-
Notifications
You must be signed in to change notification settings - Fork 62
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Does flashinfer support float datatype? #191
Comments
I got the same question. I am instantiate the SinglePrefillWithKVCacheDispatched function, but found that it has |
The decode attention operators support fp32, we just need to add fp32 to this macro: flashinfer/python/csrc/pytorch_extension_utils.h Lines 36 to 51 in 5a38066
For prefill/append attention, it's a little bit tricky, because many instructions such as |
Got it, my use case is prefill/append kernel and it looks tricky indeed. Thanks for your kind reply. I think the support of fp32 output sounds great and helpful! |
The examples are all tensors of half() type. I wonder if flashinfer supports fp32 dtype?
The text was updated successfully, but these errors were encountered: