validate manual optimization and supported features before running tr…

…aining (#7788)


Co-authored-by: Carlos Mocholí <carlossmocholi@gmail.com>

CHANGELOG.md

@@ -119,6 +119,9 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Changed `teardown()` in `Accelerator` to allow `training_type_plugin` to customize `teardown` logic ([#7579](https://github.com/PyTorchLightning/pytorch-lightning/pull/7579))
 
 
+- `Trainer.fit` now raises an error when using manual optimization with unsupported features such as `gradient_clip_val` or `accumulate_grad_batches` ([#7788](https://github.com/PyTorchLightning/pytorch-lightning/pull/7788))
+
+
 ### Deprecated
 
 

pytorch_lightning/trainer/configuration_validator.py

@@ -34,6 +34,7 @@ def verify_loop_configurations(self, model: 'pl.LightningModule') -> None:
         if self.trainer.state.fn in (TrainerFn.FITTING, TrainerFn.TUNING):
             self.__verify_train_loop_configuration(model)
             self.__verify_eval_loop_configuration(model, 'val')
+            self.__verify_manual_optimization_support(model)
         elif self.trainer.state.fn == TrainerFn.VALIDATING:
             self.__verify_eval_loop_configuration(model, 'val')
         elif self.trainer.state.fn == TrainerFn.TESTING:
@@ -112,3 +113,19 @@ def __verify_dp_batch_transfer_support(self, model: 'pl.LightningModule') -> Non
         for hook in batch_transfer_hooks:
             if self.trainer.accelerator_connector.use_dp and is_overridden(hook, model):
                 raise MisconfigurationException(f'Overriding `{hook}` is not supported in DP mode.')
+
+    def __verify_manual_optimization_support(self, model: 'pl.LightningModule') -> None:
+        if model.automatic_optimization:
+            return
+        if self.trainer.gradient_clip_val > 0:
+            raise MisconfigurationException(
+                f"Automatic gradient clipping is not supported for manual optimization."
+                f" Remove `Trainer(gradient_clip_val={self.trainer.gradient_clip_val})`"
+                f" or switch to automatic optimization."
+            )
+        if self.trainer.accumulate_grad_batches != 1:
+            raise MisconfigurationException(
+                f"Automatic gradient accumulation is not supported for manual optimization."
+                f" Remove `Trainer(accumulate_grad_batches={self.trainer.accumulate_grad_batches})`"
+                f" or switch to automatic optimization."
+            )

tests/core/test_lightning_optimizer.py

@@ -123,7 +123,6 @@ def configure_optimizers(self):
         limit_val_batches=1,
         max_epochs=1,
         weights_summary=None,
-        accumulate_grad_batches=999,  # does not do anything if manual optimization
     )
 
     with patch.multiple(torch.optim.SGD, zero_grad=DEFAULT, step=DEFAULT) as sgd, \

tests/trainer/optimization/test_manual_optimization.py

@@ -424,7 +424,6 @@ def on_train_epoch_end(self, *_, **__):
         limit_val_batches=0,
         precision=16,
         amp_backend='native',
-        accumulate_grad_batches=4,
         gpus=1,
     )
     trainer.fit(model)
@@ -631,7 +630,6 @@ def configure_optimizers(self):
         limit_val_batches=2,
         max_epochs=1,
         log_every_n_steps=1,
-        accumulate_grad_batches=2,
     )
 
     trainer.fit(model)
@@ -682,7 +680,6 @@ def configure_optimizers(self):
         limit_val_batches=2,
         max_epochs=1,
         log_every_n_steps=1,
-        accumulate_grad_batches=2,
     )
 
     trainer.fit(model)
@@ -757,7 +754,6 @@ def configure_optimizers(self):
         limit_val_batches=2,
         max_epochs=1,
         log_every_n_steps=1,
-        accumulate_grad_batches=2,
     )
 
     trainer.fit(model)
@@ -867,7 +863,6 @@ def train_manual_optimization(tmpdir, accelerator, model_cls=TesManualOptimizati
         limit_val_batches=2,
         max_epochs=1,
         log_every_n_steps=1,
-        accumulate_grad_batches=2,
         gpus=2,
         accelerator=accelerator,
         callbacks=[TestManualOptimizationDDPCallack()]

tests/trainer/test_config_validator.py

@@ -147,3 +147,17 @@ def predict_dataloader(self):
 
     with pytest.raises(MisconfigurationException, match="Dataloader not found for `Trainer.predict`"):
         trainer.predict(model)
+
+
+def test_trainer_manual_optimization_config(tmpdir):
+    """ Test error message when requesting Trainer features unsupported with manual optimization """
+    model = BoringModel()
+    model.automatic_optimization = False
+
+    trainer = Trainer(gradient_clip_val=1.0)
+    with pytest.raises(MisconfigurationException, match="Automatic gradient clipping is not supported"):
+        trainer.fit(model)
+
+    trainer = Trainer(accumulate_grad_batches=2)
+    with pytest.raises(MisconfigurationException, match="Automatic gradient accumulation is not supported"):
+        trainer.fit(model)