Adapt Dendrite or Dagre.NET to work with TorchSharp

[GeorgeS2019]

It seems you are dealing with TorchScript?

[fel88]

There is some code to load TorchScript files, but it needs to be refined. Also, I planned to implement support for '.caffemodel' files using OpenCV.DNN inference engine

[GeorgeS2019]

Start looking into PyTorch and start considering TorchSharp, the most advanced .NET Deep AI right now with respect to any other Deep AI approaches for .NET communities.

[fel88]

I like PyTorch. All my .onnx models were exported from PyTorch models. Not sure how I can use TorchSharp in this project. Besides, TorchScript is not such a popular format as Onnx

[GeorgeS2019]

PyTorch is complex so as how the Deep AI community dealing with TorchScript, fast evolving, hard to catch up. One thing sure, your view is not up to date :-) Keep learning, Keep Open Minded. There is a serious adoption of Torchscript just a few weeks back

[fel88]

I think TorchSharp can be easily integrated into Dendrite as one of the inference engines via plugin . But it's not easy to find pre-trained TorchScript models , I would be appreciated if you could share some models

[GeorgeS2019]

e.g. https://huggingface.co/docs/transformers/torchscript

[GeorgeS2019]

Also, we need someone to contribute to TorchSharp's TorchText. Please look into what others have contributed to TorchVision and TorchAudio. I believe you are talented enough to contribute to TorchText, since you have started with GPT2 last year :-)

[GeorgeS2019]

@fel88 One more suggestion. TorchSharp project follows one of the high Software Engineering industry practices: extensive Unit Tests that serve both to communicate the coverage of LibTorch API and documentation!