Spaces:
Build error
Build error
File size: 4,226 Bytes
f35cc94 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 |
# SOURCE: https://github.com/cybertronai/pytorch-lamb/
import collections
import math
import torch
from torch.optim import Optimizer
class Lamb(Optimizer):
r"""Implements Lamb algorithm.
It has been proposed in `Reducing BERT Pre-Training Time from 3 Days to 76 Minutes`_.
Arguments:
params (iterable): iterable of parameters to optimize or dicts defining
parameter groups
lr (float, optional): learning rate (default: 1e-3)
betas (Tuple[float, float], optional): coefficients used for computing
running averages of gradient and its square (default: (0.9, 0.999))
eps (float, optional): term added to the denominator to improve
numerical stability (default: 1e-8)
weight_decay (float, optional): weight decay (L2 penalty) (default: 0)
adam (bool, optional): always use trust ratio = 1, which turns this into
Adam. Useful for comparison purposes.
.. _Reducing BERT Pre-Training Time from 3 Days to 76 Minutes:
https://arxiv.org/abs/1904.00962
"""
def __init__(self, params, lr=1e-3, betas=(0.9, 0.999), eps=1e-4,
weight_decay=0, adam=False):
if not 0.0 <= lr:
raise ValueError("Invalid learning rate: {}".format(lr))
if not 0.0 <= eps:
raise ValueError("Invalid epsilon value: {}".format(eps))
if not 0.0 <= betas[0] < 1.0:
raise ValueError("Invalid beta parameter at index 0: {}".format(betas[0]))
if not 0.0 <= betas[1] < 1.0:
raise ValueError("Invalid beta parameter at index 1: {}".format(betas[1]))
defaults = dict(lr=lr, betas=betas, eps=eps,
weight_decay=weight_decay)
self.adam = adam
super(Lamb, self).__init__(params, defaults)
def step(self, closure=None):
"""Performs a single optimization step.
Arguments:
closure (callable, optional): A closure that reevaluates the model
and returns the loss.
"""
loss = None
if closure is not None:
loss = closure()
for group in self.param_groups:
for p in group['params']:
if p.grad is None:
continue
grad = p.grad.data
if grad.is_sparse:
raise RuntimeError('Lamb does not support sparse gradients, consider SparseAdam instad.')
state = self.state[p]
# State initialization
if len(state) == 0:
state['step'] = 0
# Exponential moving average of gradient values
state['exp_avg'] = torch.zeros_like(p.data)
# Exponential moving average of squared gradient values
state['exp_avg_sq'] = torch.zeros_like(p.data)
exp_avg, exp_avg_sq = state['exp_avg'], state['exp_avg_sq']
beta1, beta2 = group['betas']
state['step'] += 1
if group['weight_decay'] != 0:
grad.add_(group['weight_decay'], p.data)
# Decay the first and second moment running average coefficient
exp_avg.mul_(beta1).add_(1 - beta1, grad)
exp_avg_sq.mul_(beta2).addcmul_(1 - beta2, grad, grad)
denom = exp_avg_sq.sqrt().add_(group['eps'])
bias_correction1 = 1 - beta1 ** state['step']
bias_correction2 = 1 - beta2 ** state['step']
# Apply bias to lr to avoid broadcast.
step_size = group['lr'] * math.sqrt(bias_correction2) / bias_correction1
adam_step = exp_avg / denom
# L2 norm uses sum, but here since we're dividing, use mean to avoid overflow.
r1 = p.data.pow(2).mean().sqrt()
r2 = adam_step.pow(2).mean().sqrt()
r = 1 if r1 == 0 or r2 == 0 else min(r1/r2, 10)
state['r1'] = r1
state['r2'] = r2
state['r'] = r
if self.adam:
r = 1
p.data.add_(-step_size * r, adam_step)
return loss
|