diff --git a/pytorch_lightning/trainer/trainer.py b/pytorch_lightning/trainer/trainer.py
index b6cdbb0cf130df..70a6940588c7ba 100644
--- a/pytorch_lightning/trainer/trainer.py
+++ b/pytorch_lightning/trainer/trainer.py
@@ -371,10 +371,12 @@ def __init__(
                            ' val and test loop using a single batch')
 
         # set default save path if user didn't provide one
-        self.default_root_dir = default_root_dir
-
-        if self.default_root_dir is None:
+        if default_root_dir is None:
             self.default_root_dir = os.getcwd()
+        else:
+            # we have to do str() because the unit tests violate type annotation and pass path objecto
+            self.default_root_dir = str(default_root_dir)
+
 
         # training bookeeping
         self.total_batch_idx = 0
diff --git a/pytorch_lightning/trainer/training_io.py b/pytorch_lightning/trainer/training_io.py
index 39e45a134a4cb4..f85ba273bf6d1c 100644
--- a/pytorch_lightning/trainer/training_io.py
+++ b/pytorch_lightning/trainer/training_io.py
@@ -375,7 +375,7 @@ def restore_hpc_weights_if_needed(self, model: LightningModule):
         did_restore = False
 
         # look for hpc weights
-        folderpath = self.weights_save_path
+        folderpath = str(self.weights_save_path)
         if gfile.exists(folderpath):
             files = gfile.listdir(folderpath)
             hpc_weight_paths = [x for x in files if 'hpc_ckpt' in x]
@@ -452,6 +452,7 @@ def restore_training_state(self, checkpoint):
     # ----------------------------------
     def hpc_save(self, folderpath: str, logger):
         # make sure the checkpoint folder exists
+        folderpath = str(folderpath)  # because the tests pass a path object
         if not gfile.exists(folderpath):
             gfile.makedirs(folderpath)
 
@@ -511,7 +512,7 @@ def hpc_load(self, folderpath, on_gpu):
         log.info(f'restored hpc model from: {filepath}')
 
     def max_ckpt_in_folder(self, path, name_key='ckpt_'):
-        files = gfile.listdir(path)
+        files = gfile.listdir(str(path))
         files = [x for x in files if name_key in x]
         if len(files) == 0:
             return 0
diff --git a/tests/loggers/test_all.py b/tests/loggers/test_all.py
index a6d0f683516e11..f8a8fead41f586 100644
--- a/tests/loggers/test_all.py
+++ b/tests/loggers/test_all.py
@@ -46,7 +46,7 @@ def log_metrics(self, metrics, step):
             super().log_metrics(metrics, step)
             self.history.append((step, metrics))
 
-    logger_args = _get_logger_args(logger_class, str(tmpdir))
+    logger_args = _get_logger_args(logger_class, tmpdir)
     logger = StoreHistoryLogger(**logger_args)
 
     trainer = Trainer(
@@ -82,7 +82,7 @@ def test_loggers_pickle(tmpdir, monkeypatch, logger_class):
     import atexit
     monkeypatch.setattr(atexit, 'register', lambda _: None)
 
-    logger_args = _get_logger_args(logger_class, str(tmpdir))
+    logger_args = _get_logger_args(logger_class, tmpdir)
     logger = logger_class(**logger_args)
 
     # test pickling loggers
@@ -109,7 +109,7 @@ def test_logger_reset_correctly(tmpdir, extra_params):
     model = EvalModelTemplate()
 
     trainer = Trainer(
-        default_root_dir=str(tmpdir),
+        default_root_dir=tmpdir,
         **extra_params
     )
     logger1 = trainer.logger
diff --git a/tests/models/test_cpu.py b/tests/models/test_cpu.py
index 0656723b3f334c..051db81d1b1657 100644
--- a/tests/models/test_cpu.py
+++ b/tests/models/test_cpu.py
@@ -19,7 +19,7 @@ def test_cpu_slurm_save_load(tmpdir):
     model = EvalModelTemplate(**hparams)
 
     # logger file to get meta
-    logger = tutils.get_default_logger(str(tmpdir))
+    logger = tutils.get_default_logger(tmpdir)
     version = logger.version
 
     # fit model
@@ -28,7 +28,7 @@ def test_cpu_slurm_save_load(tmpdir):
         logger=logger,
         train_percent_check=0.2,
         val_percent_check=0.2,
-        checkpoint_callback=ModelCheckpoint(str(tmpdir)),
+        checkpoint_callback=ModelCheckpoint(tmpdir)
     )
     result = trainer.fit(model)
     real_global_step = trainer.global_step
@@ -54,13 +54,17 @@ def test_cpu_slurm_save_load(tmpdir):
 
     # test HPC saving
     # simulate snapshot on slurm
-    saved_filepath = trainer.hpc_save(str(tmpdir), logger)
+    saved_filepath = trainer.hpc_save(tmpdir, logger)
     assert os.path.exists(saved_filepath)
 
     # new logger file to get meta
-    logger = tutils.get_default_logger(str(tmpdir), version=version)
+    logger = tutils.get_default_logger(tmpdir, version=version)
 
-    trainer = Trainer(max_epochs=1, logger=logger, checkpoint_callback=ModelCheckpoint(str(tmpdir)),)
+    trainer = Trainer(
+        max_epochs=1,
+        logger=logger,
+        checkpoint_callback=ModelCheckpoint(tmpdir),
+    )
     model = EvalModelTemplate(**hparams)
 
     # set the epoch start hook so we can predict before the model does the full training
@@ -83,7 +87,7 @@ def test_early_stopping_cpu_model(tmpdir):
     """Test each of the trainer options."""
     stopping = EarlyStopping(monitor='val_loss', min_delta=0.1)
     trainer_options = dict(
-        default_root_dir=str(tmpdir),
+        default_root_dir=tmpdir,
         early_stop_callback=stopping,
         max_epochs=2,
         gradient_clip_val=1.0,
@@ -112,7 +116,7 @@ def test_multi_cpu_model_ddp(tmpdir):
     tutils.set_random_master_port()
 
     trainer_options = dict(
-        default_root_dir=str(tmpdir),
+        default_root_dir=tmpdir,
         progress_bar_refresh_rate=0,
         max_epochs=1,
         train_percent_check=0.4,
@@ -129,7 +133,7 @@ def test_multi_cpu_model_ddp(tmpdir):
 def test_lbfgs_cpu_model(tmpdir):
     """Test each of the trainer options."""
     trainer_options = dict(
-        default_root_dir=str(tmpdir),
+        default_root_dir=tmpdir,
         max_epochs=1,
         progress_bar_refresh_rate=0,
         weights_summary='top',
@@ -148,7 +152,7 @@ def test_lbfgs_cpu_model(tmpdir):
 def test_default_logger_callbacks_cpu_model(tmpdir):
     """Test each of the trainer options."""
     trainer_options = dict(
-        default_root_dir=str(tmpdir),
+        default_root_dir=tmpdir,
         max_epochs=1,
         gradient_clip_val=1.0,
         overfit_pct=0.20,
@@ -170,14 +174,14 @@ def test_running_test_after_fitting(tmpdir):
     model = EvalModelTemplate()
 
     # logger file to get meta
-    logger = tutils.get_default_logger(str(tmpdir))
+    logger = tutils.get_default_logger(tmpdir)
 
     # logger file to get weights
     checkpoint = tutils.init_checkpoint_callback(logger)
 
     # fit model
     trainer = Trainer(
-        default_root_dir=str(tmpdir),
+        default_root_dir=tmpdir,
         progress_bar_refresh_rate=0,
         max_epochs=2,
         train_percent_check=0.4,
@@ -201,7 +205,7 @@ def test_running_test_no_val(tmpdir):
     model = EvalModelTemplate()
 
     # logger file to get meta
-    logger = tutils.get_default_logger(str(tmpdir))
+    logger = tutils.get_default_logger(tmpdir)
 
     # logger file to get weights
     checkpoint = tutils.init_checkpoint_callback(logger)
@@ -280,7 +284,7 @@ def test_simple_cpu(tmpdir):
 
     # fit model
     trainer = Trainer(
-        default_root_dir=str(tmpdir),
+        default_root_dir=tmpdir,
         max_epochs=1,
         val_percent_check=0.1,
         train_percent_check=0.1,
@@ -294,7 +298,7 @@ def test_simple_cpu(tmpdir):
 def test_cpu_model(tmpdir):
     """Make sure model trains on CPU."""
     trainer_options = dict(
-        default_root_dir=str(tmpdir),
+        default_root_dir=tmpdir,
         progress_bar_refresh_rate=0,
         max_epochs=1,
         train_percent_check=0.4,
@@ -309,7 +313,7 @@ def test_cpu_model(tmpdir):
 def test_all_features_cpu_model(tmpdir):
     """Test each of the trainer options."""
     trainer_options = dict(
-        default_root_dir=str(tmpdir),
+        default_root_dir=tmpdir,
         gradient_clip_val=1.0,
         overfit_pct=0.20,
         track_grad_norm=2,
@@ -383,7 +387,7 @@ def train_dataloader(self):
 
     # fit model
     trainer = Trainer(
-        default_root_dir=str(tmpdir),
+        default_root_dir=tmpdir,
         max_epochs=1,
         truncated_bptt_steps=truncated_bptt_steps,
         val_percent_check=0,
diff --git a/tests/models/test_gpu.py b/tests/models/test_gpu.py
index 90f543bc2fb1c2..80249a727ccbbd 100644
--- a/tests/models/test_gpu.py
+++ b/tests/models/test_gpu.py
@@ -18,7 +18,7 @@
 def test_single_gpu_model(tmpdir, gpus):
     """Make sure single GPU works (DP mode)."""
     trainer_options = dict(
-        default_root_dir=str(tmpdir),
+        default_root_dir=tmpdir,
         progress_bar_refresh_rate=0,
         max_epochs=1,
         train_percent_check=0.1,
@@ -38,7 +38,7 @@ def test_multi_gpu_model(tmpdir, backend):
     tutils.set_random_master_port()
 
     trainer_options = dict(
-        default_root_dir=str(tmpdir),
+        default_root_dir=tmpdir,
         max_epochs=1,
         train_percent_check=0.4,
         val_percent_check=0.2,
@@ -84,7 +84,7 @@ def test_ddp_all_dataloaders_passed_to_fit(tmpdir):
 def test_multi_gpu_none_backend(tmpdir):
     """Make sure when using multiple GPUs the user can't use `distributed_backend = None`."""
     trainer_options = dict(
-        default_root_dir=str(tmpdir),
+        default_root_dir=tmpdir,
         progress_bar_refresh_rate=0,
         max_epochs=1,
         train_percent_check=0.1,
diff --git a/tests/trainer/test_lr_finder.py b/tests/trainer/test_lr_finder.py
index c635d87788363b..d0becff0918c65 100755
--- a/tests/trainer/test_lr_finder.py
+++ b/tests/trainer/test_lr_finder.py
@@ -14,7 +14,7 @@ def test_error_on_more_than_1_optimizer(tmpdir):
 
     # logger file to get meta
     trainer = Trainer(
-        default_root_dir=str(tmpdir),
+        default_root_dir=tmpdir,
         max_epochs=1
     )
 
@@ -29,7 +29,7 @@ def test_model_reset_correctly(tmpdir):
 
     # logger file to get meta
     trainer = Trainer(
-        default_root_dir=str(tmpdir),
+        default_root_dir=tmpdir,
         max_epochs=1
     )
 
@@ -51,7 +51,7 @@ def test_trainer_reset_correctly(tmpdir):
 
     # logger file to get meta
     trainer = Trainer(
-        default_root_dir=str(tmpdir),
+        default_root_dir=tmpdir,
         max_epochs=1
     )
 
@@ -81,7 +81,7 @@ def test_trainer_arg_bool(tmpdir):
 
     # logger file to get meta
     trainer = Trainer(
-        default_root_dir=str(tmpdir),
+        default_root_dir=tmpdir,
         max_epochs=2,
         auto_lr_find=True
     )
@@ -100,7 +100,7 @@ def test_trainer_arg_str(tmpdir):
     before_lr = model.my_fancy_lr
     # logger file to get meta
     trainer = Trainer(
-        default_root_dir=str(tmpdir),
+        default_root_dir=tmpdir,
         max_epochs=2,
         auto_lr_find='my_fancy_lr'
     )
@@ -120,7 +120,7 @@ def test_call_to_trainer_method(tmpdir):
     before_lr = hparams.get('learning_rate')
     # logger file to get meta
     trainer = Trainer(
-        default_root_dir=str(tmpdir),
+        default_root_dir=tmpdir,
         max_epochs=2,
     )
 
@@ -144,7 +144,7 @@ def test_accumulation_and_early_stopping(tmpdir):
     before_lr = hparams.get('learning_rate')
     # logger file to get meta
     trainer = Trainer(
-        default_root_dir=str(tmpdir),
+        default_root_dir=tmpdir,
         accumulate_grad_batches=2,
     )
 
@@ -167,7 +167,7 @@ def test_suggestion_parameters_work(tmpdir):
 
     # logger file to get meta
     trainer = Trainer(
-        default_root_dir=str(tmpdir),
+        default_root_dir=tmpdir,
         max_epochs=3,
     )
 
@@ -187,7 +187,7 @@ def test_suggestion_with_non_finite_values(tmpdir):
 
     # logger file to get meta
     trainer = Trainer(
-        default_root_dir=str(tmpdir),
+        default_root_dir=tmpdir,
         max_epochs=3
     )
 
diff --git a/tests/trainer/test_trainer.py b/tests/trainer/test_trainer.py
index 0c1063ea4ca8ee..f391260c139f42 100644
--- a/tests/trainer/test_trainer.py
+++ b/tests/trainer/test_trainer.py
@@ -30,12 +30,12 @@ def test_no_val_module(monkeypatch, tmpdir, tmpdir_server, url_ckpt):
     model = EvalModelTemplate()
 
     # logger file to get meta
-    logger = tutils.get_default_logger(str(tmpdir))
+    logger = tutils.get_default_logger(tmpdir)
 
     trainer = Trainer(
         max_epochs=1,
         logger=logger,
-        checkpoint_callback=ModelCheckpoint(str(tmpdir))
+        checkpoint_callback=ModelCheckpoint(tmpdir)
     )
     # fit model
     result = trainer.fit(model)
@@ -51,7 +51,7 @@ def test_no_val_module(monkeypatch, tmpdir, tmpdir_server, url_ckpt):
     assert LightningModule.CHECKPOINT_KEY_HYPER_PARAMS in ckpt.keys(), 'module_arguments missing from checkpoints'
 
     # load new model
-    hparams_path = tutils.get_data_path(logger, path_dir=str(tmpdir))
+    hparams_path = tutils.get_data_path(logger, path_dir=tmpdir)
     hparams_path = os.path.join(hparams_path, 'hparams.yaml')
     ckpt_path = f'http://{tmpdir_server[0]}:{tmpdir_server[1]}/{os.path.basename(new_weights_path)}' if url_ckpt else new_weights_path
     model_2 = EvalModelTemplate.load_from_checkpoint(
@@ -65,18 +65,18 @@ def test_no_val_module(monkeypatch, tmpdir, tmpdir_server, url_ckpt):
 def test_no_val_end_module(monkeypatch, tmpdir, tmpdir_server, url_ckpt):
     """Tests use case where trainer saves the model, and user loads it from tags independently."""
     # set $TORCH_HOME, which determines torch hub's cache path, to tmpdir
-    monkeypatch.setenv('TORCH_HOME', str(tmpdir))
+    monkeypatch.setenv('TORCH_HOME', tmpdir)
 
     model = EvalModelTemplate()
 
     # logger file to get meta
-    logger = tutils.get_default_logger(str(tmpdir))
+    logger = tutils.get_default_logger(tmpdir)
 
     # fit model
     trainer = Trainer(
         max_epochs=1,
         logger=logger,
-        checkpoint_callback=ModelCheckpoint(str(tmpdir))
+        checkpoint_callback=ModelCheckpoint(tmpdir)
     )
     result = trainer.fit(model)
 
@@ -88,7 +88,7 @@ def test_no_val_end_module(monkeypatch, tmpdir, tmpdir_server, url_ckpt):
     trainer.save_checkpoint(new_weights_path)
 
     # load new model
-    hparams_path = tutils.get_data_path(logger, path_dir=str(tmpdir))
+    hparams_path = tutils.get_data_path(logger, path_dir=tmpdir)
     hparams_path = os.path.join(hparams_path, 'hparams.yaml')
     ckpt_path = f'http://{tmpdir_server[0]}:{tmpdir_server[1]}/{os.path.basename(new_weights_path)}' if url_ckpt else new_weights_path
     model_2 = EvalModelTemplate.load_from_checkpoint(
@@ -163,7 +163,7 @@ def _optimizer_step(self, epoch, batch_idx, optimizer,
                       train_percent_check=0.1,
                       val_percent_check=0.1,
                       max_epochs=2,
-                      default_root_dir=str(tmpdir))
+                      default_root_dir=tmpdir)
 
     # for the test
     trainer.optimizer_step = _optimizer_step
@@ -179,13 +179,13 @@ def test_loading_meta_tags(tmpdir):
     hparams = EvalModelTemplate.get_default_hparams()
 
     # save tags
-    logger = tutils.get_default_logger(str(tmpdir))
+    logger = tutils.get_default_logger(tmpdir)
     logger.log_hyperparams(Namespace(some_str='a_str', an_int=1, a_float=2.0))
     logger.log_hyperparams(hparams)
     logger.save()
 
     # load hparams
-    path_expt_dir = tutils.get_data_path(logger, path_dir=str(tmpdir))
+    path_expt_dir = tutils.get_data_path(logger, path_dir=tmpdir)
     hparams_path = os.path.join(path_expt_dir, TensorBoardLogger.NAME_HPARAMS_FILE)
     hparams = load_hparams_from_yaml(hparams_path)
 
@@ -204,13 +204,13 @@ def test_loading_yaml(tmpdir):
     hparams = EvalModelTemplate.get_default_hparams()
 
     # save tags
-    logger = tutils.get_default_logger(str(tmpdir))
+    logger = tutils.get_default_logger(tmpdir)
     logger.log_hyperparams(Namespace(some_str='a_str', an_int=1, a_float=2.0))
     logger.log_hyperparams(hparams)
     logger.save()
 
     # load hparams
-    path_expt_dir = tutils.get_data_path(logger, path_dir=str(tmpdir))
+    path_expt_dir = tutils.get_data_path(logger, path_dir=tmpdir)
     hparams_path = os.path.join(path_expt_dir, 'hparams.yaml')
     tags = load_hparams_from_yaml(hparams_path)
 
@@ -262,7 +262,7 @@ def mock_save_function(filepath, *args):
     # simulated losses
     losses = [10, 9, 2.8, 5, 2.5]
 
-    checkpoint_callback = ModelCheckpoint(str(tmpdir), save_top_k=save_top_k, save_last=save_last,
+    checkpoint_callback = ModelCheckpoint(tmpdir, save_top_k=save_top_k, save_last=save_last,
                                           prefix=file_prefix, verbose=1)
     checkpoint_callback.save_function = mock_save_function
     trainer = Trainer()
@@ -291,7 +291,7 @@ def test_model_checkpoint_only_weights(tmpdir):
 
     trainer = Trainer(
         max_epochs=1,
-        checkpoint_callback=ModelCheckpoint(str(tmpdir), save_weights_only=True)
+        checkpoint_callback=ModelCheckpoint(tmpdir, save_weights_only=True)
     )
     # fit model
     result = trainer.fit(model)
@@ -367,8 +367,8 @@ def increment_on_load_checkpoint(self, _):
         max_epochs=2,
         train_percent_check=0.65,
         val_percent_check=1,
-        checkpoint_callback=ModelCheckpoint(str(tmpdir), save_top_k=-1),
-        default_root_dir=str(tmpdir),
+        checkpoint_callback=ModelCheckpoint(tmpdir, save_top_k=-1),
+        default_root_dir=tmpdir,
         early_stop_callback=False,
         val_check_interval=1.,
     )
@@ -423,7 +423,7 @@ def test_trainer_max_steps_and_epochs(tmpdir):
 
     # define less train steps than epochs
     trainer_options.update(
-        default_root_dir=str(tmpdir),
+        default_root_dir=tmpdir,
         max_epochs=3,
         max_steps=num_train_samples + 10
     )
@@ -458,7 +458,7 @@ def test_trainer_min_steps_and_epochs(tmpdir):
 
     # define callback for stopping the model and default epochs
     trainer_options.update(
-        default_root_dir=str(tmpdir),
+        default_root_dir=tmpdir,
         early_stop_callback=EarlyStopping(monitor='val_loss', min_delta=1.0),
         val_check_interval=2,
         min_epochs=1,
@@ -501,7 +501,7 @@ def test_benchmark_option(tmpdir):
 
     # fit model
     trainer = Trainer(
-        default_root_dir=str(tmpdir),
+        default_root_dir=tmpdir,
         max_epochs=1,
         benchmark=True,
     )
@@ -609,7 +609,7 @@ def training_step(self, batch, batch_idx, optimizer_idx=None):
 
     # fit model
     trainer = Trainer(
-        default_root_dir=str(tmpdir),
+        default_root_dir=tmpdir,
         max_steps=(model.test_batch_inf_loss + 1),
         terminate_on_nan=True
     )
@@ -634,7 +634,7 @@ def on_after_backward(self):
 
     model = CurrentModel()
     trainer = Trainer(
-        default_root_dir=str(tmpdir),
+        default_root_dir=tmpdir,
         max_steps=(model.test_batch_nan + 1),
         terminate_on_nan=True
     )
@@ -669,7 +669,7 @@ def on_batch_start(self, trainer, pl_module):
         train_percent_check=0.2,
         progress_bar_refresh_rate=0,
         logger=False,
-        default_root_dir=str(tmpdir),
+        default_root_dir=tmpdir,
     )
     assert not trainer.interrupted
     trainer.fit(model)
@@ -693,7 +693,7 @@ def _optimizer_step(*args, **kwargs):
         max_steps=1,
         max_epochs=1,
         gradient_clip_val=1.0,
-        default_root_dir=str(tmpdir)
+        default_root_dir=tmpdir
     )
 
     # for the test
@@ -705,7 +705,7 @@ def _optimizer_step(*args, **kwargs):
 
 def test_gpu_choice(tmpdir):
     trainer_options = dict(
-        default_root_dir=str(tmpdir),
+        default_root_dir=tmpdir,
     )
     # Only run if CUDA is available
     if not torch.cuda.is_available():
@@ -849,7 +849,7 @@ def __init__(self, **kwargs):
 def test_trainer_pickle(tmpdir):
     trainer = Trainer(
         max_epochs=1,
-        default_root_dir=str(tmpdir)
+        default_root_dir=tmpdir
     )
     pickle.dumps(trainer)
     cloudpickle.dumps(trainer)
diff --git a/tests/trainer/test_trainer_tricks.py b/tests/trainer/test_trainer_tricks.py
index 5f301aa8e0ced3..973ed32e7cd927 100755
--- a/tests/trainer/test_trainer_tricks.py
+++ b/tests/trainer/test_trainer_tricks.py
@@ -15,7 +15,7 @@ def test_model_reset_correctly(tmpdir):
 
     # logger file to get meta
     trainer = Trainer(
-        default_root_dir=str(tmpdir),
+        default_root_dir=tmpdir,
         max_epochs=1
     )
 
@@ -38,7 +38,7 @@ def test_trainer_reset_correctly(tmpdir):
 
     # logger file to get meta
     trainer = Trainer(
-        default_root_dir=str(tmpdir),
+        default_root_dir=tmpdir,
         max_epochs=1
     )
 
@@ -77,7 +77,7 @@ def test_trainer_arg(tmpdir, scale_arg):
     before_batch_size = hparams.get('batch_size')
     # logger file to get meta
     trainer = Trainer(
-        default_root_dir=str(tmpdir),
+        default_root_dir=tmpdir,
         max_epochs=1,
         auto_scale_batch_size=scale_arg,
     )
@@ -99,7 +99,7 @@ def test_call_to_trainer_method(tmpdir, scale_method):
     before_batch_size = hparams.get('batch_size')
     # logger file to get meta
     trainer = Trainer(
-        default_root_dir=str(tmpdir),
+        default_root_dir=tmpdir,
         max_epochs=1,
     )
 
@@ -118,7 +118,7 @@ def test_error_on_dataloader_passed_to_fit(tmpdir):
     # only train passed to fit
     model = EvalModelTemplate()
     trainer = Trainer(
-        default_root_dir=str(tmpdir),
+        default_root_dir=tmpdir,
         max_epochs=1,
         val_percent_check=0.1,
         train_percent_check=0.2,