Hi,
is there a special reason why you store this model as float16 and not bfloat16?Since the base model is in bfloat16...
Many thanksPhilip
Hi Philip, no particular reason for FP16
· Sign up or log in to comment