search

AutodeskAILab / UV-Net

Code for UV-Net: Learning from Boundary Representations, CVPR 2021.
MIT License
82 stars 14 forks source link

Package issues in pytorch lightning and torchmetrics leads to error. #11

Closed Ma-Weijian closed 1 year ago

Ma-Weijian commented 2 years ago

Hi Autodesk AI Lab.

Anaconda chooses torchmetrics 0.9.0 for me. And as the title shows, such api no longer exists in future versions of torchmetrics.

I believe there are newer apis in torchmetrics to replace the old one and I wonder what it is.

Any idea helps.

Thanks and best regards.

Anderson Ma

Ma-Weijian commented 2 years ago

Also, pytorch lightning 1.5 and later reports a bug that it cannot judge the batch size in self.log of class segmentation.

The error log is like

Traceback (most recent call last):
  File "/root/UV-Net/segmentation.py", line 117, in <module>
    trainer.fit(model, train_loader, val_loader)
  File "/usr/local/miniconda3/envs/uvnet/lib/python3.9/site-packages/pytorch_lightning/trainer/trainer.py", line 771, in fit
    self._call_and_handle_interrupt(
  File "/usr/local/miniconda3/envs/uvnet/lib/python3.9/site-packages/pytorch_lightning/trainer/trainer.py", line 724, in _call_and_handle_interrupt
    return trainer_fn(*args, **kwargs)
  File "/usr/local/miniconda3/envs/uvnet/lib/python3.9/site-packages/pytorch_lightning/trainer/trainer.py", line 812, in _fit_impl
    results = self._run(model, ckpt_path=self.ckpt_path)
  File "/usr/local/miniconda3/envs/uvnet/lib/python3.9/site-packages/pytorch_lightning/trainer/trainer.py", line 1237, in _run
    results = self._run_stage()
  File "/usr/local/miniconda3/envs/uvnet/lib/python3.9/site-packages/pytorch_lightning/trainer/trainer.py", line 1324, in _run_stage
    return self._run_train()
  File "/usr/local/miniconda3/envs/uvnet/lib/python3.9/site-packages/pytorch_lightning/trainer/trainer.py", line 1346, in _run_train
    self._run_sanity_check()
  File "/usr/local/miniconda3/envs/uvnet/lib/python3.9/site-packages/pytorch_lightning/trainer/trainer.py", line 1414, in _run_sanity_check
    val_loop.run()
  File "/usr/local/miniconda3/envs/uvnet/lib/python3.9/site-packages/pytorch_lightning/loops/base.py", line 204, in run
    self.advance(*args, **kwargs)
  File "/usr/local/miniconda3/envs/uvnet/lib/python3.9/site-packages/pytorch_lightning/loops/dataloader/evaluation_loop.py", line 153, in advance
    dl_outputs = self.epoch_loop.run(self._data_fetcher, dl_max_batches, kwargs)
  File "/usr/local/miniconda3/envs/uvnet/lib/python3.9/site-packages/pytorch_lightning/loops/base.py", line 204, in run
    self.advance(*args, **kwargs)
  File "/usr/local/miniconda3/envs/uvnet/lib/python3.9/site-packages/pytorch_lightning/loops/epoch/evaluation_epoch_loop.py", line 127, in advance
    output = self._evaluation_step(**kwargs)
  File "/usr/local/miniconda3/envs/uvnet/lib/python3.9/site-packages/pytorch_lightning/loops/epoch/evaluation_epoch_loop.py", line 222, in _evaluation_step
    output = self.trainer._call_strategy_hook("validation_step", *kwargs.values())
  File "/usr/local/miniconda3/envs/uvnet/lib/python3.9/site-packages/pytorch_lightning/trainer/trainer.py", line 1766, in _call_strategy_hook
    output = fn(*args, **kwargs)
  File "/usr/local/miniconda3/envs/uvnet/lib/python3.9/site-packages/pytorch_lightning/strategies/strategy.py", line 344, in validation_step
    return self.model.validation_step(*args, **kwargs)
  File "/root/UV-Net/uvnet/models.py", line 326, in validation_step
    self.log("val_loss", loss, on_step=False, on_epoch=True)
  File "/usr/local/miniconda3/envs/uvnet/lib/python3.9/site-packages/pytorch_lightning/core/lightning.py", line 417, in log
    results.log(
  File "/usr/local/miniconda3/envs/uvnet/lib/python3.9/site-packages/pytorch_lightning/trainer/connectors/logger_connector/result.py", line 493, in log
    batch_size = self._extract_batch_size(self[key], batch_size, meta)
  File "/usr/local/miniconda3/envs/uvnet/lib/python3.9/site-packages/pytorch_lightning/trainer/connectors/logger_connector/result.py", line 428, in _extract_batch_size
    batch_size = extract_batch_size(self.batch)
  File "/usr/local/miniconda3/envs/uvnet/lib/python3.9/site-packages/pytorch_lightning/utilities/data.py", line 81, in extract_batch_size
    raise MisconfigurationException(error_msg)
pytorch_lightning.utilities.exceptions.MisconfigurationException: We could not infer the batch_size from the batch. Either simplify its structure or provide the batch_size as `self.log(..., batch_size=batch_size)`.

Maybe it is necessary to change the code to fit the newer APIs.

pradeep-pyro commented 2 years ago

Thanks for reporting this. I think an easier and more stable fix would be for us to include the exact version number for PyTorch Lightning and torchmetrics in the environment.yml file. That way we can guarantee that the code works without issues when API changes in future. I will fix this.

pradeep-pyro commented 2 years ago

Hi, I added version numbers for PyTorch Lightning and torchmetrics in the environment yml file. I was able to create a new environment from scratch and run the code without any issues. Can you check if my fixes in the package_versions branch work for you?