search

JinfengX / CasFusionNet

CasFusionNet: A Cascaded Network for Point Cloud Semantic Scene Completion by Dense Feature Fusion [AAAI 2023]
19 stars 2 forks source link

LRscheduler ERROR: The provided lr scheduler `StepLR` doesn't follow PyTorch's LRScheduler API. #2

Open huangjund opened 3 months ago

huangjund commented 3 months ago

Below is the training log. The error shows: The provided lr scheduler StepLR doesn't follow PyTorch's LRScheduler API. You should override the LightningModule.lr_scheduler_step hook with your own logic if you are using a custom LR scheduler.

[2024-07-04 20:05:01,066][src.utils.utils][INFO] - Enforcing tags! [2024-07-04 20:05:01,084][src.utils.utils][INFO] - Printing config tree with Rich! [2024-07-04 20:05:02,466][main][INFO] - Instantiating datamodule [2024-07-04 20:05:04,850][main][INFO] - Instantiating model [2024-07-04 20:05:41,272][main][INFO] - Instantiating callbacks... [2024-07-04 20:05:41,273][src.utils.utils][INFO] - Instantiating callback [2024-07-04 20:05:41,284][src.utils.utils][INFO] - Instantiating callback [2024-07-04 20:05:41,286][src.utils.utils][INFO] - Instantiating callback [2024-07-04 20:05:41,287][main][INFO] - Instantiating loggers... [2024-07-04 20:05:41,288][src.utils.utils][INFO] - Instantiating logger [2024-07-04 20:05:41,293][src.utils.utils][INFO] - Instantiating logger [2024-07-04 20:05:41,298][main][INFO] - Instantiating trainer [2024-07-04 20:05:41,316][main][INFO] - Logging hyperparameters! [2024-07-04 20:05:41,371][main][INFO] - Starting training! [2024-07-04 20:05:41,700][root][INFO] - Train dataset loaded, size: 1543 [2024-07-04 20:05:41,914][root][INFO] - Validate dataset loaded, size: 398 [2024-07-04 20:05:42,152][src.utils.utils][ERROR] - Traceback (most recent call last): File "/home/jd/PointCompletion/CasFusionNet/src/utils/utils.py", line 38, in wrap metric_dict, object_dict = task_func(cfg=cfg) File "/home/jd/PointCompletion/CasFusionNet/src/train.py", line 84, in train trainer.fit(model=model, datamodule=datamodule, ckpt_path=cfg.get("ckpt_path")) File "/home/jd/miniconda3/envs/pointcomplt/lib/python3.8/site-packages/pytorch_lightning/trainer/trainer.py", line 770, in fit self._call_and_handle_interrupt( File "/home/jd/miniconda3/envs/pointcomplt/lib/python3.8/site-packages/pytorch_lightning/trainer/trainer.py", line 723, in _call_and_handle_interrupt return trainer_fn(*args, **kwargs) File "/home/jd/miniconda3/envs/pointcomplt/lib/python3.8/site-packages/pytorch_lightning/trainer/trainer.py", line 811, in _fit_impl results = self._run(model, ckpt_path=self.ckpt_path) File "/home/jd/miniconda3/envs/pointcomplt/lib/python3.8/site-packages/pytorch_lightning/trainer/trainer.py", line 1217, in _run self.strategy.setup(self) File "/home/jd/miniconda3/envs/pointcomplt/lib/python3.8/site-packages/pytorch_lightning/strategies/single_device.py", line 72, in setup super().setup(trainer) File "/home/jd/miniconda3/envs/pointcomplt/lib/python3.8/site-packages/pytorch_lightning/strategies/strategy.py", line 139, in setup self.setup_optimizers(trainer) File "/home/jd/miniconda3/envs/pointcomplt/lib/python3.8/site-packages/pytorch_lightning/strategies/strategy.py", line 128, in setup_optimizers self.optimizers, self.lr_scheduler_configs, self.optimizer_frequencies = _init_optimizers_and_lr_schedulers( File "/home/jd/miniconda3/envs/pointcomplt/lib/python3.8/site-packages/pytorch_lightning/core/optimizer.py", line 195, in _init_optimizers_and_lr_schedulers _validate_scheduler_api(lr_scheduler_configs, model) File "/home/jd/miniconda3/envs/pointcomplt/lib/python3.8/site-packages/pytorch_lightning/core/optimizer.py", line 350, in _validate_scheduler_api raise MisconfigurationException( pytorch_lightning.utilities.exceptions.MisconfigurationException: The provided lr scheduler StepLR doesn't follow PyTorch's LRScheduler API. You should override the LightningModule.lr_scheduler_step hook with your own logic if you are using a custom LR scheduler. [2024-07-04 20:05:42,156][src.utils.utils][INFO] - Output dir: /home/jd/PointCompletion/CasFusionNet/logs/train_ssc_pc/runs/2024-07-04_20-05-00 [2024-07-04 20:05:42,156][src.utils.utils][INFO] - Closing loggers...

I checked the error place: if not isinstance(scheduler, LRSchedulerTypeTuple) and not is_overridden("lr_scheduler_step", model): raise MisconfigurationException( f"The provided lr scheduler{scheduler.class.name}doesn't follow PyTorch's LRScheduler" " API. You should override theLightningModule.lr_scheduler_stephook with your own logic if" " you are using a custom LR scheduler." )

The type of scheduler is StepLR. However, the LRSchedulerTypeTuple has only two types in the torch_lightning: LRSchedulerTypeTuple = (torch.optim.lr_scheduler._LRScheduler, torch.optim.lr_scheduler.ReduceLROnPlateau). That's why the error raised.

How could I solve this problem?

huangjund commented 3 months ago

It's a version confliction problem. Can be solved in this pull request: https://github.com/JinfengX/CasFusionNet/pull/3.