Mustafa Ali

icon indicating an email [email protected]

icon company @microsoft icon location Mountain View, CA icon location Former PhD Student at Purdue University, now a hardware engineer at MSFT Azure Hardware Arch

Results 7 comments of Mustafa Ali

Failing config files

> Maybe lacking some default configuration values. Add the 'default configuration values ...' part (line 185) in cache.cfg to other cfg files. It may work. I had the same problem...

GPU model as input to timeloop

Hi @nellie-wu , following on your comments about matching the sparse throughput described by the public doc of GPU. how can we get the peak throughput from the stats of...

Unable to run benchmark of mixtral

nope, hoping we get some reply from the contributors

use tensor cores

Sorry, somehow I missed your comment, yes, I looked at the cutlass implementation and it is similar to yours. I like yours because it teaches beginners like me to learn...

[FEA] FP8 Convolution

oh cool, I only tried with cutlass profiler and couldn't find it. Could you point out an example for that?

[FEA] FP8 Convolution

Thanks a lot!

[FA3] Performance confusion

I am also seeing significant performance difference from the paper especially with causal, would be great if a benchmarking script is given that can reproduce the results from the paper.