Convert tmpdir test fixture to str after fsspec changes

requirements.txt

@@ -5,7 +5,7 @@ torch>=1.4
 future>=0.17.1  # required for builtins in setup.py
 tqdm>=4.41.0
 PyYAML>=5.1,<=5.4.1
-fsspec[http]>=2021.05.0, !=2021.06.0
+fsspec[http]>=2021.05.0
 tensorboard>=2.2.0, !=2.5.0  # 2.5.0 GPU CI error: 'Couldn't build proto file into descriptor pool!'
 torchmetrics>=0.2.0
 pyDeprecate==0.3.0

tests/callbacks/test_early_stopping.py

@@ -56,6 +56,7 @@ def test_resume_early_stopping_from_checkpoint(tmpdir):
     https://github.com/PyTorchLightning/pytorch-lightning/issues/1464
     https://github.com/PyTorchLightning/pytorch-lightning/issues/1463
     """
+    tmpdir = str(tmpdir)
     seed_everything(42)
     model = ClassificationModel()
     dm = ClassifDataModule()

tests/checkpointing/test_checkpoint_callback_frequency.py

@@ -91,6 +91,7 @@ def training_step(self, batch, batch_idx):
             self.last_coeff *= 0.999
             return loss
 
+    tmpdir = str(tmpdir)
     model = TestModel()
     trainer = Trainer(
         callbacks=[callbacks.ModelCheckpoint(dirpath=tmpdir, monitor='my_loss', save_top_k=k)],

tests/checkpointing/test_model_checkpoint.py

@@ -325,6 +325,7 @@ def test_model_checkpoint_with_non_string_input(tmpdir, save_top_k: int):
 @pytest.mark.parametrize('save_top_k', [-1, 0, 1, 2])
 def test_model_checkpoint_to_yaml(tmpdir, save_top_k: int):
     """ Test that None in checkpoint callback is valid and that chkp_path is set correctly """
+    tmpdir = str(tmpdir)
     tutils.reset_seed()
     model = LogInTwoMethods()
 
@@ -480,7 +481,7 @@ def test_model_checkpoint_file_extension(tmpdir):
     """
     Test ModelCheckpoint with different file extension.
     """
-
+    tmpdir = str(tmpdir)
     model = LogInTwoMethods()
     model_checkpoint = ModelCheckpointExtensionTest(
         monitor='early_stop_on',
@@ -535,6 +536,7 @@ def test_invalid_top_k(tmpdir):
 
 def test_none_monitor_top_k(tmpdir):
     """ Test that a warning appears for positive top_k with monitor=None. """
+    tmpdir = str(tmpdir)
     with pytest.raises(
         MisconfigurationException, match=r'ModelCheckpoint\(save_top_k=3, monitor=None\) is not a valid*'
     ):
@@ -657,6 +659,7 @@ def test_model_checkpoint_period(tmpdir, period: int):
 
 @pytest.mark.parametrize("every_n_val_epochs", list(range(4)))
 def test_model_checkpoint_every_n_val_epochs(tmpdir, every_n_val_epochs):
+    tmpdir = str(tmpdir)
     model = LogInTwoMethods()
     epochs = 5
     checkpoint_callback = ModelCheckpoint(
@@ -681,6 +684,7 @@ def test_model_checkpoint_every_n_val_epochs(tmpdir, every_n_val_epochs):
 @pytest.mark.parametrize("every_n_val_epochs", list(range(4)))
 def test_model_checkpoint_every_n_val_epochs_and_period(tmpdir, every_n_val_epochs):
     """ Tests that if period is set, it takes precedence over every_n_val_epochs for backwards compatibility. """
+    tmpdir = str(tmpdir)
     model = LogInTwoMethods()
     epochs = 5
     checkpoint_callback = ModelCheckpoint(
@@ -896,6 +900,7 @@ def test_model_checkpoint_save_last_warning(
 
 def test_model_checkpoint_save_last_checkpoint_contents(tmpdir):
     """ Tests that the save_last checkpoint contains the latest information. """
+    tmpdir = str(tmpdir)
     seed_everything(100)
     model = LogInTwoMethods()
     num_epochs = 3
@@ -931,6 +936,7 @@ def test_model_checkpoint_save_last_checkpoint_contents(tmpdir):
 @mock.patch.dict(os.environ, {"PL_DEV_DEBUG": "1"})
 @pytest.mark.parametrize('mode', ['min', 'max'])
 def test_checkpointing_with_nan_as_first(tmpdir, mode: int):
+    tmpdir = str(tmpdir)
     monitor = [float('nan')]
     monitor += [5, 7, 8] if mode == 'max' else [8, 7, 5]
 
@@ -1150,6 +1156,7 @@ def test_val_check_interval_checkpoint_files(tmpdir):
 
 def test_current_score(tmpdir):
     """ Check that the current_score value is correct and was saved """
+    tmpdir = str(tmpdir)
 
     class TestModel(BoringModel):
 
@@ -1183,6 +1190,7 @@ def training_step(self, *args):
 @pytest.mark.parametrize("mode", ["min", "max"])
 def test_current_score_when_nan(tmpdir, mode: str):
     """ Check that ModelCheckpoint handles NaN values correctly """
+    tmpdir = str(tmpdir)
 
     class TestModel(BoringModel):
 
@@ -1213,6 +1221,7 @@ def training_step(self, *args):
 
 @pytest.mark.parametrize("hparams_type", [dict, Container])
 def test_hparams_type(tmpdir, hparams_type):
+    tmpdir = str(tmpdir)
 
     class TestModel(BoringModel):
 

tests/models/test_cpu.py

@@ -94,6 +94,7 @@ def on_train_epoch_start(self, trainer, model):
 
 
 def test_early_stopping_cpu_model(tmpdir):
+    tmpdir = str(tmpdir)
 
     class ModelTrainVal(BoringModel):
 
@@ -128,6 +129,7 @@ def validation_step(self, *args, **kwargs):
 def test_multi_cpu_model_ddp(tmpdir):
     """Make sure DDP works."""
     tutils.set_random_master_port()
+    tmpdir = str(tmpdir)
 
     trainer_options = dict(
         default_root_dir=tmpdir,
@@ -295,6 +297,7 @@ def test_simple_cpu(tmpdir):
 
 def test_cpu_model(tmpdir):
     """Make sure model trains on CPU."""
+    tmpdir = str(tmpdir)
     trainer_options = dict(
         default_root_dir=tmpdir, progress_bar_refresh_rate=0, max_epochs=1, limit_train_batches=4, limit_val_batches=4
     )
@@ -305,6 +308,7 @@ def test_cpu_model(tmpdir):
 
 def test_all_features_cpu_model(tmpdir):
     """Test each of the trainer options."""
+    tmpdir = str(tmpdir)
     trainer_options = dict(
         default_root_dir=tmpdir,
         gradient_clip_val=1.0,

tests/trainer/test_trainer.py

@@ -318,6 +318,7 @@ def test_loading_yaml(tmpdir):
 )
 def test_model_checkpoint_options(tmpdir, save_top_k, save_last, expected_files):
     """Test ModelCheckpoint options."""
+    tmpdir = str(tmpdir)
 
     def mock_save_function(filepath, *args):
         open(filepath, "a").close()
@@ -1836,6 +1837,7 @@ def validation_epoch_end(self, outputs) -> None:
 @RunIf(skip_windows=True)
 def test_fit_test_synchronization(tmpdir):
     """Test that the trainer synchronizes processes before returning control back to the caller. """
+    tmpdir = str(tmpdir)
     tutils.set_random_master_port()
     model = TestDummyModelForCheckpoint()
     checkpoint = ModelCheckpoint(dirpath=tmpdir, monitor='x', mode='min', save_top_k=1)