def test_no_u(self, tmpdir):
sys.argv = [
'train',
'--output_folder', str(tmpdir),
]
args = parse_args()
assert len(args.u_connect) == 0
def test_one_to_many(self, tmpdir):
key1 = '3d_cont'
key2 = '3d_cat'
sys.argv = [
'train',
'--output_folder', str(tmpdir),
'--input_tensors', key1, key2,
'--output_tensors', key1, key2,
'--u_connect', key1, key1,
'--u_connect', key1, key2,
]
args = parse_args()
assert len(args.u_connect) == 1
assert args.u_connect[MOCK_TMAPS[key1]] == {MOCK_TMAPS[key1], MOCK_TMAPS[key2]}
def test_many_to_one(self, tmpdir):
inp_key1 = '3d_cont'
inp_key2 = '3d_cat'
sys.argv = [
'train',
'--output_folder', str(tmpdir),
'--input_tensors', inp_key1, inp_key2,
'--output_tensors', inp_key1,
'--u_connect', inp_key1, inp_key1,
'--u_connect', inp_key2, inp_key1,
]
args = parse_args()
assert len(args.u_connect) == 2
assert args.u_connect[MOCK_TMAPS[inp_key1]] == {MOCK_TMAPS[inp_key1]}
assert args.u_connect[MOCK_TMAPS[inp_key2]] == {MOCK_TMAPS[inp_key1]}
def test_simple_u(self, tmpdir):
inp_key = '3d_cont'
sys.argv = [
'train',
'--output_folder', str(tmpdir),
'--input_tensors', inp_key,
'--output_tensors', inp_key,
'--u_connect', inp_key, inp_key,
]
args = parse_args()
assert len(args.u_connect) == 1
inp, out = list(args.u_connect.items())[0]
tmap = MOCK_TMAPS[inp_key]
assert inp == tmap
assert out == {tmap, }
def default_arguments(tmpdir_factory):
temp_dir = tmpdir_factory.mktemp('data')
build_hdf5s(temp_dir, MOCK_TMAPS.values(), n=pytest.N_TENSORS)
hdf5_dir = str(temp_dir)
inp_key = '3d_cont'
out_key = '1d_cat'
sys.argv = [
'',
'--output_folder',
hdf5_dir,
'--input_tensors',
inp_key,
'--output_tensors',
out_key,
'--tensors',
hdf5_dir,
'--pool_x',
'1',
'--pool_y',
'1',
'--pool_z',
'1',
'--training_steps',
'2',
'--test_steps',
'3',
'--validation_steps',
'2',
'--epochs',
'2',
'--num_workers',
'0',
'--batch_size',
'2',
]
args = parse_args()
return args
ax2.plot(val_loss,
label=label,
linestyle=linestyles[i % 4],
color=color)
ax2.text(len(val_loss) - 1, val_loss[-1], str(i))
ax1.axhline(cutoff,
label=f'Loss display cutoff at {cutoff:.3f}',
color='k',
linestyle='--')
ax1.set_title('Training Loss')
ax2.axhline(cutoff,
label=f'Loss display cutoff at {cutoff:.3f}',
color='k',
linestyle='--')
ax2.set_title('Validation Loss')
ax3.legend(*ax2.get_legend_handles_labels(),
loc='upper center',
fontsize='x-small',
mode='expand',
ncol=5)
ax3.axis('off')
learning_path = os.path.join(figure_path, 'learning_curves' + IMAGE_EXT)
plt.tight_layout()
plt.savefig(learning_path)
logging.info('Saved learning curve plot to: {}'.format(learning_path))
if __name__ == '__main__':
args = parse_args()
run(args) # back to the top