def l1_loss(tensor, batch_norm=True, reduce_batches=True):
if struct.isstruct(tensor):
all_tensors = struct.flatten(tensor)
return sum(l1_loss(tensor, batch_norm, reduce_batches) for tensor in all_tensors)
if reduce_batches:
total_loss = math.sum(math.abs(tensor))
else:
total_loss = math.sum(math.abs(tensor), axis=list(range(1, len(tensor.shape))))
if batch_norm and reduce_batches:
batch_size = math.shape(tensor)[0]
return math.div(total_loss, math.to_float(batch_size))
else:
return total_loss
def _max_residual_condition(residual_index, accuracy):
"""continue if the maximum deviation from zero is bigger than desired accuracy"""
if accuracy is None:
return lambda *args: True
else:
return lambda *args: math.max(math.abs(args[residual_index])) > accuracy
def loop_condition(_1, _2, _3, residual, _i):
'''continue if the maximum deviation from zero is bigger than desired accuracy'''
return math.max(math.abs(residual)) >= accuracy