News for hackers. Stuff that matters.
Re:Bfloat16 support coming to Apple's Metal and PyTorch [video] by sillysaurusx on Monday July 03, 2023 @03:19PM (#36577608)
At that point it’d be better to do everything in fp32. The hardware can’t do bf16 in the way you’re saying; the conversions would consume all your time.
Re:Bfloat16 support coming to Apple's Metal and PyTorch [video] by BooneJS on Monday July 03, 2023 @04:00PM (#36578026)
Compute in F32, but then round and pack a pair of BF16 into 4 bytes.
Re:Bfloat16 support coming to Apple's Metal and PyTorch [video] by brrrrrm on Tuesday July 04, 2023 @01:40AM (#36582619)
The conversions are just a mask and shift? Super cheap