higher/optim.py (1 line):
	- line 530: # TODO: implement support for sparse gradients.