File size: 6,346 Bytes
f32c034 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 | # MIT License
#
# Copyright (c) 2019 cybertronai
#
# Permission is hereby granted, free of charge, to any person obtaining a copy
# of this software and associated documentation files (the "Software"), to deal
# in the Software without restriction, including without limitation the rights
# to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
# copies of the Software, and to permit persons to whom the Software is
# furnished to do so, subject to the following conditions:
#
# The above copyright notice and this permission notice shall be included in all
# copies or substantial portions of the Software.
#
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
# AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
# SOFTWARE.
"""Lamb optimizer."""
# From: https://github.com/cybertronai/pytorch-lamb
import collections
import math
import torch
from tensorboardX import SummaryWriter
from torch.optim import Optimizer
def log_lamb_rs(optimizer: Optimizer, event_writer: SummaryWriter, token_count: int):
"""Log a histogram of trust ratio scalars in across layers."""
results = collections.defaultdict(list)
for group in optimizer.param_groups:
for p in group['params']:
state = optimizer.state[p]
for i in ('weight_norm', 'adam_norm', 'trust_ratio'):
if i in state:
results[i].append(state[i])
for k, v in results.items():
event_writer.add_histogram(f'lamb/{k}', torch.tensor(v), token_count)
class Lamb(Optimizer):
r"""Implements Lamb algorithm.
It has been proposed in `Large Batch Optimization for Deep Learning: Training BERT in 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.
.. _Large Batch Optimization for Deep Learning: Training BERT in 76 minutes:
https://arxiv.org/abs/1904.00962
"""
def __init__(self, params, lr=1e-3, betas=(0.9, 0.999), eps=1e-6,
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
# Decay the first and second moment running average coefficient
# m_t
exp_avg.mul_(beta1).add_(grad, alpha=1 - beta1)
# v_t
exp_avg_sq.mul_(beta2).addcmul_(grad, grad, value=1 - beta2)
# Paper v3 does not use debiasing.
# 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
weight_norm = p.data.pow(2).sum().sqrt().clamp(0, 10)
adam_step = exp_avg / exp_avg_sq.sqrt().add(group['eps'])
if group['weight_decay'] != 0:
adam_step.add_(p.data, alpha=group['weight_decay'])
adam_norm = adam_step.pow(2).sum().sqrt()
if weight_norm == 0 or adam_norm == 0:
trust_ratio = 1
else:
trust_ratio = weight_norm / adam_norm
state['weight_norm'] = weight_norm
state['adam_norm'] = adam_norm
state['trust_ratio'] = trust_ratio
if self.adam:
trust_ratio = 1
p.data.add_(adam_step, alpha=-step_size * trust_ratio)
return loss
|