Igor Poletaev
Results
3
comments of
Igor Poletaev
A workaround for ones who desperately need them :) ``` // Run a command that returns a absl::Status. If the called code returns an // error status, return that status...
Is the reason of `bitsandbytes 8bit` being slower than even default fp16 - the flash attention kernels?
Same issue is still there even for 1.20.0 release version. Ubuntu 20.04 / x86 with CuDNN 9.5.1 and CUDA 12.2.