data = xr.open_dataset(predictor_file, chunks={'sample': batch_size})
if 'time_step' in data.dims:
time_dim = data.dims['time_step']
else:
time_dim = 1
n_sample = data.dims['sample']
if crop_north_pole:
data = data.isel(lat=(data.lat < 90.0))
#%% Build a model and the data generators
dlwp = DLWPNeuralNet(is_convolutional=model_is_convolutional,
is_recurrent=model_is_recurrent,
time_dim=time_dim,
scaler_type=None,
scale_targets=False)
# Find the validation set
if isinstance(validation_set, int):
n_sample = data.dims['sample']
ts, val_set = train_test_split_ind(n_sample, validation_set, method='last')
if train_set is None:
train_set = ts
elif isinstance(train_set, int):
train_set = list(range(train_set))
validation_data = data.isel(sample=val_set)
train_data = data.isel(sample=train_set)
elif validation_set is None:
if train_set is None: