ExecuTorch in the context of TVM

Dear team,

Are you guys aware (perhaps participated) of the new route to bring models on-device for Pytorch users, ExecuTorch?

Could anyone shed light about the scopes of ExecuTorch & TVM? Iā€™m mostly interested in

  • similarity/overlap of both tools
  • pros/cons
  • differences
  • potential synergies

Feel free to share your sentiment as well :laughing: . BTW, I also created a post in PyTorch forum.

Thanks in advance!

I think there are great synergies. e.g. we could leverage the export from ExecuTorch and connect some of the relax operator offloadings to what ExecuTorch can offer. In the meantime, we can also bring compilation support together

1 Like