0.0.5 Expanded functionality and testing

mplement CPU Override Feature for Explicit Device Control

• Added new cpu:-1 device specification to force CPU usage
• Enhanced device redirection logic to recognize and respect CPU override
• Simplified device handling with more robust device selection mechanism
• Updated README and CHANGELOG with new CPU override feature details
• Created dedicated test modules for CPU and MPS operations
• Improved logging and error handling for device selection
• Bumped version to 0.0.5 to reflect significant feature addition

Interim release while improving test suites and refacoring all code

Interrim release while refactoring and improving tests.

Minor updates to README prior to refactoring

Concentrating on extensively testing this. Using Transformers since it has many issues working on Apple Silicon.

• The original TorchDevice has become too large and refactoring before continuing testing.
• Wanted to make this interim release which works with most things available.
• Works with most things, though still finding parameter mismatches and other issues.
• Methodically working through the Transformers Regression tests using TorchDevice to redirect CUDA and CPU to MPS for Apple Silicon
• This is not just for CUDA -> MPS migration, it will spot and log a message when it performs a GPU redirect.
• Should, though I don't have CUDA hardware to work with be compatible with CUDA devices
• Should also allow MPS -> CUDA transition as well.

Project Goal is to ease migration from one hardware architecture to another since it should "just work" and alert developers where code needs to be changed or fixed.

Pre relese before refactoring and complete Transformers testing

Concentrating on extensively testing this. Using Transformers since it has many issues working on Apple Silicon.

• The original TorchDevice has become too large and refactoring before continuing testing.
• Wanted to make this interim release which works with most things available.
• Works with most things, though still finding parameter mismatches and other issues.
• Methodically working through the Transformers Regression tests using TorchDevice to redirect CUDA and CPU to MPS for Apple Silicon
• This is not just for CUDA -> MPS migration, it will spot and log a message when it performs a GPU redirect.
• Should, though I don't have CUDA hardware to work with be compatible with CUDA devices
• Should also allow MPS -> CUDA transition as well.

Project Goal is to ease migration from one hardware architecture to another since it should "just work" and alert developers where code needs to be changed or fixed.

First release.

Key features include:
• Automatic Device Redirection: Redirects torch.device based on hardware (CUDA, MPS, CPU).
• Mocked CUDA Functions: Run CUDA-based code on MPS.
• Unified Memory Handling: Smooth transition between memory management systems.
• Performance Boost: Includes a recipe to recompile NumPy with Accelerate on macOS, yielding 8x performance gains.

Also in the README there is a recipe for improving your NumPy performance.

Feedback always welcome.