acoustic-van-30942
05/03/2023, 10:25 PMtrn1
instances. I do see support for inferentia. Our use case is we want to try refactoring one of our customer's models to leverage AWS neuron sdk just to see what the performance gain is.
Looking through these docs, it appears like there would be some retooling involved to get Metaflow's Batch decorator compatible with trn1
.