easy5 pts · first answer countstransformerstraining

Why training uses bf16

Most LLM training today uses bfloat16 (bf16) rather than float16 (fp16). What is the main advantage of bf16?

A.It stores more decimal digits of precision than fp32B.It uses one byte per value instead of twoC.It only works on CPUs, which are cheaper than GPUsD.It has the same exponent range as fp32, so values rarely overflow

Discussion 0

Join free to post in the discussion.

Why training uses bf16

Discussion 0

Related questions