CHEN Dong
CHEN Dong
Thanks for your reply. Now I have a better understanding for the threat model of Occlum PKU. By the way, do you have any idea on SGX SFI? For the...
Thanks @Bonjourz, your discussion is really inspiring. > but we are still glad to learn some fun ideas if you are willing to share with us. Currently, I just want...
Sure, I have changed the name for this issue. But it seems that I can't remove the "bug" label, since I'm not a contributor of this repo. Could you help...
I guess no more discussion now. Thanks very much!
> [@ubergarm](https://github.com/ubergarm) I happen to know the people who are doing the ktransformer project. Its idea of utilizing Xeon (large memory) to host MoE and GPU for other layers is...
@mingfeima @chunyuan-w , hello, one question for this fp8 gemm on AMX CPU: will this use AMX_INT8 to compute fp8?
> > @mingfeima @chunyuan-w , hello, one question for this fp8 gemm on AMX CPU: will this use AMX_INT8 to compute fp8? > > The activation is BF16. We dequantized...
One follow up: my worker image has 40 layers, while the "[mcr.microsoft.com/azurelinux/base/pytorch:2.2.2-1-azl3.0.20240824-amd64](http://mcr.microsoft.com/azurelinux/base/pytorch:2.2.2-1-azl3.0.20240824-amd64)" has about 13 layers. I guess this issue is: 1) one layer is mapped to one virtio-pci device,...