Develop

[singjc]

This pull request introduces several enhancements and updates to the ReDeeM project, focusing on adding CUDA support, improving data handling, and extending model capabilities. Key changes include the addition of a CUDA-enabled Dockerfile, updates to data handling with metadata support, and the introduction of new models in the redeem-properties crate. Below is a categorized summary of the most important changes:

Infrastructure and Build Updates:

• Added a new Dockerfile to enable building the project with NVIDIA CUDA support, setting up the environment for GPU-accelerated computations.
• Updated the Cargo.toml workspace to include the new redeem-cli crate.

Data Handling Enhancements:

• Introduced a new PsmMetadata struct in data_handling.rs to store metadata such as spectrum IDs, file IDs, and feature names. This metadata is now integrated into the Experiment struct. [1] [2] [3]
• Refactored example scripts (gbdt_semi_supervised_learning.rs, svm_semi_supervised_learning.rs, xgb_semi_supervised_learning.rs) to use a new load_test_psm_csv function for loading data with metadata, replacing older TSV-based methods. [1] [2] [3]

Model and Feature Updates:

• Expanded the list of models in the redeem-properties crate, adding CNN-Transformer-based models for RT and CCS prediction.
• Updated model names in the README.md for consistency with the new architecture.

Code Refactoring:

• Replaced the license field in redeem-classifiers/Cargo.toml with a blank entry, possibly for future updates.
• Enhanced logging and error handling in the example scripts to improve debugging and maintainability. [1] [2] [3]

These changes collectively improve the project's usability, scalability, and support for advanced features like GPU acceleration and metadata-driven workflows.