CannyLab / tsne-cuda

GPU Accelerated t-SNE for CUDA with Python bindings
BSD 3-Clause "New" or "Revised" License
1.78k stars 126 forks source link

tsnecuda equivalent for M1 chip #113

Closed abis330 closed 1 year ago

abis330 commented 1 year ago

Is it possible to take advantage of M1 chip GPU and Neural Engine ecosystem for running tsne efficiently?

DavidMChan commented 1 year ago

It would be possible - but it would require a very different architecture to what we provide in this repository. Since the M1 relies on the metal libraries for code acceleration (and not CUDA), we would have to re-write the core code for this repository to support M1 (which isn't feasible at this time).