Tests v0 3 5

[pattonw]

Fixed a couple bugs

1. CNNectome UNet wasn't following the api defined by the Architecture ABC. It was supposed to be returning Coordinate class instances for voxel_size, input_shape etc.
2. CNNectome UNet had logical errors in the definition of kernel_size_up and down.
3. New tests was expecting multi channel data for the 2D UNet and single channel for the 3D unet despite always getting single channel data.
4. fix voxel_size attr in test fixtures

[pattonw]

There are still some tests failing.
They fall into two categories

1. (size of tensor a (x) must match the size of tensor b (x//2). This is caused by the upsample test. We have given raw data and gt data at the same resolution, but then asked the model to upsample and compare to the non upsampled gt. We get a resolution mis match and the shapes will not be comparable between upsampled output and non upsampled target.
2. AssertionError tensor(y). This seems to be caused by an assumption that the weights should change more than some minimum amount in a few test iterations. I'm not sure if this assumption holds.

FAILED tests/operations/test_train.py::test_train_unet[True-True-True-True-gunpowder_trainer-distance_task-six_class_datasplit] - RuntimeError: The size of tensor a (188) must match the size of tensor b (94) at non-singleton dimensi...
FAILED tests/operations/test_train.py::test_train_unet[True-True-True-False-gunpowder_trainer-distance_task-six_class_datasplit] - RuntimeError: The size of tensor a (188) must match the size of tensor b (94) at non-singleton dimensi...
FAILED tests/operations/test_train.py::test_train_unet[True-True-False-True-gunpowder_trainer-distance_task-six_class_datasplit] - AssertionError: tensor(0.)
FAILED tests/operations/test_train.py::test_train_unet[True-False-True-True-gunpowder_trainer-distance_task-six_class_datasplit] - RuntimeError: The size of tensor a (188) must match the size of tensor b (94) at non-singleton dimensi...
FAILED tests/operations/test_train.py::test_train_unet[True-False-True-False-gunpowder_trainer-distance_task-six_class_datasplit] - RuntimeError: The size of tensor a (188) must match the size of tensor b (94) at non-singleton dimensi...
FAILED tests/operations/test_train.py::test_train_unet[True-False-False-True-gunpowder_trainer-distance_task-six_class_datasplit] - AssertionError: tensor(0.)
FAILED tests/operations/test_train.py::test_train_unet[False-True-True-True-gunpowder_trainer-distance_task-six_class_datasplit] - RuntimeError: The size of tensor a (28) must match the size of tensor b (14) at non-singleton dimension 4
FAILED tests/operations/test_train.py::test_train_unet[False-True-True-False-gunpowder_trainer-distance_task-six_class_datasplit] - RuntimeError: The size of tensor a (28) must match the size of tensor b (14) at non-singleton dimension 4
FAILED tests/operations/test_train.py::test_train_unet[False-True-False-True-gunpowder_trainer-distance_task-six_class_datasplit] - AssertionError: tensor(1.1921e-07)
FAILED tests/operations/test_train.py::test_train_unet[False-False-True-True-gunpowder_trainer-distance_task-six_class_datasplit] - RuntimeError: The size of tensor a (28) must match the size of tensor b (14) at non-singleton dimension 4
FAILED tests/operations/test_train.py::test_train_unet[False-False-True-False-gunpowder_trainer-distance_task-six_class_datasplit] - RuntimeError: The size of tensor a (28) must match the size of tensor b (14) at non-singleton dimension 4
FAILED tests/operations/test_train.py::test_train_unet[False-False-False-True-gunpowder_trainer-distance_task-six_class_datasplit] - AssertionError: tensor(5.9605e-08)

[mzouink]

nice, can you please review tha changes i did in the tests_v0_3_5 branch