Changes in version 1.0.0                        

  - New Feature: Added full support for the tidymodels ecosystem,
    including integration with parsnip, recipes, workflows, and tune.
  - New Feature: Added GPU acceleration support for fitting Bayesian
    Neural Networks via OpenCL and cmdstanr.
  - New Feature: Introduced model evaluation metrics using Leave-One-Out
    (LOO) cross-validation (loo.bnns()) and Watanabe-Akaike Information
    Criterion (waic.bnns()).
  - New Feature: Added save_bnns() and load_bnns() functions for easy
    saving and loading of trained models.
  - New Feature: Added a plot.bnns() method for visualizing posterior
    distributions and model convergence.
  - New Feature: Added option to provide character input for act_fn and
    out_act_fn arguments.
  - Changed the package to stable lifecycle and prepared for CRAN
    submission.
  - Removed cmdstanr from package dependencies. It is now accessed
    dynamically to comply with CRAN policies.
  - Removed the manual rhub GitHub Actions badge from the README.
  - Added CODE_OF_CONDUCT.md.

                     Changes in version 0.1.2.9000                      

  - Added a NEWS.md file to track changes to the package.