ML.net CLI support using GPU

[xsa-dev]

1. The current version of the ML.net CLI tool does not support GPU acceleration, which limits its performance and efficiency when working with large and complex datasets. This is a significant issue, as the use of GPUs in machine learning has become increasingly prevalent in recent years.

2. I propose adding GPU support to the ML.net CLI tool. This would allow users to harness the power of GPUs when running their machine learning models, greatly increasing performance and efficiency. Specifically, I suggest adding an option to the existing command line interface to enable GPU support, allowing users to easily switch between CPU and GPU mode as needed.

3. While there are alternative tools available for working with GPU-accelerated machine learning, such as the Model Builder tool, tool often lack the flexibility and customization options provided by the ML.net CLI tool. Additionally, many users are more comfortable working with command-line interfaces than with graphical user interfaces.

4. The addition of GPU support to the ML.net CLI tool would align it with the current industry standards and would bring parity with the ML.net GUI tool, which already supports GPU acceleration. This would allow users to work with their preferred interface while still being able to take advantage of the performance benefits of GPUs.

[xsa-dev]

[luisquintanilla]

cc @LittleLittleCloud

[JakeRadMSFT]

@xsa-dev I think we can make this happen once we migrate our image-classification to TorchSharp.

Just to set expectations ... this is probably 6-9 months out.

[ynnob]

@JakeRadMSFT is there any update on this?