.com/PyTorchLightning/pytorch-lightning/pull/926))
- Added `reload_dataloaders_every_epoch=False` flag for trainer. Some users require reloading data every epoch ([#926](https://github.com/PyTorchLightning/pytorch-lightning/pull/926))
- Added `progress_bar_refresh_rate=50` flag for trainer. Throttle refresh rate on notebooks ([#926](https://github.com/PyTorchLightning/pytorch-lightning/pull/926))
- Added automatic sampler setup. Depending on DDP or TPU, lightning configures the sampler correctly (user needs to do nothing) ([#926](https://github.com/PyTorchLightning/pytorch-lightning/pull/926))
- Added `reload_dataloaders_every_epoch=False` flag for trainer. Some users require reloading data every epoch ([#926](https://github.com/PyTorchLightning/pytorch-lightning/pull/926))
- Added `progress_bar_refresh_rate=50` flag for trainer. Throttle refresh rate on notebooks ([#926](https://github.com/PyTorchLightning/pytorch-lightning/pull/926))
- Updated governance docs
- Added a check to ensure that the metric used for early stopping exists before training commences ([#542](https://github.com/PyTorchLightning/pytorch-lightning/pull/542))
- Added `optimizer_idx` argument to `backward` hook ([#733](https://github.com/PyTorchLightning/pytorch-lightning/pull/733))
@@ -95,7 +97,6 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
- Added TPU gradient clipping ([#963](https://github.com/PyTorchLightning/pytorch-lightning/pull/963))
- Added max/min number of steps in `Trainer` ([#728](https://github.com/PyTorchLightning/pytorch-lightning/pull/728))
### Changed
- Improved `NeptuneLogger` by adding `close_after_fit` argument to allow logging after training([#908](https://github.com/PyTorchLightning/pytorch-lightning/pull/1084))
@@ -107,9 +108,9 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
- Freezed models `hparams` as `Namespace` property ([#1029](https://github.com/PyTorchLightning/pytorch-lightning/pull/1029))
- Dropped `logging` config in package init ([#1015](https://github.com/PyTorchLightning/pytorch-lightning/pull/1015))
- Renames model steps ([#1051](https://github.com/PyTorchLightning/pytorch-lightning/pull/1051))
* `training_end` >> `training_epoch_end`
* `validation_end` >> `validation_epoch_end`
* `test_end` >> `test_epoch_end`
- `training_end` >> `training_epoch_end`
- `validation_end` >> `validation_epoch_end`
- `test_end` >> `test_epoch_end`
- Refactor dataloading, supports infinite dataloader ([#955](https://github.com/PyTorchLightning/pytorch-lightning/pull/955))
- Create single file in `TensorBoardLogger` ([#777](https://github.com/PyTorchLightning/pytorch-lightning/pull/777))
@@ -117,7 +118,7 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
- Deprecated `pytorch_lightning.logging` ([#767](https://github.com/PyTorchLightning/pytorch-lightning/pull/767))
- Deprecated `LightningModule.load_from_metrics` in favour of `LightningModule.load_from_checkpoint` ([#995](https://github.com/PyTorchLightning/pytorch-lightning/pull/995), [#1079](https://github.com/PyTorchLightning/pytorch-lightning/pull/1079))
- Deprecated `@data_loader` decorator ([#926](https://github.com/PyTorchLightning/pytorch-lightning/pull/926))
- Deprecated `@data_loader` decorator ([#926](https://github.com/PyTorchLightning/pytorch-lightning/pull/926))
- Deprecated model steps `training_end`, `validation_end` and `test_end` ([#1051](https://github.com/PyTorchLightning/pytorch-lightning/pull/1051), [#1056](https://github.com/PyTorchLightning/pytorch-lightning/pull/1056))
### Removed
@@ -307,9 +308,7 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
### Added
- Added the flag `log_gpu_memory` to `Trainer` to deactivate logging of GPU
memory utilization
- Added SLURM resubmit functionality (port from test-tube)
- Added the flag `log_gpu_memory` to `Trainer` to deactivate logging of GPU memory utilization
- Added optional weight_save_path to trainer to remove the need for a checkpoint_callback when using cluster training
- Added option to use single gpu per node with `DistributedDataParallel`
pytorch_lightning/trainer/__init__.py
+4 -5
View File
@@ -646,6 +646,8 @@ Example::
# default used by the Trainer
trainer = Trainer(progress_bar_refresh_rate=1)
# disable progress bar
trainer = Trainer(progress_bar_refresh_rate=0)
reload_dataloaders_every_epoch
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
@@ -702,12 +704,9 @@ use_amp:
show_progress_bar
^^^^^^^^^^^^^^^^^
If true shows tqdm progress bar
.. warning:: .. deprecated:: 0.7.2
Example::
# default used by the Trainer
trainer = Trainer(show_progress_bar=True)
Set `progress_bar_refresh_rate` to 0 instead. Will remove 0.9.0.
test_percent_check
^^^^^^^^^^^^^^^^^^
pytorch_lightning/trainer/deprecated_api.py
+19
View File
@@ -87,3 +87,22 @@ class TrainerDeprecatedAPITillVer0_8(ABC):
"`num_sanity_val_steps` since v0.5.0"
" and this method will be removed in v0.8.0", DeprecationWarning)
self.num_sanity_val_steps = nb
class TrainerDeprecatedAPITillVer0_9(ABC):
def __init__(self):
super().__init__() # mixin calls super too
@property
def show_progress_bar(self):
"""Back compatibility, will be removed in v0.9.0"""
warnings.warn("Argument `show_progress_bar` is now set by `progress_bar_refresh_rate` since v0.7.2"
" and this method will be removed in v0.9.0", DeprecationWarning)
return self.progress_bar_refresh_rate >= 1
@show_progress_bar.setter
def show_progress_bar(self, tf):
"""Back compatibility, will be removed in v0.9.0"""
warnings.warn("Argument `show_progress_bar` is now set by `progress_bar_refresh_rate` since v0.7.2"
" and this method will be removed in v0.9.0", DeprecationWarning)
pytorch_lightning/trainer/distrib_data_parallel.py
+1 -1
View File
@@ -281,7 +281,7 @@ class TrainerDDPMixin(ABC):
self.node_rank = 0
# show progressbar only on progress_rank 0
self.show_progress_bar = self.show_progress_bar and self.node_rank == 0 and gpu_idx == 0
self.progress_bar_refresh_rate = self.progress_bar_refresh_rate if self.node_rank == 0 and gpu_idx == 0 else 0
# determine which process we are and world size
if self.use_ddp:
pytorch_lightning/trainer/distrib_parts.py
+1 -1
View File
@@ -480,7 +480,7 @@ class TrainerDPMixin(ABC):
self.tpu_global_core_rank = xm.get_ordinal()
# avoid duplicating progress bar
self.show_progress_bar = self.show_progress_bar and self.tpu_global_core_rank == 0
self.progress_bar_refresh_rate = self.progress_bar_refresh_rate if self.tpu_global_core_rank == 0 else 0
# track current tpu
self.current_tpu_idx = tpu_core_idx
pytorch_lightning/trainer/evaluation_loop.py
+2 -3
View File
@@ -163,7 +163,6 @@ class TrainerEvaluationLoopMixin(ABC):
num_val_batches: int
fast_dev_run: ...
process_position: ...
show_progress_bar: ...
process_output: ...
training_tqdm_dict: ...
proc_rank: int
@@ -278,7 +277,7 @@ class TrainerEvaluationLoopMixin(ABC):
dl_outputs.append(output)
# batch done
if batch_idx % self.progress_bar_refresh_rate == 0:
if self.progress_bar_refresh_rate >= 1 and batch_idx % self.progress_bar_refresh_rate == 0:
if test_mode:
self.test_progress_bar.update(self.progress_bar_refresh_rate)
else:
@@ -361,7 +360,7 @@ class TrainerEvaluationLoopMixin(ABC):
desc = 'Testing' if test_mode else 'Validating'
total = max_batches if max_batches != float('inf') else None
pbar = tqdm(desc=desc, total=total, leave=test_mode, position=position,
disable=not self.show_progress_bar, dynamic_ncols=True, file=sys.stdout)
disable=not self.progress_bar_refresh_rate, dynamic_ncols=True, file=sys.stdout)
setattr(self, f'{"test" if test_mode else "val"}_progress_bar', pbar)
# run evaluation
pytorch_lightning/trainer/trainer.py
+14 -7
View File
@@ -21,7 +21,8 @@ from pytorch_lightning.trainer.auto_mix_precision import TrainerAMPMixin
from pytorch_lightning.trainer.callback_config import TrainerCallbackConfigMixin
from pytorch_lightning.trainer.callback_hook import TrainerCallbackHookMixin
from pytorch_lightning.trainer.data_loading import TrainerDataLoadingMixin
from pytorch_lightning.trainer.deprecated_api import TrainerDeprecatedAPITillVer0_8
from pytorch_lightning.trainer.deprecated_api import (TrainerDeprecatedAPITillVer0_8,
TrainerDeprecatedAPITillVer0_9)
from pytorch_lightning.trainer.distrib_data_parallel import TrainerDDPMixin
from pytorch_lightning.trainer.distrib_parts import TrainerDPMixin, parse_gpu_ids, determine_root_gpu_device
from pytorch_lightning.trainer.evaluation_loop import TrainerEvaluationLoopMixin
@@ -66,12 +67,13 @@ class Trainer(
TrainerCallbackConfigMixin,
TrainerCallbackHookMixin,
TrainerDeprecatedAPITillVer0_8,
TrainerDeprecatedAPITillVer0_9,
):
DEPRECATED_IN_0_8 = (
'gradient_clip', 'nb_gpu_nodes', 'max_nb_epochs', 'min_nb_epochs',
'add_row_log_interval', 'nb_sanity_val_steps'
)
DEPRECATED_IN_0_9 = ('use_amp',)
DEPRECATED_IN_0_9 = ('use_amp', 'show_progress_bar')
def __init__(
self,
@@ -86,7 +88,7 @@ class Trainer(
gpus: Optional[Union[List[int], str, int]] = None,
num_tpu_cores: Optional[int] = None,
log_gpu_memory: Optional[str] = None,
show_progress_bar: bool = True,
show_progress_bar=None, # backward compatible, todo: remove in v0.9.0
progress_bar_refresh_rate: int = 1,
overfit_pct: float = 0.0,
track_grad_norm: int = -1,
@@ -161,9 +163,12 @@ class Trainer(
log_gpu_memory: None, 'min_max', 'all'. Might slow performance
show_progress_bar: If true shows tqdm progress bar
show_progress_bar:
.. warning:: .. deprecated:: 0.7.2
progress_bar_refresh_rate: How often to refresh progress bar (in steps)
Set `progress_bar_refresh_rate` to postive integer to enable. Will remove 0.9.0.
progress_bar_refresh_rate: How often to refresh progress bar (in steps). Value ``0`` disables progress bar.
overfit_pct: How much of training-, validation-, and test dataset to check.
@@ -414,7 +419,9 @@ class Trainer(
# can't init progress bar here because starting a new process
# means the progress_bar won't survive pickling
self.show_progress_bar = show_progress_bar
# backward compatibility
if show_progress_bar is not None:
self.show_progress_bar = show_progress_bar
# logging
self.log_save_interval = log_save_interval
@@ -820,7 +827,7 @@ class Trainer(
pbar = tqdm(desc='Validation sanity check',
total=self.num_sanity_val_steps * len(self.val_dataloaders),
leave=False, position=2 * self.process_position,
disable=not self.show_progress_bar, dynamic_ncols=True)
disable=not self.progress_bar_refresh_rate, dynamic_ncols=True)
self.main_progress_bar = pbar
# dummy validation progress bar
self.val_progress_bar = tqdm(disable=True)
pytorch_lightning/trainer/training_loop.py
+1 -1
View File
@@ -623,7 +623,7 @@ class TrainerTrainLoopMixin(ABC):
self.get_model().on_batch_end()
# update progress bar
if batch_idx % self.progress_bar_refresh_rate == 0:
if self.progress_bar_refresh_rate >= 1 and batch_idx % self.progress_bar_refresh_rate == 0:
self.main_progress_bar.update(self.progress_bar_refresh_rate)
self.main_progress_bar.set_postfix(**self.training_tqdm_dict)
tests/models/test_amp.py
+1 -5
View File
@@ -21,7 +21,6 @@ def test_amp_single_gpu(tmpdir):
trainer_options = dict(
default_save_path=tmpdir,
show_progress_bar=True,
max_epochs=1,
gpus=1,
distributed_backend='ddp',
@@ -42,7 +41,6 @@ def test_no_amp_single_gpu(tmpdir):
trainer_options = dict(
default_save_path=tmpdir,
show_progress_bar=True,
max_epochs=1,
gpus=1,
distributed_backend='dp',
@@ -66,7 +64,6 @@ def test_amp_gpu_ddp(tmpdir):
trainer_options = dict(
default_save_path=tmpdir,
show_progress_bar=True,
max_epochs=1,
gpus=2,
distributed_backend='ddp',
@@ -90,7 +87,6 @@ def test_amp_gpu_ddp_slurm_managed(tmpdir):
model = LightningTestModel(hparams)
trainer_options = dict(
show_progress_bar=True,
max_epochs=1,
gpus=[0],
distributed_backend='ddp',
@@ -128,7 +124,7 @@ def test_cpu_model_with_amp(tmpdir):
trainer_options = dict(
default_save_path=tmpdir,
show_progress_bar=False,
progress_bar_refresh_rate=0,
logger=tutils.get_default_testtube_logger(tmpdir),
max_epochs=1,
train_percent_check=0.4,
tests/models/test_cpu.py
+7 -8
View File
@@ -27,7 +27,6 @@ def test_early_stopping_cpu_model(tmpdir):
gradient_clip_val=1.0,
overfit_pct=0.20,
track_grad_norm=2,
show_progress_bar=True,
logger=tutils.get_default_testtube_logger(tmpdir),
train_percent_check=0.1,
val_percent_check=0.1,
@@ -48,7 +47,7 @@ def test_lbfgs_cpu_model(tmpdir):
trainer_options = dict(
default_save_path=tmpdir,
max_epochs=2,
show_progress_bar=False,
progress_bar_refresh_rate=0,
weights_summary='top',
train_percent_check=1.0,
val_percent_check=0.2,
@@ -67,7 +66,7 @@ def test_default_logger_callbacks_cpu_model(tmpdir):
max_epochs=1,
gradient_clip_val=1.0,
overfit_pct=0.20,
show_progress_bar=False,
progress_bar_refresh_rate=0,
train_percent_check=0.01,
val_percent_check=0.01,
)
@@ -95,7 +94,7 @@ def test_running_test_after_fitting(tmpdir):
trainer_options = dict(
default_save_path=tmpdir,
show_progress_bar=False,
progress_bar_refresh_rate=0,
max_epochs=8,
train_percent_check=0.4,
val_percent_check=0.2,
@@ -133,7 +132,7 @@ def test_running_test_without_val(tmpdir):
checkpoint = tutils.init_checkpoint_callback(logger)
trainer_options = dict(
show_progress_bar=False,
progress_bar_refresh_rate=0,
max_epochs=1,
train_percent_check=0.4,
val_percent_check=0.2,
@@ -226,7 +225,7 @@ def test_cpu_model(tmpdir):
trainer_options = dict(
default_save_path=tmpdir,
show_progress_bar=False,
progress_bar_refresh_rate=0,
logger=tutils.get_default_testtube_logger(tmpdir),
max_epochs=1,
train_percent_check=0.4,
@@ -247,7 +246,7 @@ def test_all_features_cpu_model(tmpdir):
gradient_clip_val=1.0,
overfit_pct=0.20,
track_grad_norm=2,
show_progress_bar=False,
progress_bar_refresh_rate=0,
logger=tutils.get_default_testtube_logger(tmpdir),
accumulate_grad_batches=2,
max_epochs=1,
@@ -344,7 +343,7 @@ def test_single_gpu_model(tmpdir):
trainer_options = dict(
default_save_path=tmpdir,
show_progress_bar=False,
progress_bar_refresh_rate=0,
max_epochs=1,
train_percent_check=0.1,
val_percent_check=0.1,
tests/models/test_gpu.py
+4 -5
View File
@@ -27,7 +27,6 @@ def test_multi_gpu_model_ddp2(tmpdir):
model, hparams = tutils.get_default_model()
trainer_options = dict(
default_save_path=tmpdir,
show_progress_bar=True,
max_epochs=1,
train_percent_check=0.4,
val_percent_check=0.2,
@@ -49,7 +48,7 @@ def test_multi_gpu_model_ddp(tmpdir):
model, hparams = tutils.get_default_model()
trainer_options = dict(
default_save_path=tmpdir,
show_progress_bar=False,
progress_bar_refresh_rate=0,
max_epochs=1,
train_percent_check=0.4,
val_percent_check=0.2,
@@ -69,7 +68,7 @@ def test_ddp_all_dataloaders_passed_to_fit(tmpdir):
model, hparams = tutils.get_default_model()
trainer_options = dict(default_save_path=tmpdir,
show_progress_bar=False,
progress_bar_refresh_rate=0,
max_epochs=1,
train_percent_check=0.4,
val_percent_check=0.2,
@@ -165,7 +164,7 @@ def test_multi_gpu_none_backend(tmpdir):
model, hparams = tutils.get_default_model()
trainer_options = dict(
default_save_path=tmpdir,
show_progress_bar=False,
progress_bar_refresh_rate=0,
max_epochs=1,
train_percent_check=0.1,
val_percent_check=0.1,
@@ -184,7 +183,7 @@ def test_multi_gpu_model_dp(tmpdir):
model, hparams = tutils.get_default_model()
trainer_options = dict(
default_save_path=tmpdir,
show_progress_bar=False,
progress_bar_refresh_rate=0,
distributed_backend='dp',
max_epochs=1,
train_percent_check=0.1,
tests/models/test_restore.py
+4 -6
View File
@@ -33,7 +33,7 @@ def test_running_test_pretrained_model_ddp(tmpdir):
checkpoint = tutils.init_checkpoint_callback(logger)
trainer_options = dict(
show_progress_bar=False,
progress_bar_refresh_rate=0,
max_epochs=1,
train_percent_check=0.4,
val_percent_check=0.2,
@@ -81,7 +81,7 @@ def test_running_test_pretrained_model(tmpdir):
checkpoint = tutils.init_checkpoint_callback(logger)
trainer_options = dict(
show_progress_bar=False,
progress_bar_refresh_rate=0,
max_epochs=4,
train_percent_check=0.4,
val_percent_check=0.2,
@@ -114,7 +114,7 @@ def test_load_model_from_checkpoint(tmpdir):
model = LightningTestModel(hparams)
trainer_options = dict(
show_progress_bar=False,
progress_bar_refresh_rate=0,
max_epochs=2,
train_percent_check=0.4,
val_percent_check=0.2,
@@ -165,7 +165,6 @@ def test_running_test_pretrained_model_dp(tmpdir):
checkpoint = tutils.init_checkpoint_callback(logger)
trainer_options = dict(
show_progress_bar=True,
max_epochs=2,
train_percent_check=0.4,
val_percent_check=0.2,
@@ -202,7 +201,6 @@ def test_dp_resume(tmpdir):
model = LightningTestModel(hparams)
trainer_options = dict(
show_progress_bar=True,
max_epochs=1,
gpus=2,
distributed_backend='dp',
@@ -329,7 +327,7 @@ def test_model_saving_loading(tmpdir):
def test_load_model_with_missing_hparams(tmpdir):
trainer_options = dict(
show_progress_bar=False,
progress_bar_refresh_rate=0,
max_epochs=1,
checkpoint_callback=ModelCheckpoint(tmpdir, save_top_k=-1),
logger=False,
tests/test_deprecated.py
+9
View File
@@ -48,6 +48,15 @@ def test_tbd_remove_in_v0_8_0_trainer():
'Wrongly passed deprecated argument "%s" to attribute "%s"' % (attr_old, attr_new)
def test_tbd_remove_in_v0_9_0_trainer():
# test show_progress_bar set by progress_bar_refresh_rate
trainer = Trainer(progress_bar_refresh_rate=0, show_progress_bar=True)
assert not getattr(trainer, 'show_progress_bar')
trainer = Trainer(progress_bar_refresh_rate=50, show_progress_bar=False)
assert getattr(trainer, 'show_progress_bar')
def test_tbd_remove_in_v0_9_0_module_imports():
from pytorch_lightning.core.decorators import data_loader # noqa: F811
tests/trainer/test_callbacks.py
+1 -1
View File
@@ -99,7 +99,7 @@ def test_trainer_callback_system(tmpdir):
'max_epochs': 1,
'val_percent_check': 0.1,
'train_percent_check': 0.2,
'show_progress_bar': False
'progress_bar_refresh_rate': 0
}
assert not test_callback.on_init_start_called
tests/trainer/test_trainer.py
+1 -1
View File
@@ -332,7 +332,7 @@ def test_resume_from_checkpoint_epoch_restored(tmpdir):
model = _new_model()
trainer_options = dict(
show_progress_bar=False,
progress_bar_refresh_rate=0,
max_epochs=2,
train_percent_check=0.65,
val_percent_check=1,