https://en.wikipedia.org/wiki/Bfloat16_floating-point_format
Yes, however it’s a different format from standard fp16, it trades precision for greater dynamic range.