v2.4.2
This is a patch release containing the following changes to v2.4.1:
- Fixed performance regression for convolution primitive for the shapes with 3D spatial for future Intel Xeon Scalable processor (code name Sapphire Rapids) (aca0af1)
- Fixed segmentation fault in bfloat16 forward and backward inner product primitive or future Intel Xeon Scalable processor (code name Sapphire Rapids) (ae8cf18, 3de9549)
- Fixed reorder primitive with compensation (6ba086a)
- Fixed issue in scratch pad size calculation for BRGEMM-based convolutions (dd9eceb)