Open 7wik3vedi16 opened 3 weeks ago
Hey! This isn't necessary for performance, you could convert these to full precision floats! :)
@7wik3vedi16 you can try changing thedtype = float16
from the experiment.py,
but that would be really slower as compared with bfloat16
ive tried it in google colab and same error happened, maybe you can try with a100/h100 gpu
Error: "Feature "bf16" requires .target sm_80 or higher
Is there any other way other than upgrading my GPU ?