v2.2.0

Fix upload modules, scripts, and params correctly to fit in this case.

v2.1.0

Able to select num_workers of DataLoader using workers_ratio and using gpu devices.
Able to select num_workers of DataLoader using all cpu workers.
Fix tuner suggest param types at scheduler.
Refactor a scheduler.
Change some default parameter values.
Match optuna suggest ranges with default param values.

v2.0.0

Scheduler works correctly.
Hyper-parameters(optimizer, scheduler) scientific notation.
Refactor some parameter names.
Python version and packages update.
Fix typo.

v1.7.0

Change default values of parameters to fit the task.
Get total number of train batches in whole training process correctly.

v1.6.0

Upload model based on upload_tag and model_detail.
Change default scripts.

v1.5.2

Versions of libraries update.

v1.5.1

Parameter names change clearly.
eta_min depends on LR.

v1.5.0

Adapt weight decay at deepspeed Adam optimizers.
Weight decay hparams tuning.
Model safetensors keep trained dtype.

v1.4.1

Requirements updated.

v1.4.0

Gather and convert as a model.pt after all deepspeed strategy training for uploading hub.
Postprocessing(uploading hub) modules and scripts.