Batch to space / space to batch perf improvements (again)

microsoft / tensorflow-directml

Fork of TensorFlow accelerated by DirectML

Apache License 2.0

457 stars 32 forks source link

Batch to space / space to batch perf improvements (again) #328

Closed Jamather closed 2 years ago

Jamather commented 3 years ago

I added an extra exception for api compatibility when block sizes are not all the same, since DML does not support this. This fixes the test failures associated with the change.