what I was looking for was support for onnx runtime instead of the python runtime.
jjmachan
05/04/2020, 6:26 AM
I'm not sure if there is any use case for your lib but the main advantage I was looking for was support using hardware accelerators with my packed models
c
Chaoyu
05/04/2020, 7:16 AM
got it, yea onnx runtime will definitely be one of the onnx backend we want to support
j
jjmachan
05/05/2020, 6:09 AM
yes, but for the time being the OnnxModelArtifact works. Once the onnx backend is set we can use it to serve the onnx models by default.