I am not super familiar with the Unity direction, but keeping BYOC sounds like a good idea. I don't know if this is how its supposed to be used, but I am using it as "catch all" way to extend TVM. I'm currently adding some custom opencl kernels for depthwise conv2d: the way that I am planning to do it is to just perform a graph substitution with all relax nodes that are dw conv2d layers (using BYOC). Is there an easier way to do this?
I can't express the changes in tensor expression, so this looked like my only option. --- [Visit Topic](https://discuss.tvm.apache.org/t/phasing-out-legacy-components/17703/11) to respond. You are receiving this because you enabled mailing list mode. To unsubscribe from these emails, [click here](https://discuss.tvm.apache.org/email/unsubscribe/48ceb647263a6b82e0a8cab0e5d166981618aea51a4ef32d5a294e334266b867).
