3 ms·

The plan is to eventually implement with CUDA: "Currently, I am working on [...] direct CUDA implementation, which will be significantly faster and probably co

by dnlzro 3y ago

The plan is to eventually implement with CUDA: "Currently, I am working on [...] direct CUDA implementation, which will be significantly faster and probably come close to PyTorch."