pytorch-sqrtm
pytorch-sqrtm copied to clipboard
How to accelerate the running speed
I try to calculate sqrtm by this method, but soon I found it so slow to run, How can I accelerate the running speed?
I dont think we can do much here. Most of the latency s behould due to converting your data from GPU to cpu back and forth and not necessarily suboptimal code. If you know a better implementation please let meknow.
I find discussions in this thread useful: https://github.com/pytorch/pytorch/issues/25481