csarofeen / pytorch

Tensors and Dynamic neural networks in Python with strong GPU acceleration
http://pytorch.org
Other
26 stars 7 forks source link

disable welford translate if the original scheduler before translate is not persistent #2508

Closed liqiangxl closed 1 year ago

liqiangxl commented 1 year ago

🚀 The feature, motivation and pitch

simple cases such as var_mean will use reduction scheduler and then change to persistent scheduler after translate. However, this translation doesn't lead to performance change but introduced additional lowering overhead.

Alternatives

No response

Additional context

No response