consolidate callbacks and hooks (#950)

* consolidate callbacks and hooks

* ensure callbacks recieve proper arg types

* remove model from init callback events

* clean up early stopping event

* update changelog

* remove on_fit_start and on_fit_end

* fix args for on_init_start and on_init_end

* handle case where early stopping is not used

* show all callback methods

* wrap checkpoint callback logic into proper class

* fix check for main process in checkpoint callback

* move callbacks test to separate file

* refactor arg checks

* get model and call hook on same line

* define trainer_options dict in one call

* add more asserts to callback test

Author: jeremyjordan

CHANGELOG.md

@@ -22,6 +22,7 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Added Tensor Processing Unit (TPU) support ([#868](https://github.com/PyTorchLightning/pytorch-lightning/pull/868))
 - Added semantic segmentation example ([#751](https://github.com/PyTorchLightning/pytorch-lightning/pull/751),[#876](https://github.com/PyTorchLightning/pytorch-lightning/pull/876))
 - Split callbacks in multiple files ([#849](https://github.com/PyTorchLightning/pytorch-lightning/pull/849))
+- Support for user defined callbacks ([#889](https://github.com/PyTorchLightning/pytorch-lightning/pull/889) and [#950](https://github.com/PyTorchLightning/pytorch-lightning/pull/950))
 - Added support for multiple loggers to be passed to `Trainer` as an iterable (e.g. list, tuple, etc.) ([#903](https://github.com/PyTorchLightning/pytorch-lightning/pull/903))
 
 ### Changed

docs/source/callbacks.rst

@@ -44,8 +44,4 @@ Callback Class
         _del_model,
         _save_model,
         _abc_impl,
-        on_epoch_end,
-        on_train_end,
-        on_epoch_start,
         check_monitor_top_k,
-        on_train_start,

pytorch_lightning/callbacks/base.py

@@ -12,19 +12,11 @@ class Callback(abc.ABC):
     """Abstract base class used to build new callbacks."""
 
     def on_init_start(self, trainer):
-        """Called when the trainer initialization begins."""
+        """Called when the trainer initialization begins, model has not yet been set."""
         pass
 
     def on_init_end(self, trainer):
-        """Called when the trainer initialization ends."""
-        pass
-
-    def on_fit_start(self, trainer, pl_module):
-        """Called when the fit begins."""
-        pass
-
-    def on_fit_end(self, trainer, pl_module):
-        """Called when the fit ends."""
+        """Called when the trainer initialization ends, model has not yet been set."""
         pass
 
     def on_epoch_start(self, trainer, pl_module):

pytorch_lightning/callbacks/early_stopping.py

@@ -64,8 +64,6 @@ def __init__(self, monitor: str = 'val_loss', min_delta: float = 0.0, patience:
         self.monitor_op = mode_dict[mode]
         self.min_delta *= 1 if self.monitor_op == np.greater else -1
 
-        self.on_train_start(None, None)
-
     def check_metrics(self, logs):
         monitor_val = logs.get(self.monitor)
         error_msg = (f'Early stopping conditioned on metric `{self.monitor}`'

pytorch_lightning/callbacks/model_checkpoint.py

@@ -118,6 +118,10 @@ def check_monitor_top_k(self, current):
         return self.monitor_op(current, self.best_k_models[self.kth_best_model])
 
     def on_validation_end(self, trainer, pl_module):
+        # only run on main process
+        if trainer.proc_rank != 0:
+            return
+
         logs = trainer.callback_metrics
         epoch = trainer.current_epoch
         self.epochs_since_last_check += 1

pytorch_lightning/trainer/callback_hook.py

@@ -12,25 +12,15 @@ def __init__(self):
         self.callbacks: list[Callback] = []
         self.get_model: Callable = ...
 
-    def on_init_start(self, trainer):
-        """Called when the trainer initialization begins."""
+    def on_init_start(self):
+        """Called when the trainer initialization begins, model has not yet been set."""
         for callback in self.callbacks:
-            callback.on_init_start(trainer)
+            callback.on_init_start(self)
 
-    def on_init_end(self, trainer):
-        """Called when the trainer initialization ends."""
+    def on_init_end(self):
+        """Called when the trainer initialization ends, model has not yet been set."""
         for callback in self.callbacks:
-            callback.on_init_end(trainer)
-
-    def on_fit_start(self):
-        """Called when the fit begins."""
-        for callback in self.callbacks:
-            callback.on_fit_start(self, self.get_model())
-
-    def on_fit_end(self):
-        """Called when the fit ends."""
-        for callback in self.callbacks:
-            callback.on_fit_end(self, self.get_model())
+            callback.on_init_end(self)
 
     def on_epoch_start(self):
         """Called when the epoch begins."""

pytorch_lightning/trainer/evaluation_loop.py

@@ -374,14 +374,13 @@ def run_evaluation(self, test_mode: bool = False):
         else:
             self.val_progress_bar.close()
 
-        # model checkpointing
-        if self.proc_rank == 0 and self.checkpoint_callback is not None and not test_mode:
-            self.checkpoint_callback.on_validation_end(self, self.get_model())
-
         # Validation/Test end callbacks
         if test_mode:
             self.on_test_end()
         else:
+            # model checkpointing
+            if self.checkpoint_callback is not None:
+                self.checkpoint_callback.on_validation_end(self, self.get_model())
             self.on_validation_end()
 
     def evaluation_forward(self, model, batch, batch_idx, dataloader_idx, test_mode: bool = False):

pytorch_lightning/trainer/trainer.py

@@ -618,7 +618,7 @@ def on_train_end(self):
 
         # Init callbacks
         self.callbacks = callbacks
-        self.on_init_start(self)
+        self.on_init_start()
 
         # benchmarking
         self.benchmark = benchmark
@@ -808,7 +808,7 @@ def on_train_end(self):
         self.init_amp(use_amp)
 
         # Callback system
-        self.on_init_end(self)
+        self.on_init_end()
 
     @property
     def slurm_job_id(self) -> int:
@@ -941,9 +941,6 @@ def fit(
         # bind logger
         model.logger = self.logger
 
-        # Fit begin callbacks
-        self.on_fit_start()
-
         # set up the passed in dataloaders (if needed)
         self.__set_fit_dataloaders(model, train_dataloader, val_dataloaders, test_dataloaders)
 
@@ -1006,9 +1003,6 @@ def fit(
 
             self.run_pretrain_routine(model)
 
-        # Fit end callbacks
-        self.on_fit_end()
-
         # return 1 when finished
         # used for testing or when we need to know that training succeeded
         return 1

pytorch_lightning/trainer/training_loop.py

@@ -302,9 +302,15 @@ def train(self):
         self.reset_train_dataloader(model)
         self.reset_val_dataloader(model)
 
-        # Train begin callbacks
-        model.on_train_start()
-        self.on_train_start()
+        # Train start events
+        with self.profiler.profile('on_train_start'):
+            # callbacks
+            self.on_train_start()
+            # initialize early stop callback
+            if self.early_stop_callback is not None:
+                self.early_stop_callback.on_train_start(self, self.get_model())
+            # model hooks
+            model.on_train_start()
 
         try:
             # run all epochs
@@ -347,9 +353,6 @@ def train(self):
                 desc = f'Epoch {epoch + 1}' if not self.is_infinite_dataloader(self.train_dataloader) else ''
                 self.main_progress_bar.set_description(desc)
 
-                # changing gradient according accumulation_scheduler
-                self.accumulation_scheduler.on_epoch_start(self, self.get_model())
-
                 # -----------------
                 # RUN TNG EPOCH
                 # -----------------
@@ -369,23 +372,21 @@ def train(self):
                     self.reduce_lr_on_plateau_scheduler.step(val_loss)
 
                 if self.max_steps and self.max_steps == self.global_step:
-                    self.main_progress_bar.close()
-                    model.on_train_end()
-                    self.on_train_end()
+                    self.run_training_teardown()
                     return
 
                 # early stopping
                 met_min_epochs = epoch >= self.min_epochs - 1
                 met_min_steps = self.global_step >= self.min_steps if self.min_steps else True
 
+                # TODO wrap this logic into the callback
                 if self.enable_early_stop and not self.disable_validation and is_val_epoch:
                     if ((met_min_epochs and met_min_steps) or self.fast_dev_run):
                         should_stop = self.early_stop_callback.on_epoch_end(self, self.get_model())
                         # stop training
                         stop = should_stop and met_min_epochs
                         if stop:
                             self.run_training_teardown()
-                            self.on_train_end()
                             return
 
             self.run_training_teardown()
@@ -394,19 +395,17 @@ def train(self):
             log.info('Detected KeyboardInterrupt, attempting graceful shutdown...')
             self.run_training_teardown()
 
-        # Train end callbacks
-        self.on_train_end()
-
     def run_training_epoch(self):
 
-        # Epoch begin callbacks
-        self.on_epoch_start()
-
-        # before epoch hook
-        if self.is_function_implemented('on_epoch_start'):
-            model = self.get_model()
-            with self.profiler.profile('on_epoch_start'):
-                model.on_epoch_start()
+        # Epoch start events
+        with self.profiler.profile('on_epoch_start'):
+            # callbacks
+            self.on_epoch_start()
+            # changing gradient according accumulation_scheduler
+            self.accumulation_scheduler.on_epoch_start(self, self.get_model())
+            # model hooks
+            if self.is_function_implemented('on_epoch_start'):
+                self.get_model().on_epoch_start()
 
         # reset train dataloader
         if self.reload_dataloaders_every_epoch:
@@ -485,14 +484,13 @@ def run_training_epoch(self):
             if early_stop_epoch or self.fast_dev_run:
                 break
 
-        # epoch end hook
-        if self.is_function_implemented('on_epoch_end'):
-            model = self.get_model()
-            with self.profiler.profile('on_epoch_end'):
-                model.on_epoch_end()
-
-        # Epoch begin callbacks
-        self.on_epoch_end()
+        # Epoch end events
+        with self.profiler.profile('on_epoch_end'):
+            # callbacks
+            self.on_epoch_end()
+            # model hooks
+            if self.is_function_implemented('on_epoch_end'):
+                self.get_model().on_epoch_end()
 
     def run_training_batch(self, batch, batch_idx):
         # track grad norms
@@ -507,17 +505,15 @@ def run_training_batch(self, batch, batch_idx):
         if batch is None:
             return 0, grad_norm_dic, {}
 
-        # Batch begin callbacks
-        self.on_batch_start()
-
-        # hook
-        if self.is_function_implemented('on_batch_start'):
-            model_ref = self.get_model()
-            with self.profiler.profile('on_batch_start'):
-                response = model_ref.on_batch_start(batch)
-
-            if response == -1:
-                return -1, grad_norm_dic, {}
+        # Batch start events
+        with self.profiler.profile('on_batch_start'):
+            # callbacks
+            self.on_batch_start()
+            # hooks
+            if self.is_function_implemented('on_batch_start'):
+                response = self.get_model().on_batch_start(batch)
+                if response == -1:
+                    return -1, grad_norm_dic, {}
 
         splits = [batch]
         if self.truncated_bptt_steps is not None:
@@ -612,14 +608,13 @@ def optimizer_closure():
                     self.batch_loss_value = 0
                     self.avg_loss = np.mean(self.running_loss[-100:])
 
-        # activate batch end hook
-        if self.is_function_implemented('on_batch_end'):
-            model = self.get_model()
-            with self.profiler.profile('on_batch_end'):
-                model.on_batch_end()
-
-        # Batch end callbacks
-        self.on_batch_end()
+        # Batch end events
+        with self.profiler.profile('on_batch_end'):
+            # callbacks
+            self.on_batch_end()
+            # model hooks
+            if self.is_function_implemented('on_batch_end'):
+                self.get_model().on_batch_end()
 
         # update progress bar
         if batch_idx % self.progress_bar_refresh_rate == 0:
@@ -635,12 +630,15 @@ def optimizer_closure():
         return 0, grad_norm_dic, all_log_metrics
 
     def run_training_teardown(self):
-        model = self.get_model()
-
         self.main_progress_bar.close()
 
+        # Train end events
         with self.profiler.profile('on_train_end'):
-            model.on_train_end()
+            # callbacks
+            self.on_train_end()
+            # model hooks
+            if self.is_function_implemented('on_train_end'):
+                self.get_model().on_train_end()
 
         if self.logger is not None:
             self.logger.finalize("success")