Hi @tqchen and @maheshambule,
I'm glad to share our code with you. The below figure is our current flowchart: ![|1221x249](https://msx.itri.org.tw/owa/service.svc/s/GetFileAttachment?id=AAMkADRhZjVmYjNiLWY1YTMtNGMyMS05NjZmLTlkZDU4MGI0YzRmOQBGAAAAAAAdWmvjQCH1R4whAGTAtKnfBwCd05YIuBehQ53IESALM0txAAAAAAEMAACd05YIuBehQ53IESALM0txAABGC833AAABEgAQAKqTMhV0bBpEqjmBmcHbpBI%3D&X-OWA-CANARY=DM3JyHwkwk2PboGdn04rOqrPW3_p5NcIubGVy8BmcvvJyUKZ5BFU7SA1b2GU3zfzlpzEQNnFFH4.) We partition Relay graph to two subgraphs, one is for accelerator which supports ONNX model, and the other is for other devices. The ONNX model is compiled by the 3rd compiler. And then build the graph following the original TVM codegen. Finally, we intend to create a new runtime integrate these codes. We also interest in your proposed Option C1. I hope we can cooperate with each other. --- [Visit Topic](https://discuss.tvm.ai/t/rfc-relay-to-onnx/6101/15) to respond. You are receiving this because you enabled mailing list mode. To unsubscribe from these emails, [click here](https://discuss.tvm.ai/email/unsubscribe/b1f00fa5afbe7e5fa67dd24601fc63af1deed9dea294c8fed1ced600b5cb8de7).