TensorFlow.NET VS m2cgen

But to answer your question. I've run UNet in C#. I trained the data originally using python and used SciSharp to run the model using GPU for a solution more than 5 years ago.

TensorFlow .Net

Yes, exactly. Since you will probably use scikit / tensorflow / pytorch with Python, you can call your model directly in a Django controller. Using other frameworks you will probably have to create a separate microservice with a model exposed via Rest API or use bindings like TensorFlow.NET

Not even close , It is close to python's scikit learn , and it does not have it is own deep learning framework like tensorflow or pytorch , while there are some developers that are trying to implement tensorflow in dotnet , they are doing a good job at providing TensorFlow's low-level C++ API , However c# is not as good as python in manipulating data , there are tons of online materials for python's data science libraries like numpy , pandas , scikit learn , tensorflow , pytorch , quick bug fixes , even some support for unique cases .

But Ive actually been looking at TensorFlow.NET. Its uses .NET binding for TensorFlow machine learning library. Heres the GitHub page: https://github.com/SciSharp/TensorFlow.NET

If you're pointing out that Tensorflow.NET is exciting (https://github.com/SciSharp/TensorFlow.NET), I agree!

Looks like TensorFlow is callable directly from .NET. https://github.com/SciSharp/TensorFlow.NET

Building neural nets from scratch is going to be really hard, for a lot of reasons that you're not even aware of yet. You're better off using preexisting machine learning frameworks. For example here's a library for using tensorflow in c#: https://github.com/SciSharp/TensorFlow.NET

Why don't you try this: https://www.nuget.org/packages/TensorFlow.NET/ (and maybe other nugets).

Check out: https://github.com/BayesWitnesses/m2cgen

Notice that I’m using random values to populate the Sentiment field. You might compute the values for your fields based on other fields or actually use an ML model to perform the transformation. E.g. you could make use of m2cgen to transform trained models to pure python code and load them in **RedisGears **to be executed in a *GearsBuilder *instance. Another option is to pull out the big guns and go straight to RedisAI.

I wish that it was more common for model outputs to be converted the way bayeswitness does with mc2gen https://github.com/BayesWitnesses/m2cgen

And that’s it! All the magic in just two lines of code. I would like to thank the authors of the m2cgen library and encourage you to try it out.

I forgot to update the README. We just replaced RNN with xgboost that has a better f1 and is very quick, as the decision trees are translated to plain rust using m2cgen.

Probably, as many frameworks come with a Rust support (or there are wrappers). Some models, like decision tree, can also be automatically translated to plain Rust (in my company we use m2cgen to translate xgboost models to plain rust code).

These repositories on GitHub are good start I think: https://github.com/BayesWitnesses/m2cgen and https://github.com/vickylance/dart_nn