Skip to content

First CUDA release

Compare
Choose a tag to compare
@Manza12 Manza12 released this 20 Apr 11:14
· 110 commits to main since this release
8e7931c

In this release, nnMorpho changed the paradigm from the torch.Tensor.unfold strategy to pure CUDA code.
The memory usage is now optimal and the speed imporvement relative to Scipy is of around x200.