def assert_model_trains(input_tmaps: List[TensorMap],
output_tmaps: List[TensorMap],
m: Optional[tf.keras.Model] = None,
skip_shape_check: bool = False):
if m is None:
m = make_multimodal_multitask_model(
input_tmaps,
output_tmaps,
**DEFAULT_PARAMS,
)
if not skip_shape_check:
for tmap, tensor in zip(input_tmaps, m.inputs):
assert tensor.shape[1:] == tmap.shape
assert tensor.shape[1:] == tmap.shape
for tmap, tensor in zip(parent_sort(output_tmaps), m.outputs):
assert tensor.shape[1:] == tmap.shape
assert tensor.shape[1:] == tmap.shape
data = make_training_data(input_tmaps, output_tmaps)
history = m.fit(data,
steps_per_epoch=2,
epochs=2,
validation_data=data,
validation_steps=2)
for tmap in output_tmaps:
for metric in tmap.metrics:
metric_name = metric if type(metric) == str else metric.__name__
name = f'{tmap.output_name()}_{metric_name}' if len(
output_tmaps) > 1 else metric_name
assert name in history.history
def _process_args(args):
now_string = datetime.datetime.now().strftime('%Y-%m-%d_%H-%M')
args_file = os.path.join(args.output_folder, args.id, 'arguments_' + now_string + '.txt')
command_line = f"\n./scripts/tf.sh {' '.join(sys.argv)}\n"
if not os.path.exists(os.path.dirname(args_file)):
os.makedirs(os.path.dirname(args_file))
with open(args_file, 'w') as f:
f.write(command_line)
for k, v in sorted(args.__dict__.items(), key=operator.itemgetter(0)):
f.write(k + ' = ' + str(v) + '\n')
load_config(args.logging_level, os.path.join(args.output_folder, args.id), 'log_' + now_string, args.min_sample_id)
args.u_connect = _process_u_connect_args(args.u_connect, args.tensormap_prefix)
args.pairs = _process_pair_args(args.pairs, args.tensormap_prefix)
args.tensor_maps_in = []
args.tensor_maps_out = []
if args.text_file is not None:
del args.input_tensors[:2]
del args.output_tensors[0]
input_map, burn_in, output_map = generate_random_text_tensor_maps(args.text_file, args.text_window, args.text_one_hot)
if args.text_one_hot:
args.tensor_maps_in.append(input_map)
else:
args.tensor_maps_in.extend([input_map, burn_in])
args.tensor_maps_out.append(output_map)
args.tensor_maps_in.extend([tensormap_lookup(it, args.tensormap_prefix) for it in args.input_tensors])
if args.continuous_file is not None:
# Continuous TensorMap generated from file is given the name specified by the first output_tensors argument
args.tensor_maps_out.append(
generate_continuous_tensor_map_from_file(
args.continuous_file,
args.continuous_file_column,
args.output_tensors.pop(0),
args.continuous_file_normalize,
args.continuous_file_discretization_bounds,
),
)
args.tensor_maps_out.extend([tensormap_lookup(ot, args.tensormap_prefix) for ot in args.output_tensors])
args.tensor_maps_out = parent_sort(args.tensor_maps_out)
args.tensor_maps_protected = [tensormap_lookup(it, args.tensormap_prefix) for it in args.protected_tensors]
args.sample_weight = tensormap_lookup(args.sample_weight, args.tensormap_prefix) if args.sample_weight else None
if args.sample_weight:
assert args.sample_weight.shape == (1,)
args.bottleneck_type = BOTTLENECK_STR_TO_ENUM[args.bottleneck_type]
if args.bottleneck_type == BottleneckType.NoBottleNeck:
check_no_bottleneck(args.u_connect, args.tensor_maps_out)
if args.learning_rate_schedule is not None and args.patience < args.epochs:
raise ValueError(f'learning_rate_schedule is not compatible with ReduceLROnPlateau. Set patience > epochs.')
np.random.seed(args.random_seed)
logging.info(f"Command Line was: {command_line}")
logging.info(f"Arguments are {args}\n")
if args.eager:
import tensorflow as tf
tf.config.experimental_run_functions_eagerly(True)
def test_parent_sort_cycle(tmaps):
with pytest.raises(ValueError):
parent_sort(tmaps)
def test_parent_sort_idempotent(tmaps):
assert parent_sort(tmaps) == parent_sort(
parent_sort(tmaps)) == parent_sort(parent_sort(parent_sort(tmaps)))
def test_parent_sort(tmaps):
assert parent_sort(tmaps) == PARENT_TMAPS