【transformer】【pytorch】TransFG代码【Scheduler.py】

LambdaLR类：torch.optim.lr_scheduler
作用是，可以自定义学习率学习曲线，官方代码如下：

lass LambdaLR(_LRScheduler):
    """Sets the learning rate of each parameter group to the initial lr
    times a given function. When last_epoch=-1, sets initial lr as lr.

    Args:
        optimizer (Optimizer): Wrapped optimizer.
        lr_lambda (function or list): A function which computes a multiplicative
            factor given an integer parameter epoch, or a list of such
            functions, one for each group in optimizer.param_groups.
        last_epoch (int): The index of last epoch. Default: -1.

    Example:
        >>> # Assuming optimizer has two groups.
        >>> lambda1 = lambda epoch: epoch // 30
        >>> lambda2 = lambda epoch: 0.95 ** epoch
        >>> scheduler = LambdaLR(optimizer, lr_lambda=[lambda1, lambda2])
        >>> for epoch in range(100):
        >>>     train(...)
        >>>     validate(...)
        >>>     scheduler.step()
    """

    def __init__(self, optimizer, lr_lambda, last_epoch=-1):
    	...

其中，optimizer是优化器实例，如SGD的实例，lr_lambda是学习率变化的函数或者一组函数，函数的参数是当前的epoch，last_epoch是前一个epoch，比如epoch=2训练完后，中断了，继续训练的话last_epoch=2，内部会自动加1.

import logging
import math

from torch.optim.lr_scheduler import LambdaLR

logger = logging.getLogger(__name__)

class ConstantLRSchedule(LambdaLR):#常数
    """ Constant learning rate schedule.
    """
    def __init__(self, optimizer, last_epoch=-1):
        super(ConstantLRSchedule, self).__init__(optimizer, lambda _: 1.0, last_epoch=last_epoch)


class WarmupConstantSchedule(LambdaLR):#梯形，并且是左边斜边
    """ Linear warmup and then constant.
        Linearly increases learning rate schedule from 0 to 1 over `warmup_steps` training steps.
        Keeps learning rate schedule equal to 1. after warmup_steps.
    """
    def __init__(self, optimizer, warmup_steps, last_epoch=-1):
        self.warmup_steps = warmup_steps
        super(WarmupConstantSchedule, self).__init__(optimizer, self.lr_lambda, last_epoch=last_epoch)

    def lr_lambda(self, step):
        if step < self.warmup_steps:
            return float(step) / float(max(1.0, self.warmup_steps))
        return 1.


class WarmupLinearSchedule(LambdaLR):#等腰三角形
    """ Linear warmup and then linear decay. 
        Linearly increases learning rate from 0 to 1 over `warmup_steps` training steps.
        Linearly decreases learning rate from 1. to 0. over remaining `t_total - warmup_steps` steps.
    """
    def __init__(self, optimizer, warmup_steps, t_total, last_epoch=-1):
        self.warmup_steps = warmup_steps
        self.t_total = t_total
        super(WarmupLinearSchedule, self).__init__(optimizer, self.lr_lambda, last_epoch=last_epoch)

    def lr_lambda(self, step):
        if step < self.warmup_steps:
            return float(step) / float(max(1, self.warmup_steps))
        return max(0.0, float(self.t_total - step) / float(max(1.0, self.t_total - self.warmup_steps)))


class WarmupCosineSchedule(LambdaLR):#左边斜线，右边cos，并且cos的程度可以通过cycles设定
    """ Linear warmup and then cosine decay.
        Linearly increases learning rate from 0 to 1 over `warmup_steps` training steps.
        Decreases learning rate from 1. to 0. over remaining `t_total - warmup_steps` steps following a cosine curve.
        If `cycles` (default=0.5) is different from default, learning rate follows cosine function after warmup.
    """
    def __init__(self, optimizer, warmup_steps, t_total, cycles=.5, last_epoch=-1):
        self.warmup_steps = warmup_steps
        self.t_total = t_total
        self.cycles = cycles
        super(WarmupCosineSchedule, self).__init__(optimizer, self.lr_lambda, last_epoch=last_epoch)

    def lr_lambda(self, step):
        if step < self.warmup_steps:
            return float(step) / float(max(1.0, self.warmup_steps))
        # progress after warmup
        progress = float(step - self.warmup_steps) / float(max(1, self.t_total - self.warmup_steps))
        return max(0.0, 0.5 * (1. + math.cos(math.pi * float(self.cycles) * 2.0 * progress)))
    #cycle<0.5：不到pi