How to do the deep learning calculation in the keystone?

[ww ruan]

Hi， 

 

Thanks for your reply in https://github.com/keystone-enclave/keystone/issues/181#issuecomment-630456038.

 

Actually I tried to use libtorch(torch in C++) in the keystone, which compile(use gcc) successfully, but when ran in the qemu, I got the error: the runtime could not handle exception 2. I guessed the error may be caused by insufficient memory space? Is there any advice for the inference?

Thank you

Ruan

[David William Kohlbrenner]

Unfortunately our runtime errors aren't very informative right now.

The most likely scenario is that the runtime you built either didn't support needed operations, or you ran out of memory in the enclave.

You can compile Eyrie with strace-style debug printing (see build.sh in Eyrie for options) that may help some. (https://github.com/keystone-enclave/keystone-runtime/blob/master/build.sh)

It may also be worth running things under gdb once you have an idea of what is actually failing.

If you look at keystone-bench (https://github.com/keystone-enclave/keystone-bench/tree/master/bench-runner) the benchmark runner is a good example of the available options when creating an enclave (including setting total memory available.)

Good luck! And update us if you figure out what the problem is.

-David

--
You received this message because you are subscribed to the Google Groups "Keystone Enclave Forum" group.
To unsubscribe from this group and stop receiving emails from it, send an email to keystone-enclave-...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/keystone-enclave-forum/cfd37fb9-8967-4b51-9af3-c0d7ed271750%40googlegroups.com.