Support latest PyTorch lightning #3418
Description
Expected behavior
Optuna works with the latest PyTorch-Lightning (1.6.0).
Environment
- Optuna version: 3.0.0b0.dev
- Python version:
- OS: Independent
- (Optional) Other libraries and their versions: PyTorch-Lightning 1.6.0
Error messages, stack traces, or logs
tests/integration_tests/test_pytorch_lightning.py:158:
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
optuna/study/study.py:399: in optimize
_optimize(
optuna/study/_optimize.py:68: in _optimize
_optimize_sequential(
optuna/study/_optimize.py:162: in _optimize_sequential
trial = _run_trial(study, func, catch)
optuna/study/_optimize.py:262: in _run_trial
raise func_err
optuna/study/_optimize.py:211: in _run_trial
value_or_values = func(trial)
tests/integration_tests/test_pytorch_lightning.py:143: in objective
trainer = pl.Trainer(
/opt/hostedtoolcache/Python/3.8.12/x64/lib/python3.8/site-packages/pytorch_lightning/utilities/argparse.py:339: in insert_env_defaults
return fn(self, **kwargs)
/opt/hostedtoolcache/Python/3.8.12/x64/lib/python3.8/site-packages/pytorch_lightning/trainer/trainer.py:561: in __init__
self._call_callback_hooks("on_init_start")
/opt/hostedtoolcache/Python/3.8.12/x64/lib/python3.8/site-packages/pytorch_lightning/trainer/trainer.py:1617: in _call_callback_hooks
fn(self, *args, **kwargs)
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
self = <optuna.integration.pytorch_lightning.PyTorchLightningPruningCallback object at 0x7f892c1c5730>
trainer = <pytorch_lightning.trainer.trainer.Trainer object at 0x7f892c1c5040>
def on_init_start(self, trainer: Trainer) -> None:
> self.is_ddp_backend = trainer._accelerator_connector.distributed_backend is not None
E AttributeError: 'AcceleratorConnector' object has no attribute 'distributed_backend'
optuna/integration/pytorch_lightning.py:60: AttributeErrorhttps://github.com/optuna/optuna/runs/5745775785?check_suite_focus=true
tests/integration_tests/test_pytorch_lightning.py:6: in <module>
import pytorch_lightning as pl
../../../hostedtoolcache/Python/3.8.12/x64/lib/python3.8/site-packages/pytorch_lightning/__init__.py:30: in <module>
from pytorch_lightning.callbacks import Callback # noqa: E402
../../../hostedtoolcache/Python/3.8.12/x64/lib/python3.8/site-packages/pytorch_lightning/callbacks/__init__.py:26: in <module>
from pytorch_lightning.callbacks.pruning import ModelPruning
../../../hostedtoolcache/Python/3.8.12/x64/lib/python3.8/site-packages/pytorch_lightning/callbacks/pruning.py:31: in <module>
from pytorch_lightning.core.lightning import LightningModule
../../../hostedtoolcache/Python/3.8.12/x64/lib/python3.8/site-packages/pytorch_lightning/core/__init__.py:16: in <module>
from pytorch_lightning.core.lightning import LightningModule
../../../hostedtoolcache/Python/3.8.12/x64/lib/python3.8/site-packages/pytorch_lightning/core/lightning.py:40: in <module>
from pytorch_lightning.loggers import LightningLoggerBase
../../../hostedtoolcache/Python/3.8.12/x64/lib/python3.8/site-packages/pytorch_lightning/loggers/__init__.py:18: in <module>
from pytorch_lightning.loggers.tensorboard import TensorBoardLogger
../../../hostedtoolcache/Python/3.8.12/x64/lib/python3.8/site-packages/pytorch_lightning/loggers/tensorboard.py:26: in <module>
from torch.utils.tensorboard import SummaryWriter
../../../hostedtoolcache/Python/3.8.12/x64/lib/python3.8/site-packages/torch/utils/tensorboard/__init__.py:4: in <module>
LooseVersion = distutils.version.LooseVersion
E AttributeError: module 'distutils' has no attribute 'version'
https://github.com/optuna/optuna/runs/5745734509?check_suite_focus=true
Steps to reproduce
See our CI failures.
Additional context (optional)
It may be the simplest way to support PyTorch v1.11.0 (Lightning-AI/pytorch-lightning#12324).
🔗 #3417