Inference in go (trained in python and deploy in golang)

[manjunathsc24]

Hello @chewxy ,

I have a model(autoencoder) trained in python , now i would like to deploy (do inference) in golang, i tried with onnx-go but that is not good in performance , could you please suggest me how to do inference in golang gorgonia that will be great help from your side.

Thank You
Manjunath

[auxten]

I think it's pretty difficult(impossible) to import a model trained by other frameworks to gorgonia.
For now, gorgonia build the computing graph arch in golang, the Forward and Backprop behavior may not identical.
To achieve that, there are at least 3 things to do manually:

1. Parse the model description for PyTorch or TF
2. Build the missing Node type needed by autoencoder
3. Create the whole computing graph and recover the weights, maybe something like https://github.com/gorgonia/parser/