Do you know of any examples of compiling models from Hugging Face transformers to Relax, getting the computational graph, and launching kernels for each node? I have looked around the Relax code a bit and didn't see anything that would enable this level of fine-grained control.
--- [Visit Topic](https://discuss.tvm.apache.org/t/kernel-launching-and-dynamic-kernel-selection-for-llms-with-relay/17635/3) 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/5cd1a2ab1fd60fccfd8425e208145665e0cca07559b8e2a46b54b9418e360271).
