The params of consine scheduler are as below in the documetation: [image] and I use params in TrainingArguments as below: [image] But it shows an error: [image] So here are my questions: Why this error occurs? If I need assign optimizer param in get_cosine_schedule_with_warmup?

How to use lr_scheduler_kwargs param in TrainingArguments?

John6666 June 25, 2025, 9:00am 10

warmup_steps

Oh. Let’s use it.