They were using float64 for FP32 kernel before.tonyyang-svail-feed-op-desgin
parent
55b68c6e90
commit
6efcbc4fcb
Loading…
Reference in new issue