Open WenheLI opened 4 years ago
I think we could convert the job's model when generating the output: https://github.com/alibaba/pipcook/blob/master/packages/daemon/src/service/pipeline.ts#L391. However, it requires that daemon knows what neural network framework is used for the pipeline, any idea @FeelyChau?
After investigation on tvm, I found that we could use tvm to tune a model & export it to a wasm executable in both nodejs and browsers. Later, I will upload the
mvp
to github for reference.The next step is to merge this feature into our system. My idea is to implement a separate CLI tool for producing a bundle with an arbitrary model.