pkg syscall (freebsd-arm64), const DLT_LAPB_WITH_DIR ideal-int
pkg syscall (freebsd-arm64), const DLT_LAPD = 203
pkg syscall (freebsd-arm64), const DLT_LAPD ideal-int
pkg syscall (freebsd-arm64), const DLT_LIN = 212
pkg syscall (freebsd-arm64), const DLT_LIN ideal-int
pkg syscall (freebsd-arm64), const DLT_LINUX_EVDEV = 216
pkg syscall (freebsd-arm64), const DLT_LINUX_EVDEV ideal-int

 * If you use masked losses with Keras the loss values may be different in TensorFlow `2.12` compared to previous versions.
 * In cases where the mask was previously ignored, you will now get an error if you pass a mask with an incompatible shape.

### Major Features and Improvements

`tf.keras`:

CX RORL $0x11, AX MOVL CX, DX RORL $0x13, CX SHRL $0x0a, DX MOVL 152(BP), BX XORL CX, AX MOVL BX, CX XORL DX, AX RORL $0x07, BX MOVL CX, DX SHRL $0x03, DX RORL $0x12, CX ADDL 184(BP), AX XORL CX, BX XORL DX, BX ADDL 148(BP), BX ADDL BX, AX MOVL AX, 212(BP) ADDL AX, R10 MOVL R15, AX ADDL $0x4ed8aa4a, R10 MOVL R15, CX RORL $0x06, AX MOVL R15, DX RORL $0x0b, CX XORL CX, AX MOVL R15, CX RORL $0x19, DX ANDL R8, CX XORL AX, DX MOVL R15, AX NOTL AX ADDL DX, R10 ANDL R9, AX XORL CX, AX ADDL R10, AX MOVL R11,...