transformers

94e51ee3 - fix(trainer): pass optim_args to SGD, Adagrad, and RMSprop optimizers (#44203)

fix(trainer): pass optim_args to SGD, Adagrad, and RMSprop optimizers (#44203) SGD, Adagrad, and RMSprop ignored optim_args from TrainingArguments, unlike AdamW variants which properly parse and apply them. This adds optim_args support so users can customize optimizer hyperparameters via --optim_args for these optimizers. Fixes #44199 Co-authored-by: nightcityblade <nightcityblade@gmail.com> Co-authored-by: Marc Sun <57196510+SunMarc@users.noreply.github.com>