cppflow VS onnxruntime

It’s been awhile since I’ve looked at it, so not sure how hard it would be to get to work. I only commented since you mentioned that you would support other operating systems. For others interested in cross platform support there is also cppflow.

what about start with keras and convert model to c++ ? https://github.com/pplonski/keras2cpp https://github.com/serizba/cppflow

My solution so far: I am using a compiled Tensorflow C DLL in combination with cppflow (https://github.com/serizba/cppflow). However, I get problems when I take models which use operations from the tensorflow_text python module since I don’t know how to get their C++ API.

I use cppflow to run compiled tensorflow models natively in C++. It works like a charm :)

I toyed around with this idea a while back but I never got around to finishing the implementation. If all you need is inference with no training and you are relatively familiar with c++ you could look into creating a module for Godot that interfaces with the Tensorflow C API. Something like cppflow would provide an even easier API to work with. Looking into that project could also explain how they interface with the Tensorflow C API if you'd rather cut out the middle man. A module like this would let you train your model in Python and then load it and perform inference in Godot natively.

If your re using keras with TensorFlow you can save it as a saved model format and then you can easily use cppflow to perform inference with it.

The whole game is written in C++ using SFML for the graphics, entt as Entity-Component-System and tensorflow for the Neural Network. Tensorflow itself is written in C, so I use cppflow to integrate it into my C++ framework.

Found this which seems more recent (uses TF 2, updated 4 days ago): https://github.com/serizba/cppflow

ONNX Runtime: ONNX Runtime: cross-platform, high performance ML inferencing and training accelerator

Embedding generation uses the ONNX runtime under the hood. This is a cross-platform inferencing library that supports multiple execution providers from CPU to specialized GPUs.

tfjs is dead, looking at the commit history. The standard now is to convert PyTorch to onnx, then use onnxruntime (https://github.com/microsoft/onnxruntime/tree/main/js/web) to run the model on the browsdr.

- slang[2] has the potential, but the meta programming part is not as strong as C++, existing libraries cannot be used.

The above conclusion is drawn from my work https://github.com/microsoft/onnxruntime/tree/dev/opencl, purely nightmare to work with thoes drivers and jit compilers. Hopefully Vcc can take compute shader more seriously.

[1]: https://www.circle-lang.org/

I did. It depends what you want, for an overview of how ONNX Runtime works then Microsoft have a bunch of things on https://onnxruntime.ai, but the Java content is a bit lacking on there as I've not had time to write much. Eventually I'll probably write something similar to the C# SD tutorial they have on there but for the Java API.

For writing ONNX models from Java we added an ONNX export system to Tribuo in 2022 which can be used by anything on the JVM to export ONNX models in an easier way than writing a protobuf directly. Tribuo doesn't have full coverage of the ONNX spec, but we're happy to accept PRs to expand it, otherwise it'll fill out as we need it.

What about models besides GPT? Most of the popular vector encoding models aren't using this architecture.

If you really didn't want PyTorch/Transformers, you could consider exporting your models to ONNX (https://github.com/microsoft/onnxruntime).