You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I don't know of any proposal to include fp8 types in numpy itself, but you might take a look at the ml_dtypes package, which defines several flavors of fp8 dtypes for use with NumPy.
We may end up incorporating bfloat16 into numpy itself at some point, but float8 seems like a bridge too far, at least in the near future (and no one has proposed actually doing this anyway). So I'll close this since the question is answered and a solution available. Thanks @xiguadong and @jakevdp
Proposed new feature or change:
Hi :
Do you have any plan to support FP8 which proposed in https://arxiv.org/pdf/2209.05433 ?
The text was updated successfully, but these errors were encountered: