Prevent Triton from getting eagerly imported when importing torch._inductor (#100374)
This makes 'import torch._inductor.utils' go from 3.5s to 2.1s
See also https://github.com/openai/triton/issues/1599
Signed-off-by: Edward Z. Yang <ezyang@meta.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/100374
Approved by: https://github.com/voznesenskym