chore: update README and REPORT with performance insights and dataset changes

chore: update dependencies and configuration for improved training

feat(distiller): add option to skip post-training re-regularization

refactor(distiller): improve beam distillation and tokenlearn integration

feat(distiller): add checkpointing and refactor analyze.py

chore: remove unused scripts and update dependencies

feat: overhaul distiller package with unified CLI, enhanced evaluation, and modular structure

feat: created a cli to manage the complete generation process