segmentation-tutorial.ipynb crashes without Apex #437
Description
Describe the bug
In examples/notebooks/segmentation-tutorial.ipynb, if you skip NVIDIA Apex installation (as per the comment in the first code cell saying if Your machine doesn't support FP16, comment this lines), the notebook will crash on runner.train(...).
To Reproduce
Steps to reproduce the behavior:
- Comment out
!git clone https://github.com/NVIDIA/apex
!pip install -v --no-cache-dir --global-option="--cpp_ext" --global-option="--cuda_ext" ./apex
- Run all cells.
Expected behavior
Successful execution.
Additional context
Traceback (tested on a Colab instance with K80):
---------------------------------------------------------------------------
ModuleNotFoundError Traceback (most recent call last)
/usr/local/lib/python3.6/dist-packages/catalyst/utils/torch.py in assert_fp16_available()
105 try:
--> 106 __import__("apex")
107 except ImportError:
ModuleNotFoundError: No module named 'apex'
During handling of the above exception, another exception occurred:
AssertionError Traceback (most recent call last)
8 frames
/usr/local/lib/python3.6/dist-packages/catalyst/dl/core/runner.py in run_experiment(self, experiment, check)
227 for stage in self.experiment.stages:
--> 228 self._run_stage(stage)
229 except (Exception, KeyboardInterrupt) as ex:
/usr/local/lib/python3.6/dist-packages/catalyst/dl/core/runner.py in _run_stage(self, stage)
191 def _run_stage(self, stage: str):
--> 192 self._prepare_for_stage(stage)
193 loaders = self.experiment.get_loaders(stage)
/usr/local/lib/python3.6/dist-packages/catalyst/dl/core/runner.py in _prepare_for_stage(self, stage)
78 self.model, criterion, optimizer, scheduler, self.device = \
---> 79 self._get_experiment_components(stage)
80
/usr/local/lib/python3.6/dist-packages/catalyst/dl/core/runner.py in _get_experiment_components(self, stage)
61 scheduler=scheduler,
---> 62 distributed_params=self.experiment.distributed_params
63 )
/usr/local/lib/python3.6/dist-packages/catalyst/dl/utils/torch.py in process_components(model, criterion, optimizer, scheduler, distributed_params)
34 assert isinstance(model, nn.Module)
---> 35 utils.assert_fp16_available()
36 from apex import amp
/usr/local/lib/python3.6/dist-packages/catalyst/utils/torch.py in assert_fp16_available()
108 assert False, \
--> 109 "NVidia Apex package must be installed. " \
110 "See https://github.com/NVIDIA/apex."
AssertionError: NVidia Apex package must be installed. See https://github.com/NVIDIA/apex.
During handling of the above exception, another exception occurred:
AttributeError Traceback (most recent call last)
<ipython-input-25-ee9db9bd610d> in <module>()
54
55 # prints train logs
---> 56 verbose=True
57 )
/usr/local/lib/python3.6/dist-packages/catalyst/dl/runner/supervised.py in train(self, model, criterion, optimizer, loaders, logdir, callbacks, scheduler, resume, num_epochs, valid_loader, main_metric, minimize_metric, verbose, state_kwargs, checkpoint_data, fp16, monitoring_params, check)
195 monitoring_params=monitoring_params
196 )
--> 197 self.run_experiment(experiment, check=check)
198
199 def infer(
/usr/local/lib/python3.6/dist-packages/catalyst/dl/core/runner.py in run_experiment(self, experiment, check)
228 self._run_stage(stage)
229 except (Exception, KeyboardInterrupt) as ex:
--> 230 self.state.exception = ex
231 self._run_event("exception")
232
AttributeError: 'NoneType' object has no attribute 'exception'