espnet.scheduler.scheduler.CyclicCosineScheduler
Less than 1 minute
espnet.scheduler.scheduler.CyclicCosineScheduler
class espnet.scheduler.scheduler.CyclicCosineScheduler(key: str, args: Namespace)
Bases: SchedulerInterface
Cyclic cosine annealing.
- Parameters:
- cosine_warmup (int) – number of warmup iterations.
- cosine_total (int) – number of total annealing iterations.
Notes
Proposed in https://openreview.net/pdf?id=BJYwwY9ll (and https://arxiv.org/pdf/1608.03983.pdf). Used in the GPT2 config of Megatron-LM https://github.com/NVIDIA/Megatron-LM
Initialize class.
alias = 'cosine'
scale(n_iter)
Scale of lr.