Like its Python-based cousin, TorchSharp is a rich and capable numerics library, especially well suited for machine learning using neural networks, that is, deep learning.
You can dive into TorchSharp by studying the examples provided, which provides an end-to-end view. On the other hand, if you would like a more methodical, step-by-step introduction to the many concepts and capabilities it offers, these tutorials aim to bring you one step at a time toward a comprehensive understanding of what it can do, and how.
The tutorials are organized to start with the very basics, creating and using tensors, which are generalized vectors and matrices, and the fundamental data type of all deep learning. Once we have treated tensors in depth, we will move on to using tensors to compute things. There are a ton of numerical operators available, and you can express just about anything using TorchSharp.
Once we have looked at numerics, it is time to move on to constructing models from numerics, and then to train them using optimizers and learning rate schedulers.
Most tutorials are presented in the form of an interactive notebook, which is intended to be executed under .NET Interactive. These notebooks have been developed and tested using Visual Studio Code with the .NET Interactive extension installed.
For more information on installing the .NET Interactive extension, see:
Installing .NET Interactive Notebooks
or
For your tutorials, please choose one of: C# or F#.
More tutorials are coming.