Construction of MultivariateNormal much slower on GPU than CPU #23780
Labels
module: cuda
Related to torch.cuda, and CUDA support in general
module: distributions
Related to torch.distributions
module: performance
Issues related to performance, either of kernel code or framework glue
triaged
This issue has been looked at a team member, and triaged and prioritized into an appropriate module
馃悰 Bug
Constructing a MultivariateNormal distribution is much slower when inputting GPU-based
FloatTensor
s than CPU-based ones.On my machine the GPU version is ~33x slower than CPU.
To Reproduce
Steps to reproduce the behavior:
Output on my machine:
Expected behavior
I'd expect the GPU to be faster, or at least of a comparable speed to CPU.
Environment
The text was updated successfully, but these errors were encountered: