Refactor trainer init (#43807)
* refactor trainer init
* update init
* simplify liger
* udapte
* better
* comments
* do_train not reliable at all
* This should make more sense now
* Apply suggestion from @qgallouedec
Co-authored-by: Quentin Gallouédec <45557362+qgallouedec@users.noreply.github.com>
* Apply suggestion from @qgallouedec
Co-authored-by: Quentin Gallouédec <45557362+qgallouedec@users.noreply.github.com>
* Apply suggestion from @qgallouedec
Co-authored-by: Quentin Gallouédec <45557362+qgallouedec@users.noreply.github.com>
* Apply suggestion from @qgallouedec
Co-authored-by: Quentin Gallouédec <45557362+qgallouedec@users.noreply.github.com>
* move sagemaker mixed precision to validation
* Apply repo consistency fixes
---------
Co-authored-by: Quentin Gallouédec <45557362+qgallouedec@users.noreply.github.com>
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
