data = voldata_train
labels = vollabels_train
volume_shape = data.shape
crop_seq_train = VolumeCropSequence(
data_volume=data,
labels_volume=labels,
batch_size=batch_size,
meta_crop_generator=MetaCrop3DGenerator(
volume_shape=volume_shape,
crop_shape=crop_shape,
is_2halfd=(model_type == T2SModelType.d2half),
x0y0z0_generator=(grid_pos_gen := t2s_volseq.UniformGridPosition.build_from_volume_crop_shapes(
volume_shape=volume_shape,
crop_shape=crop_shape,
random_state=RandomState(args.random_state_seed),
)),
gt_field=t2s_volseq.GTUniformEverywhere(
gt_type=gt_type,
grid_position_generator=grid_pos_gen,
random_state=RandomState(args.random_state_seed),
),
et_field=t2s_volseq.ET3DConstantEverywhere.build_no_displacement(grid_position_generator=grid_pos_gen),
grid_pos_gen = t2s_volseq.UniformGridPosition.build_from_volume_crop_shapes(
volume_shape=volume_shape,
crop_shape=crop_shape,
random_state=RandomState(args.random_state_seed),
)
crop_seq_train = VolumeCropSequence(
data_volume=data,
labels_volume=labels,
batch_size=batch_size,
# data augmentation
meta_crop_generator=MetaCrop3DGenerator.build_no_augmentation(
grid_pos_gen=grid_pos_gen,
volume_shape=volume_shape,
crop_shape=crop_shape,
common_random_state_seed=args.random_state_seed,
gt_type=gt_type,
is_2halfd=(model_type == T2SModelType.d2half),
),
# this volume cropper only returns random crops,
# so the number of crops per epoch/batch is w/e i want
epoch_size=10,
**vol_crop_seq_common_kwargs,
meta_crops_hist_path=t2s_model.train_metacrop_history_path,
)
# ## Val
data = voldata_val
labels = vollabels_val