photontorch
photontorch copied to clipboard
Integrating sparse tensors
A lot of memory usage can probably be avoided when transitioning to sparse tensor representations under the hood.