Handle KeyboardInterrupt during training (#2134)

* Handle KeyboardInterrupt during training

Fixes #2079.

* chlog

* Fix whitespace

* Update callback_hook.py

* Update base.py

* Update training_loop.py

* Update test_trainer.py

* Update CHANGELOG.md

Co-authored-by: Adrian Wälchli <aedu.waelchli@gmail.com>

* Update CHANGELOG.md

* on_keyboard_interrupt

Co-authored-by: Jirka <jirka@pytorchlightning.ai>
Co-authored-by: William Falcon <waf2107@columbia.edu>
Co-authored-by: Jirka Borovec <Borda@users.noreply.github.com>
Co-authored-by: Adrian Wälchli <aedu.waelchli@gmail.com>

CHANGELOG.md

@@ -21,7 +21,6 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 
 ### Added
 
-- Remove explicit flush from tensorboard logger ([#2126](https://github.com/PyTorchLightning/pytorch-lightning/pull/2126))
 - Added metric Base classes ([#1326](https://github.com/PyTorchLightning/pytorch-lightning/pull/1326), [#1877](https://github.com/PyTorchLightning/pytorch-lightning/pull/1877))
 - Added Sklearn metrics classes ([#1327](https://github.com/PyTorchLightning/pytorch-lightning/pull/1327))
 - Added Native torch metrics ([#1488](https://github.com/PyTorchLightning/pytorch-lightning/pull/1488))
@@ -35,6 +34,7 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Added [black](https://black.readthedocs.io/en/stable/) formatter for the code with code-checker on pull ([1610](https://github.com/PyTorchLightning/pytorch-lightning/pull/1610))
 - Added back the slow spawn ddp implementation as `ddp_spawn` ([#2115](https://github.com/PyTorchLightning/pytorch-lightning/pull/2115))
 - Added loading checkpoints from URLs ([#1667](https://github.com/PyTorchLightning/pytorch-lightning/issues/1667))
+- Added a callback method `on_keyboard_interrupt` for handling KeyboardInterrupt events during training ([#2134](https://github.com/PyTorchLightning/pytorch-lightning/pull/2134))
 
 ### Changed
 
@@ -46,6 +46,7 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Changed the default value of the Trainer argument `weights_summary` from `full` to `top` ([#2029](https://github.com/PyTorchLightning/pytorch-lightning/pull/2029))
 - Raise an error when lightning replaces an existing sampler ([#2020](https://github.com/PyTorchLightning/pytorch-lightning/pull/2020))
 - Enabled prepare_data from correct processes - clarify local vs global rank ([#2166](https://github.com/PyTorchLightning/pytorch-lightning/pull/2166))
+- Remove explicit flush from tensorboard logger ([#2126](https://github.com/PyTorchLightning/pytorch-lightning/pull/2126))
 
 ### Deprecated
 

pytorch_lightning/callbacks/base.py

@@ -85,3 +85,6 @@ class Callback(abc.ABC):
     def on_test_end(self, trainer, pl_module):
         """Called when the test ends."""
         pass
+
+    def on_keyboard_interrupt(self, trainer, pl_module):
+        """Called when the training is interrupted by KeyboardInterrupt."""

pytorch_lightning/trainer/callback_hook.py

@@ -100,3 +100,8 @@ class TrainerCallbackHookMixin(ABC):
         """Called when the test ends."""
         for callback in self.callbacks:
             callback.on_test_end(self, self.get_model())
+
+    def on_keyboard_interrupt(self):
+        """Called when the training is interrupted by KeyboardInterrupt."""
+        for callback in self.callbacks:
+            callback.on_keyboard_interrupt(self, self.get_model())

pytorch_lightning/trainer/training_loop.py

@@ -237,6 +237,7 @@ class TrainerTrainLoopMixin(ABC):
     checkpoint_callback: ...
     terminate_on_nan: bool
     tpu_id: int
+    interactive_ddp_procs: ...
 
     # Callback system
     callbacks: List[Callback]
@@ -247,6 +248,7 @@ class TrainerTrainLoopMixin(ABC):
     on_epoch_start: Callable
     on_epoch_end: Callable
     on_validation_end: Callable
+    on_keyboard_interrupt: Callable
 
     @abstractmethod
     def get_model(self) -> LightningModule:
@@ -395,6 +397,7 @@ class TrainerTrainLoopMixin(ABC):
             # user could press ctrl+c many times... only shutdown once
             if not self.interrupted:
                 self.interrupted = True
+                self.on_keyboard_interrupt()
 
                 for proc in self.interactive_ddp_procs:
                     subprocess.Popen.kill(proc)

tests/trainer/test_trainer.py

@@ -3,6 +3,7 @@ import math
 import os
 import pickle
 import types
+import sys
 from argparse import Namespace
 
 import cloudpickle
@@ -10,10 +11,10 @@ import pytest
 import torch
 
 import tests.base.utils as tutils
-from pytorch_lightning import Callback, LightningModule
-from pytorch_lightning import Trainer
+from pytorch_lightning import Callback, LightningModule, Trainer
 from pytorch_lightning.callbacks import EarlyStopping, ModelCheckpoint
-from pytorch_lightning.core.saving import load_hparams_from_tags_csv, load_hparams_from_yaml, save_hparams_to_tags_csv
+from pytorch_lightning.core.saving import (
+    load_hparams_from_tags_csv, load_hparams_from_yaml, save_hparams_to_tags_csv)
 from pytorch_lightning.loggers import TensorBoardLogger
 from pytorch_lightning.trainer.logging import TrainerLoggingMixin
 from pytorch_lightning.utilities.io import load as pl_load
@@ -660,10 +661,19 @@ def test_trainer_interrupted_flag(tmpdir):
         def on_batch_start(self, trainer, pl_module):
             raise KeyboardInterrupt
 
+    class HandleInterruptCallback(Callback):
+        def __init__(self):
+            super().__init__()
+            self.exc_info = None
+
+        def on_keyboard_interrupt(self, trainer, pl_module):
+            self.exc_info = sys.exc_info()
+
     interrupt_callback = InterruptCallback()
+    handle_interrupt_callback = HandleInterruptCallback()
 
     trainer = Trainer(
-        callbacks=[interrupt_callback],
+        callbacks=[interrupt_callback, handle_interrupt_callback],
         max_epochs=1,
         val_percent_check=0.1,
         train_percent_check=0.2,
@@ -672,8 +682,10 @@ def test_trainer_interrupted_flag(tmpdir):
         default_root_dir=tmpdir,
     )
     assert not trainer.interrupted
+    assert handle_interrupt_callback.exc_info is None
     trainer.fit(model)
     assert trainer.interrupted
+    assert isinstance(handle_interrupt_callback.exc_info[1], KeyboardInterrupt)
 
 
 def test_gradient_clipping(tmpdir):