Initial commit of C# port - does not build

[MichaelSpencerJr]

🎉 Issues resolved:

• closes #

🧪 How to Test

1. Do this
2. Then this

[louisgv]

This is awesome! However I feel like it's better to put this in a separate repo :-?

[MichaelSpencerJr]

You might be right, but I hope to persuade people otherwise:

• I have no ability or interest in advertising my work, or in trying to wrest eyeballs away from this amazing project. Isolating this port in its own anonymous repo seems like it would doom the work to obscurity.
• I'm not really lead developer material at this point in my career, and will likely over-complicate things or focus on features people don't actually care about, without other developers to bounce ideas off of.

[JKamsker]

Comments for atleast the public properties would be awesome.
I can't figure out what the most of them are for by just looking at the code.

[MichaelSpencerJr]

Don't worry - I can't either. :-) I'm running the Python project in one Visual Studio debugger and am porting statements to C# in another as I encounter them. I haven't actually confirmed TensorFlow.net can successfully load the GPT-2 model into memory yet.

Once that is done, I'll rework this into something more product-shaped than proof-of-concept-shaped.

Multiple people on Discord were asking to see what I had so far, so that's why this is here (despite its current sorry state).

[JKamsker]

Ahh didn't know that xD
Well then keep it up :P

[JushBJJ]

insane

[iggyglass]

If need be, you could just run the code in Iron Python, which is a .NET python interpreter.

[dustinlacewell]

What is the overall goal of porting to C#?

[Deltrego]

Anyone interested in writing a C# port might find this useful: https://github.com/losttech/Gradient-Samples/tree/master/GPT-2 (Gradient is Tensorflow's .NET binding)