Just wondering if anyone had any resources for learning about sparse neural nets. I am not worried about the possibility of speed improvements, and would actually be fine with a small speed decrease. I am interested in sparse operations for the possibility of decrease over memory requirements of large matrix multiplies. Therefore I am looking for something that implements it in a “truly sparse” way as opposed to mm+mask. I am not sure how pytorch implements sparse mm, or if anyone has found a good resource for learning about using sparse operations in pytorch.