Version 1.0.3

• Added support for YAML configuration.
• Increased flexibility and control over the pruning process.
• Enhanced the README.md with a description of AutoML WIP and example usage.
• Included a TODO list with GitHub-compatible checkboxes for future enhancements.
• Improved documentation for better clarity and usability.

v1.0.2

Update README and main.py to Improve Pruning Functionality

Added new arguments for pruning methods and normalization options:

New Arguments:

• --prune_method: Specifies the method to use for pruning. Currently, the supported options are "mk_prune" (alias: "mk") and "mk_prune_adjusted" (alias: "mka"). The default is set to mk_prune.
• --use_normalized_weights: Enables the use of normalized weights to compute the final weights. The default setting is False.
• --use_layer_norm_tweaks: Applies layer normalization adjustments to account for the effects of pruned neurons. The default is False.
• --layer_norm_scale: Sets the layer normalization scale, applicable only if --use_layer_norm_tweaks is enabled. The default value is 4.0.

Version 1.0.1

Enhance model pruning functionality and update documentation

• Added new arguments for pruning methods and model updates in README.
• Introduced new functions for calculating importance scores and normalizing weights.
• Updated .gitignore to exclude additional temporary files and directories.
• Improved output function to return logits alongside generated text.