3 ms·
And TensorFlow for C# is alive and kicking: https://losttech.software/gradient.html https://losttech.software/gradient.html I guess this is sort of an ad. Swi
by lostmsu 6y ago
And TensorFlow for C# is alive and kicking: https://losttech.software/gradient.html https://losttech.software/gradient.html
I guess this is sort of an ad.
Swift was a weird choice for statically typed TensorFlow, being only
popular on the platform, that does not have GPU/TPU support in TensorFlow,
which is, basically, a requirement for any serious work. The fact, that
they had to fork the compiler did not help either.
TensorFlow for C# is much like TensorFlow for Swift. There's a statically
typed binding to the core types with the rest of the TensorFlow API available
through an open source Python.NET project [1]. Unlike Swift version though,
that second part is also mostly statically typed, so you can get IDE
autocompletion hints. Also, .NET runtime has native support for dynamic
languages.
Like with Swift for TensorFlow, we ported all recent interesting
neural network architectures (and maybe even more): Transformers (GPT-2) [2],
CNNs (YOLOv4) [3], Q-learning (e.g. RL, actor-critic) [4], and even some
cool ones, that have lots of unexplored potential like Siren [5] (this one
uses sin/cos as activation functions).
Although we have not worked on automatic differentiation yet, unlike
Swift, it will not need compiler support. .NET (and Java) can inspect and
generate code at runtime, so autodiff can be implemented in a library.
We also have integration with Unity ML Agents for training robotic agents [4].
[1] https://github.com/pythonnet/pythonnet/ https://github.com/pythonnet/pythonnet/
[2] https://github.com/losttech/Gradient-Samples/tree/master/GPT-2 https://github.com/losttech/Gradient-Samples/tree/master/GPT...
[3] https://github.com/losttech/YOLOv4 https://github.com/losttech/YOLOv4
[4] https://github.com/losttech/Gradient-Samples/tree/master/RL-MLAgents https://github.com/losttech/Gradient-Samples/tree/master/RL-...
[5] https://github.com/losttech/Siren https://github.com/losttech/Siren https://vsitzmann.github.io/siren/ https://vsitzmann.github.io/siren/
- astrange 6y ago> Swift was a weird choice for statically typed TensorFlow, being only popular on the platform, that does not have GPU/TPU support in TensorFlow, which is, basically, a requirement for any serious work. The fact, that they had to fork the compiler did not help either. This is available now (not sure if it's in mainline). https://blog.tensorflow.org/2020/11/accelerating-tensorflow-performance-on-mac.html https://blog.tensorflow.org/2020/11/accelerating-tensorflow-...
- The_rationalist 6y agoDo you know if tensorflow for Java is as mature as tensorflow for C#? Btw of you didn't saw it, Facebook is implementation support for auto differentiation to the Kotlin compiler
- lostmsu 6y agoTensorFlow for Java is mostly for running pretrained models. Even though it might have training capabilities, they are extremely limited.