kazum opened a new pull request #5634: URL: https://github.com/apache/incubator-tvm/pull/5634
This PR implements CoreML codegen using the BYOC feature and enables us to offload subgraphs to Apple’s Neural Engine on iOS devices. - The CoreML codegen is implemented with Python coremltools. - The `CoreMLRuntime` class is splitted into two classes, `CoreMLModel` and `CoreMLRuntime`, to support multiple models (subgraphs). - The CoreML runtime supports SaveToBinary and LoadFromBinary. Compiled CoreML directories (mlmodelc) are serialized into a binary format with NSFileWrapper. - A mobilenet test is added to illustrate the usage of the CoreML codegen. We can try it with [iOS TVM RPC](https://github.com/apache/incubator-tvm/tree/master/apps/ios_rpc) workflow. RFC: https://discuss.tvm.ai/t/rfc-coreml-codegen/6415 @tqchen @ZihengJiang @FrozenGene Could you help to review this? ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org
