[Tensorize] how to use tensorize for composition op(eg. conv+relu)

Hi all,
If my hardware can only do the composition op like conv+relu, I can only write this in two compute for tvm.sum must be the top level of a compute.
But tensorize can not deal with composition op yet.
So what can we do in this situation.

2 Likes

Is your question similar to the following question?
In tensorize tutorials, if the Hardware support gemv+relu,how to use the feature of hardware instead of doing gemv and relu separately ?

Yes, and the tutorial only show the case that tensorize a non-composition op.

1 Like

I am also looking at a similar use case. Is there any solution or update?

I am also interested in an update.

@tqchen @yzhliu @zhiics

2 Likes

+1 for my team. Current work around is using pragmas for hints about fusion at codegen stage

Hi, can you describe your schemes a bit more thoroughly? I encounted this problem as well.

@xwrock For now I think the proper way to work around this is to go the BYOC route until this part of TVM becomes more flexible for custom accelerators.

@xwrock another way is to bypass tensorize like the ppl of VTA did. It requires some manipulation of the low level AST, but I guess the end effect are the same.

@aca88 Looks like i’m running into the same issue :frowning: Can you point to the code of VTA where they do this tensorization bypass?

Thanks!