tests

Author: Borda

tests/accelerators/legacy/test_ddp_spawn.py

@@ -21,6 +21,9 @@
 from pytorch_lightning.trainer import Trainer
 from pytorch_lightning.trainer.states import TrainerState
 from tests.base import EvalModelTemplate
+from tests.helpers import BoringModel
+from tests.helpers.datamodules import ClassifDataModule
+from tests.helpers.simple_models import ClassificationModel
 
 
 @pytest.mark.skipif(torch.cuda.device_count() < 2, reason="test requires multi-GPU machine")
@@ -38,8 +41,9 @@ def test_multi_gpu_early_stop_ddp_spawn(tmpdir):
         accelerator='ddp_spawn',
     )
 
-    model = EvalModelTemplate()
-    tpipes.run_model_test(trainer_options, model)
+    dm = ClassifDataModule()
+    model = ClassificationModel()
+    tpipes.run_model_test(trainer_options, model, dm)
 
 
 @pytest.mark.skipif(torch.cuda.device_count() < 2, reason="test requires multi-GPU machine")
@@ -56,7 +60,7 @@ def test_multi_gpu_model_ddp_spawn(tmpdir):
         progress_bar_refresh_rate=0,
     )
 
-    model = EvalModelTemplate()
+    model = BoringModel()
 
     tpipes.run_model_test(trainer_options, model)
 

tests/accelerators/legacy/test_dp.py

@@ -20,6 +20,9 @@
 from pytorch_lightning.callbacks import EarlyStopping
 from pytorch_lightning.core import memory
 from tests.base import EvalModelTemplate
+from tests.helpers import BoringModel
+from tests.helpers.datamodules import ClassifDataModule
+from tests.helpers.simple_models import ClassificationModel
 
 PRETEND_N_OF_GPUS = 16
 
@@ -39,8 +42,9 @@ def test_multi_gpu_early_stop_dp(tmpdir):
         accelerator='dp',
     )
 
-    model = EvalModelTemplate()
-    tpipes.run_model_test(trainer_options, model)
+    dm = ClassifDataModule()
+    model = ClassificationModel()
+    tpipes.run_model_test(trainer_options, model, dm)
 
 
 @pytest.mark.skipif(torch.cuda.device_count() < 2, reason="test requires multi-GPU machine")
@@ -57,7 +61,7 @@ def test_multi_gpu_model_dp(tmpdir):
         progress_bar_refresh_rate=0,
     )
 
-    model = EvalModelTemplate()
+    model = BoringModel()
 
     tpipes.run_model_test(trainer_options, model)
 

tests/helpers/pipelines.py

@@ -54,7 +54,13 @@ def run_model_test_without_loggers(
 
 
 def run_model_test(
-    trainer_options, model, on_gpu: bool = True, version=None, with_hpc: bool = True, min_acc: float = 0.25
+    trainer_options,
+    model: LightningModule,
+    data: LightningDataModule = None,
+    on_gpu: bool = True,
+    version=None,
+    with_hpc: bool = True,
+    min_acc: float = 0.25
 ):
 
     reset_seed()
@@ -66,7 +72,7 @@ def run_model_test(
 
     trainer = Trainer(**trainer_options)
     initial_values = torch.tensor([torch.sum(torch.abs(x)) for x in model.parameters()])
-    trainer.fit(model)
+    trainer.fit(model, datamodule=data)
     post_train_values = torch.tensor([torch.sum(torch.abs(x)) for x in model.parameters()])
 
     assert trainer.state == TrainerState.FINISHED, f"Training failed with {trainer.state}"
@@ -78,7 +84,7 @@ def run_model_test(
     pretrained_model = load_model_from_checkpoint(logger, trainer.checkpoint_callback.best_model_path, type(model))
 
     # test new model accuracy
-    test_loaders = model.test_dataloader()
+    test_loaders = model.test_dataloader() if not data else data.test_dataloader()
     if not isinstance(test_loaders, list):
         test_loaders = [test_loaders]
 

tests/models/test_amp.py

@@ -25,6 +25,7 @@
 from pytorch_lightning.utilities import _APEX_AVAILABLE
 from pytorch_lightning.utilities.exceptions import MisconfigurationException
 from tests.base import EvalModelTemplate
+from tests.helpers import BoringModel
 
 
 @pytest.mark.skip(reason='dp + amp not supported currently')  # TODO
@@ -156,7 +157,7 @@ def test_cpu_model_with_amp(tmpdir):
         precision=16,
     )
 
-    model = EvalModelTemplate()
+    model = BoringModel()
 
     with pytest.raises((MisconfigurationException, ModuleNotFoundError)):
         tpipes.run_model_test(trainer_options, model, on_gpu=False)

tests/models/test_cpu.py

@@ -23,7 +23,6 @@
 from pytorch_lightning import Trainer
 from pytorch_lightning.callbacks import Callback, EarlyStopping, ModelCheckpoint
 from pytorch_lightning.trainer.states import TrainerState
-from tests.base import EvalModelTemplate
 from tests.helpers import BoringModel
 from tests.helpers.datamodules import ClassifDataModule
 from tests.helpers.simple_models import ClassificationModel
@@ -304,7 +303,7 @@ def test_cpu_model(tmpdir):
         limit_val_batches=0.4
     )
 
-    model = EvalModelTemplate()
+    model = BoringModel()
 
     tpipes.run_model_test(trainer_options, model, on_gpu=False)
 

tests/models/test_gpu.py

@@ -25,6 +25,7 @@
 from pytorch_lightning.utilities import device_parser
 from pytorch_lightning.utilities.exceptions import MisconfigurationException
 from tests.helpers import BoringModel
+from tests.helpers.datamodules import ClassifDataModule
 from tests.helpers.simple_models import ClassificationModel
 
 PRETEND_N_OF_GPUS = 16
@@ -43,8 +44,9 @@ def test_multi_gpu_none_backend(tmpdir):
         gpus=2,
     )
 
+    dm = ClassifDataModule()
     model = ClassificationModel()
-    tpipes.run_model_test(trainer_options, model)
+    tpipes.run_model_test(trainer_options, model, dm)
 
 
 @pytest.mark.skipif(torch.cuda.device_count() < 2, reason="test requires multi-GPU machine")

tests/models/test_onnx.py

@@ -92,7 +92,7 @@ def test_model_saves_on_multi_gpu(tmpdir):
         progress_bar_refresh_rate=0,
     )
 
-    model = EvalModelTemplate()
+    model = BoringModel()
 
     tpipes.run_model_test(trainer_options, model)
 

tests/models/test_tpu.py

@@ -26,6 +26,7 @@
 from pytorch_lightning.utilities import _TPU_AVAILABLE
 from pytorch_lightning.utilities.exceptions import MisconfigurationException
 from tests.base import EvalModelTemplate
+from tests.helpers import BoringModel
 from tests.helpers.datasets import TrialMNIST
 from tests.helpers.utils import pl_multi_process_test
 
@@ -55,7 +56,7 @@ def test_model_tpu_cores_1(tmpdir):
         limit_val_batches=0.4,
     )
 
-    model = EvalModelTemplate()
+    model = BoringModel()
     tpipes.run_model_test(trainer_options, model, on_gpu=False, with_hpc=False)
 
 
@@ -73,7 +74,7 @@ def test_model_tpu_index(tmpdir, tpu_core):
         limit_val_batches=0.4,
     )
 
-    model = EvalModelTemplate()
+    model = BoringModel()
     tpipes.run_model_test(trainer_options, model, on_gpu=False, with_hpc=False)
     assert torch_xla._XLAC._xla_get_default_device() == f'xla:{tpu_core}'
 
@@ -113,7 +114,7 @@ def test_model_16bit_tpu_cores_1(tmpdir):
         limit_val_batches=0.4,
     )
 
-    model = EvalModelTemplate()
+    model = BoringModel()
     tpipes.run_model_test(trainer_options, model, on_gpu=False)
     assert os.environ.get('XLA_USE_BF16') == str(1), "XLA_USE_BF16 was not set in environment variables"
 
@@ -133,7 +134,7 @@ def test_model_16bit_tpu_index(tmpdir, tpu_core):
         limit_val_batches=0.2,
     )
 
-    model = EvalModelTemplate()
+    model = BoringModel()
     tpipes.run_model_test(trainer_options, model, on_gpu=False)
     assert torch_xla._XLAC._xla_get_default_device() == f'xla:{tpu_core}'
     assert os.environ.get('XLA_USE_BF16') == str(1), "XLA_USE_BF16 was not set in environment variables"
@@ -192,19 +193,25 @@ def test_tpu_grad_norm(tmpdir):
         gradient_clip_val=0.1,
     )
 
-    model = EvalModelTemplate()
+    model = BoringModel()
     tpipes.run_model_test(trainer_options, model, on_gpu=False, with_hpc=False)
 
 
 @pytest.mark.skipif(not _TPU_AVAILABLE, reason="test requires TPU machine")
 @pl_multi_process_test
 def test_dataloaders_passed_to_fit(tmpdir):
     """Test if dataloaders passed to trainer works on TPU"""
-
-    model = EvalModelTemplate()
-
-    trainer = Trainer(default_root_dir=tmpdir, max_epochs=1, tpu_cores=8)
-    trainer.fit(model, train_dataloader=model.train_dataloader(), val_dataloaders=model.val_dataloader())
+    model = BoringModel()
+    trainer = Trainer(
+        default_root_dir=tmpdir,
+        max_epochs=1,
+        tpu_cores=8,
+    )
+    trainer.fit(
+        model,
+        train_dataloader=model.train_dataloader(),
+        val_dataloaders=model.val_dataloader(),
+    )
     assert trainer.state == TrainerState.FINISHED, f"Training failed with {trainer.state}"