onnx VS hummingbird

Once your model has been trained and validated using Anomalib, the next step is to prepare it for real-time implementation. This is where ONNX (Open Neural Network Exchange) or OpenVINO (Open Visual Inference and Neural network Optimization) comes into play.

ONNX is an open format from the Linux Foundation to represent machine learning models. It is becoming extensively adopted by the Machine Learning community and is compatible with most of the machine learning frameworks like PyTorch, TensorFlow, etc. Converting a model between any of those formats and ONNX is really simple and can be done in most cases with a single command.

ONNX[0], model-as-protosbufs, continuing to gain adoption will hopefully solve this issue.

[0] https://github.com/onnx/onnx

Running inference for large models on CPU is not a new problem and fortunately there has been great development in many different optimization frameworks for speeding up matrix and tensor computations on CPU. We explored multiple optimization frameworks and methods to improve latency, namely TorchScript, BetterTransformer and ONNX.

ONNX is a format for representing machine learning models in a portable way. Additionally, ONNX models can be easily optimized and thus become smaller and faster.

I would say onnx.ai [0] provides more information about ONNX for those who aren’t working with ML/DL.

[0] https://onnx.ai

It's not clear why you thing this sub is appropriate for some third party system with a Python interface. Why don't you try their discussion group: https://github.com/onnx/onnx/discussions

This kind of indirection through standardisation is pretty common to make compatibility between different kinds of software components easier. Some other good examples are the LSP project from Microsoft and ONNX to represent machine learning models. The first provides a standard so that IDEs don't have to re-invent the weel for every programming language. The latter decouples training frameworks from inference frameworks. Going back to WSGI, you can find a pretty extensive rationale for the WSGI standard here if interested.

If are interested in just predictions you can try Hummingbird. It is part of the PyTorch ecosystem. We get already trained scikit-learn models and translate them into PyTorch models. From them you can run your model on any hardware support by PyTorch, export it into TVM, ONNX, etc. Performance on hardware acceleration is quite good (orders of magnitude better than scikit-learn is some cases)

I think Rapids AI's cuML tried to go into this direction (essentially scikit-learn on the GPU): https://docs.rapids.ai/api/cuml/stable/api.html#logistic-reg.... For some reason it never took really off though.

Btw., going on a tangent, you might like Hummingbird (https://github.com/microsoft/hummingbird). It allows you trained scikit-learn tree-based models to PyTorch. I watched the SciPy talk last year, and it's a super smart & elegant idea.

ONNX opens an avenue for direct inference using a number of languages and platforms. For example, a model could be run directly on Android to limit data sent to a third party service. ONNX is an exciting development with a lot of promise. Microsoft has also released Hummingbird which enables exporting traditional models (sklearn, decision trees, logistical regression..) to ONNX.

And Python.

For inference, you can convert your models to other formats that support GPU acceleration. See Hummingbird https://github.com/microsoft/hummingbird

The surprising thing is that Hummingbird can be faster than the GPU implementation of LightGBM (and XGBoost) if you use tensor compilers such as TVM. [The paper](https://www.usenix.org/conference/osdi20/presentation/nakandala) describes our findings. We have also open sourced the [benchmark code](https://github.com/microsoft/hummingbird/tree/main/benchmarks) so you try yourself!

I took their sample code from Github and tweaked it to spit out times for each model's prediction, as well as increase the number of rows to 5 million. I used Google's Colab and selected GPU for my hardware accelerator. This gives an option to run code on GPU, not that all computations will happen on the GPU.