def test_trainer_validation(configuration):
# start model
model = Model(
forecast_minutes=configuration.input_data.default_forecast_minutes,
history_minutes=configuration.input_data.default_history_minutes,
output_variable="gsp_yield",
)
# create fake data loader
train_dataset = FakeDataset(configuration=configuration)
train_dataloader = torch.utils.data.DataLoader(train_dataset,
batch_size=None)
# set up trainer
trainer = pl.Trainer(gpus=0, max_epochs=1)
with tempfile.TemporaryDirectory() as tmpdirname:
model.results_file_name = f'{tmpdirname}/temp'
# test over validation set
_ = trainer.validate(model, train_dataloader)
# check csv file of validation results has been made
results_df = pd.read_csv(f'{model.results_file_name}_0.csv')
assert len(results_df) == len(
train_dataloader
) * configuration.process.batch_size * model.forecast_len_30
assert 't0_datetime_utc' in results_df.keys()
assert 'target_datetime_utc' in results_df.keys()
assert 'gsp_id' in results_df.keys()
assert "actual_gsp_pv_outturn_mw" in results_df.keys()
assert "forecast_gsp_pv_outturn_mw" in results_df.keys()
def test_model_forward_no_satellite(configuration_conv3d):
config_file = "tests/configs/model/conv3d_sat_nwp.yaml"
config = load_config(config_file)
config['include_future_satellite'] = False
# start model
model = Model(**config)
dataset_configuration = configuration_conv3d
dataset_configuration.input_data.nwp.nwp_image_size_pixels = 16
# create fake data loader
train_dataset = FakeDataset(configuration=dataset_configuration)
train_dataloader = torch.utils.data.DataLoader(train_dataset,
batch_size=None)
x = next(iter(train_dataloader))
# run data through model
y = model(x)
# check out put is the correct shape
assert len(y.shape) == 2
assert y.shape[0] == 2
assert y.shape[1] == model.forecast_len_30
def test_model_forward(configuration):
# start model
model = Model(
forecast_minutes=configuration.input_data.default_forecast_minutes,
history_minutes=configuration.input_data.default_history_minutes,
output_variable="gsp_yield",
)
# create fake data loader
train_dataset = FakeDataset(configuration=configuration)
train_dataloader = torch.utils.data.DataLoader(train_dataset,
batch_size=None)
# satellite data
x = next(iter(train_dataloader))
# run data through model
y = model(x)
# check out put is the correct shape
assert len(y.shape) == 2
assert y.shape[0] == configuration.process.batch_size
assert y.shape[
1] == configuration.input_data.default_forecast_minutes // 30
def test_model_forward(configuration_perceiver):
dataset_configuration = configuration_perceiver
dataset_configuration.input_data.nwp.nwp_image_size_pixels = 64
dataset_configuration.input_data.satellite.satellite_image_size_pixels = 16
model = PerceiverModel(history_minutes=30,
forecast_minutes=60,
nwp_channels=params["nwp_channels"],
embedding_dem=2048) # doesnt do anything
# set up fake data
train_dataset = FakeDataset(configuration=dataset_configuration)
train_dataloader = torch.utils.data.DataLoader(train_dataset,
batch_size=None)
# get data
x = next(iter(train_dataloader))
# run data through model
y = model(x)
# check out put is the correct shape
assert len(y.shape) == 2
assert y.shape[0] == dataset_configuration.process.batch_size
assert y.shape[1] == 60 // 5
def val_dataloader(self):
if self.fake_data:
val_dataset = FakeDataset(configuration=self.configuration)
else:
val_dataset = NetCDFDataset(self.n_val_data,
os.path.join(self.data_path, "test"),
os.path.join(self.temp_path, "test"),
configuration=self.configuration)
return torch.utils.data.DataLoader(val_dataset,
**self.dataloader_config)
def test_dataloader(self):
if self.fake_data:
test_dataset = FakeDataset(configuration=self.configuration)
else:
# TODO need to change this to a test folder
test_dataset = NetCDFDataset(self.n_val_data,
os.path.join(self.data_path, "test"),
os.path.join(self.temp_path, "test"),
configuration=self.configuration)
return torch.utils.data.DataLoader(test_dataset,
**self.dataloader_config)
def test_train(configuration_conv3d):
config_file = "tests/configs/model/conv3d.yaml"
config = load_config(config_file)
dataset_configuration = configuration_conv3d
# start model
model = Model(**config)
# create fake data loader
train_dataset = FakeDataset(configuration=dataset_configuration)
train_dataset.length = 2
train_dataloader = torch.utils.data.DataLoader(train_dataset,
batch_size=None)
# fit model
trainer = pl.Trainer(gpus=0, max_epochs=1)
trainer.fit(model, train_dataloader)
# predict over training set
_ = trainer.predict(model, train_dataloader)
def test_trainer(configuration):
# start model
model = Model(
forecast_minutes=configuration.input_data.default_forecast_minutes,
history_minutes=configuration.input_data.default_history_minutes,
output_variable="gsp_yield",
)
# create fake data loader
train_dataset = FakeDataset(configuration=configuration)
train_dataloader = torch.utils.data.DataLoader(train_dataset,
batch_size=None)
# set up trainer
trainer = pl.Trainer(gpus=0, max_epochs=1)
# test over training set
_ = trainer.test(model, train_dataloader)
def test_model_validation(configuration):
# start model
model = Model(
forecast_minutes=configuration.input_data.default_forecast_minutes,
history_minutes=configuration.input_data.default_history_minutes,
output_variable="gsp_yield",
)
# create fake data loader
train_dataset = FakeDataset(configuration=configuration)
train_dataloader = torch.utils.data.DataLoader(train_dataset,
batch_size=None)
# satellite data
x = next(iter(train_dataloader))
# run data through model
model.validation_step(x, 0)
def test_model_forward(configuration_conv3d):
config_file = "tests/configs/model/conv3d.yaml"
config = load_config(config_file)
dataset_configuration = configuration_conv3d
# start model
model = Model(**config)
# create fake data loader
train_dataset = FakeDataset(configuration=dataset_configuration)
train_dataloader = torch.utils.data.DataLoader(train_dataset,
batch_size=None)
x = next(iter(train_dataloader))
# run data through model
y = model(x)
# check out put is the correct shape
assert len(y.shape) == 2
assert y.shape[0] == 2
assert y.shape[1] == model.forecast_len_5