test: replacing tmpdir with tmp_path in tests_pytorch/core (#19644

)

refactored tmpdir from the tests_pytorch/core dir

tests/tests_pytorch/core/test_datamodules.py

@@ -145,7 +145,7 @@ def test_dm_pickle_after_init():
 
 
 @RunIf(sklearn=True)
-def test_train_loop_only(tmpdir):
+def test_train_loop_only(tmp_path):
     seed_everything(7)
 
     dm = ClassifDataModule()
@@ -154,7 +154,7 @@ def test_train_loop_only(tmpdir):
     model.validation_step = None
     model.test_step = None
 
-    trainer = Trainer(default_root_dir=tmpdir, max_epochs=1, enable_model_summary=False)
+    trainer = Trainer(default_root_dir=tmp_path, max_epochs=1, enable_model_summary=False)
 
     # fit model
     trainer.fit(model, datamodule=dm)
@@ -163,23 +163,23 @@ def test_train_loop_only(tmpdir):
 
 
 @RunIf(sklearn=True)
-def test_train_val_loop_only(tmpdir):
+def test_train_val_loop_only(tmp_path):
     seed_everything(7)
 
     dm = ClassifDataModule()
     model = ClassificationModel()
 
     model.validation_step = None
 
-    trainer = Trainer(default_root_dir=tmpdir, max_epochs=1, enable_model_summary=False)
+    trainer = Trainer(default_root_dir=tmp_path, max_epochs=1, enable_model_summary=False)
 
     # fit model
     trainer.fit(model, datamodule=dm)
     assert trainer.state.finished, f"Training failed with {trainer.state}"
     assert trainer.callback_metrics["train_loss"] < 1.1
 
 
-def test_dm_checkpoint_save_and_load(tmpdir):
+def test_dm_checkpoint_save_and_load(tmp_path):
     class CustomBoringModel(BoringModel):
         def validation_step(self, batch, batch_idx):
             out = super().validation_step(batch, batch_idx)
@@ -197,12 +197,12 @@ def load_state_dict(self, state_dict: Dict[str, Any]) -> None:
     model = CustomBoringModel()
 
     trainer = Trainer(
-        default_root_dir=tmpdir,
+        default_root_dir=tmp_path,
         max_epochs=1,
         limit_train_batches=2,
         limit_val_batches=1,
         enable_model_summary=False,
-        callbacks=[ModelCheckpoint(dirpath=tmpdir, monitor="early_stop_on")],
+        callbacks=[ModelCheckpoint(dirpath=tmp_path, monitor="early_stop_on")],
     )
 
     # fit model
@@ -219,13 +219,13 @@ def load_state_dict(self, state_dict: Dict[str, Any]) -> None:
 
 
 @RunIf(sklearn=True)
-def test_full_loop(tmpdir):
+def test_full_loop(tmp_path):
     seed_everything(7)
 
     dm = ClassifDataModule()
     model = ClassificationModel()
 
-    trainer = Trainer(default_root_dir=tmpdir, max_epochs=1, enable_model_summary=False, deterministic="warn")
+    trainer = Trainer(default_root_dir=tmp_path, max_epochs=1, enable_model_summary=False, deterministic="warn")
 
     # fit model
     trainer.fit(model, dm)
@@ -243,7 +243,7 @@ def test_full_loop(tmpdir):
     assert result[0]["test_acc"] > 0.57
 
 
-def test_dm_reload_dataloaders_every_n_epochs(tmpdir):
+def test_dm_reload_dataloaders_every_n_epochs(tmp_path):
     """Test datamodule, where trainer argument reload_dataloaders_every_n_epochs is set to a non negative integer."""
 
     class CustomBoringDataModule(BoringDataModule):
@@ -262,7 +262,9 @@ def train_dataloader(self):
     model.validation_step = None
     model.test_step = None
 
-    trainer = Trainer(default_root_dir=tmpdir, max_epochs=3, limit_train_batches=2, reload_dataloaders_every_n_epochs=2)
+    trainer = Trainer(
+        default_root_dir=tmp_path, max_epochs=3, limit_train_batches=2, reload_dataloaders_every_n_epochs=2
+    )
     trainer.fit(model, dm)
 
 

tests/tests_pytorch/core/test_lightning_module.py

@@ -75,29 +75,29 @@ def test_property_local_rank():
     assert model.local_rank == 123
 
 
-def test_property_logger(tmpdir):
+def test_property_logger(tmp_path):
     """Test that the logger in LightningModule is accessible via the Trainer."""
     model = BoringModel()
     assert model.logger is None
 
-    logger = TensorBoardLogger(tmpdir)
+    logger = TensorBoardLogger(tmp_path)
     trainer = Trainer(logger=logger)
     model.trainer = trainer
     assert model.logger == logger
 
 
-def test_property_loggers(tmpdir):
+def test_property_loggers(tmp_path):
     """Test that loggers in LightningModule is accessible via the Trainer."""
     model = BoringModel()
     assert model.loggers == []
 
-    logger = TensorBoardLogger(tmpdir)
+    logger = TensorBoardLogger(tmp_path)
     trainer = Trainer(logger=logger)
     model.trainer = trainer
     assert model.loggers == [logger]
 
-    logger0 = TensorBoardLogger(tmpdir)
-    logger1 = TensorBoardLogger(tmpdir)
+    logger0 = TensorBoardLogger(tmp_path)
+    logger1 = TensorBoardLogger(tmp_path)
     trainer = Trainer(logger=[logger0, logger1])
     model.trainer = trainer
     assert model.loggers == [logger0, logger1]
@@ -120,7 +120,7 @@ def test_1_optimizer_toggle_model():
     assert not model._param_requires_grad_state
 
 
-def test_toggle_untoggle_2_optimizers_no_shared_parameters(tmpdir):
+def test_toggle_untoggle_2_optimizers_no_shared_parameters(tmp_path):
     class TestModel(BoringModel):
         def __init__(self):
             super().__init__()
@@ -178,11 +178,11 @@ def configure_optimizers(self):
 
     model = TestModel()
 
-    trainer = Trainer(max_epochs=1, default_root_dir=tmpdir, limit_train_batches=8, limit_val_batches=0)
+    trainer = Trainer(max_epochs=1, default_root_dir=tmp_path, limit_train_batches=8, limit_val_batches=0)
     trainer.fit(model)
 
 
-def test_toggle_untoggle_3_optimizers_shared_parameters(tmpdir):
+def test_toggle_untoggle_3_optimizers_shared_parameters(tmp_path):
     class TestModel(BoringModel):
         def __init__(self):
             super().__init__()
@@ -279,7 +279,7 @@ def configure_optimizers(self):
             return [optimizer_1, optimizer_2, optimizer_3]
 
     model = TestModel()
-    trainer = Trainer(max_epochs=1, default_root_dir=tmpdir, limit_train_batches=8)
+    trainer = Trainer(max_epochs=1, default_root_dir=tmp_path, limit_train_batches=8)
     trainer.fit(model)
 
 
@@ -290,9 +290,9 @@ def configure_optimizers(self):
         pytest.param("mps", "mps:0", marks=RunIf(mps=True)),
     ],
 )
-def test_device_placement(tmpdir, accelerator, device):
+def test_device_placement(tmp_path, accelerator, device):
     model = BoringModel()
-    trainer = Trainer(default_root_dir=tmpdir, fast_dev_run=True, accelerator=accelerator, devices=1)
+    trainer = Trainer(default_root_dir=tmp_path, fast_dev_run=True, accelerator=accelerator, devices=1)
     trainer.fit(model)
 
     def assert_device(device: torch.device) -> None:
@@ -342,7 +342,7 @@ def __init__(self, spec):
     ), "Expect the shards to be same after `m_1` loading `m_0`'s state dict"
 
 
-def test_lightning_module_configure_gradient_clipping(tmpdir):
+def test_lightning_module_configure_gradient_clipping(tmp_path):
     """Test custom gradient clipping inside `configure_gradient_clipping` hook."""
 
     class TestModel(BoringModel):
@@ -359,7 +359,7 @@ def configure_gradient_clipping(self, optimizer, gradient_clip_val, gradient_cli
 
     model = TestModel()
     trainer = Trainer(
-        default_root_dir=tmpdir, max_epochs=1, limit_train_batches=1, limit_val_batches=0, gradient_clip_val=1e-4
+        default_root_dir=tmp_path, max_epochs=1, limit_train_batches=1, limit_val_batches=0, gradient_clip_val=1e-4
     )
     trainer.fit(model)
 
@@ -371,7 +371,7 @@ def configure_gradient_clipping(self, optimizer, gradient_clip_val, gradient_cli
                 assert p.grad.max() <= model.custom_gradient_clip_val
 
 
-def test_lightning_module_configure_gradient_clipping_different_argument_values(tmpdir):
+def test_lightning_module_configure_gradient_clipping_different_argument_values(tmp_path):
     """Test that setting gradient clipping arguments in `Trainer` and cusotmizing gradient clipping inside
     `configure_gradient_clipping` with different values raises an exception."""
 
@@ -383,7 +383,7 @@ def configure_gradient_clipping(self, optimizer, gradient_clip_val, gradient_cli
 
     model = TestModel()
     trainer = Trainer(
-        default_root_dir=tmpdir, max_epochs=1, limit_train_batches=2, limit_val_batches=0, gradient_clip_val=1e-4
+        default_root_dir=tmp_path, max_epochs=1, limit_train_batches=2, limit_val_batches=0, gradient_clip_val=1e-4
     )
     with pytest.raises(
         MisconfigurationException,
@@ -399,7 +399,7 @@ def configure_gradient_clipping(self, optimizer, gradient_clip_val, gradient_cli
 
     model = TestModel()
     trainer = Trainer(
-        default_root_dir=tmpdir,
+        default_root_dir=tmp_path,
         max_epochs=1,
         limit_train_batches=2,
         limit_val_batches=0,

tests/tests_pytorch/core/test_lightning_optimizer.py

@@ -25,7 +25,7 @@
 
 
 @pytest.mark.parametrize("auto", [True, False])
-def test_lightning_optimizer(tmpdir, auto):
+def test_lightning_optimizer(tmp_path, auto):
     """Test that optimizer are correctly wrapped by our LightningOptimizer."""
 
     class TestModel(BoringModel):
@@ -39,23 +39,23 @@ def configure_optimizers(self):
 
     model = TestModel()
     trainer = Trainer(
-        default_root_dir=tmpdir, limit_train_batches=1, limit_val_batches=1, max_epochs=1, enable_model_summary=False
+        default_root_dir=tmp_path, limit_train_batches=1, limit_val_batches=1, max_epochs=1, enable_model_summary=False
     )
     trainer.fit(model)
 
     lightning_opt = model.optimizers()
     assert str(lightning_opt) == "Lightning" + str(lightning_opt.optimizer)
 
 
-def test_init_optimizers_resets_lightning_optimizers(tmpdir):
+def test_init_optimizers_resets_lightning_optimizers(tmp_path):
     """Test that the Trainer resets the `lightning_optimizers` list everytime new optimizers get initialized."""
 
     def compare_optimizers():
         assert trainer.strategy._lightning_optimizers[0].optimizer is trainer.optimizers[0]
 
     model = BoringModel()
     model.lr = 0.2
-    trainer = Trainer(default_root_dir=tmpdir, max_epochs=1)
+    trainer = Trainer(default_root_dir=tmp_path, max_epochs=1)
     tuner = Tuner(trainer)
 
     tuner.lr_find(model)
@@ -69,7 +69,7 @@ def compare_optimizers():
     compare_optimizers()
 
 
-def test_lightning_optimizer_manual_optimization_and_accumulated_gradients(tmpdir):
+def test_lightning_optimizer_manual_optimization_and_accumulated_gradients(tmp_path):
     """Test that the user can use our LightningOptimizer.
 
     Not recommended.
@@ -110,7 +110,7 @@ def configure_optimizers(self):
 
     model = TestModel()
     trainer = Trainer(
-        default_root_dir=tmpdir, limit_train_batches=8, limit_val_batches=1, max_epochs=1, enable_model_summary=False
+        default_root_dir=tmp_path, limit_train_batches=8, limit_val_batches=1, max_epochs=1, enable_model_summary=False
     )
 
     with patch.multiple(torch.optim.SGD, zero_grad=DEFAULT, step=DEFAULT) as sgd, patch.multiple(
@@ -174,7 +174,7 @@ def test_state_mutation():
     assert lightning_optimizer2.param_groups[0]["lr"] == 1.0
 
 
-def test_lightning_optimizer_automatic_optimization_optimizer_zero_grad(tmpdir):
+def test_lightning_optimizer_automatic_optimization_optimizer_zero_grad(tmp_path):
     """Test overriding zero_grad works in automatic_optimization."""
 
     class TestModel(BoringModel):
@@ -189,15 +189,15 @@ def configure_optimizers(self):
 
     model = TestModel()
     trainer = Trainer(
-        default_root_dir=tmpdir, limit_train_batches=20, limit_val_batches=1, max_epochs=1, enable_model_summary=False
+        default_root_dir=tmp_path, limit_train_batches=20, limit_val_batches=1, max_epochs=1, enable_model_summary=False
     )
 
     with patch("torch.optim.SGD.zero_grad") as sgd_zero_grad:
         trainer.fit(model)
     assert sgd_zero_grad.call_count == 10
 
 
-def test_lightning_optimizer_automatic_optimization_optimizer_step(tmpdir):
+def test_lightning_optimizer_automatic_optimization_optimizer_step(tmp_path):
     """Test overriding step works in automatic_optimization."""
 
     class TestModel(BoringModel):
@@ -218,7 +218,7 @@ def configure_optimizers(self):
 
     limit_train_batches = 8
     trainer = Trainer(
-        default_root_dir=tmpdir,
+        default_root_dir=tmp_path,
         limit_train_batches=limit_train_batches,
         limit_val_batches=1,
         max_epochs=1,
@@ -232,7 +232,7 @@ def configure_optimizers(self):
     assert sgd["zero_grad"].call_count == limit_train_batches
 
 
-def test_lightning_optimizer_automatic_optimization_lbfgs_zero_grad(tmpdir):
+def test_lightning_optimizer_automatic_optimization_lbfgs_zero_grad(tmp_path):
     """Test zero_grad is called the same number of times as LBFGS requires for reevaluation of the loss in
     automatic_optimization."""
 
@@ -242,7 +242,7 @@ def configure_optimizers(self):
 
     model = TestModel()
     trainer = Trainer(
-        default_root_dir=tmpdir, limit_train_batches=1, limit_val_batches=1, max_epochs=1, enable_model_summary=False
+        default_root_dir=tmp_path, limit_train_batches=1, limit_val_batches=1, max_epochs=1, enable_model_summary=False
     )
 
     with patch("torch.optim.LBFGS.zero_grad") as zero_grad:
@@ -308,7 +308,7 @@ def test_lightning_optimizer_keeps_hooks():
     assert len(optimizer._fwd_handles) == 1
 
 
-def test_params_groups_and_state_are_accessible(tmpdir):
+def test_params_groups_and_state_are_accessible(tmp_path):
     class TestModel(BoringModel):
         def on_train_start(self):
             # Update the learning rate manually on the unwrapped optimizer
@@ -337,5 +337,5 @@ def optimizer_step(self, epoch, batch_idx, optimizer, optimizer_closure, **__):
             assert loss == self.__loss
 
     model = TestModel()
-    trainer = Trainer(max_epochs=1, default_root_dir=tmpdir, limit_train_batches=1, limit_val_batches=0)
+    trainer = Trainer(max_epochs=1, default_root_dir=tmp_path, limit_train_batches=1, limit_val_batches=0)
     trainer.fit(model)