# coding=utf-8
# Copyright 2018 The Google AI Language Team Authors and The HuggingFace Inc. team.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""PyTorch optimization for BERT model."""
import logging
import math
from typing import Callable, Iterable, Tuple
import torch
from torch.optim import Optimizer
from torch.optim.lr_scheduler import LambdaLR
logger = logging.getLogger(__name__)
[docs]def get_constant_schedule(optimizer: Optimizer, last_epoch: int = -1):
"""
Create a schedule with a constant learning rate, using the learning rate set in optimizer.
Args:
optimizer (:class:`~torch.optim.Optimizer`):
The optimizer for which to schedule the learning rate.
last_epoch (:obj:`int`, `optional`, defaults to -1):
The index of the last epoch when resuming training.
Return:
:obj:`torch.optim.lr_scheduler.LambdaLR` with the appropriate schedule.
"""
return LambdaLR(optimizer, lambda _: 1, last_epoch=last_epoch)
[docs]def get_constant_schedule_with_warmup(optimizer: Optimizer, num_warmup_steps: int, last_epoch: int = -1):
"""
Create a schedule with a constant learning rate preceded by a warmup period during which the learning rate
increases linearly between 0 and the initial lr set in the optimizer.
Args:
optimizer (:class:`~torch.optim.Optimizer`):
The optimizer for which to schedule the learning rate.
num_warmup_steps (:obj:`int`):
The number of steps for the warmup phase.
last_epoch (:obj:`int`, `optional`, defaults to -1):
The index of the last epoch when resuming training.
Return:
:obj:`torch.optim.lr_scheduler.LambdaLR` with the appropriate schedule.
"""
def lr_lambda(current_step: int):
if current_step < num_warmup_steps:
return float(current_step) / float(max(1.0, num_warmup_steps))
return 1.0
return LambdaLR(optimizer, lr_lambda, last_epoch=last_epoch)
[docs]def get_linear_schedule_with_warmup(optimizer, num_warmup_steps, num_training_steps, last_epoch=-1):
"""
Create a schedule with a learning rate that decreases linearly from the initial lr set in the optimizer to 0,
after a warmup period during which it increases linearly from 0 to the initial lr set in the optimizer.
Args:
optimizer (:class:`~torch.optim.Optimizer`):
The optimizer for which to schedule the learning rate.
num_warmup_steps (:obj:`int`):
The number of steps for the warmup phase.
num_training_steps (:obj:`int`):
The totale number of training steps.
last_epoch (:obj:`int`, `optional`, defaults to -1):
The index of the last epoch when resuming training.
Return:
:obj:`torch.optim.lr_scheduler.LambdaLR` with the appropriate schedule.
"""
def lr_lambda(current_step: int):
if current_step < num_warmup_steps:
return float(current_step) / float(max(1, num_warmup_steps))
return max(
0.0, float(num_training_steps - current_step) / float(max(1, num_training_steps - num_warmup_steps))
)
return LambdaLR(optimizer, lr_lambda, last_epoch)
[docs]def get_cosine_schedule_with_warmup(
optimizer: Optimizer, num_warmup_steps: int, num_training_steps: int, num_cycles: float = 0.5, last_epoch: int = -1
):
"""
Create a schedule with a learning rate that decreases following the values of the cosine function between the
initial lr set in the optimizer to 0, after a warmup period during which it increases linearly between 0 and the
initial lr set in the optimizer.
Args:
optimizer (:class:`~torch.optim.Optimizer`):
The optimizer for which to schedule the learning rate.
num_warmup_steps (:obj:`int`):
The number of steps for the warmup phase.
num_training_steps (:obj:`int`):
The total number of training steps.
num_cycles (:obj:`float`, `optional`, defaults to 0.5):
The number of waves in the cosine schedule (the defaults is to just decrease from the max value to 0
following a half-cosine).
last_epoch (:obj:`int`, `optional`, defaults to -1):
The index of the last epoch when resuming training.
Return:
:obj:`torch.optim.lr_scheduler.LambdaLR` with the appropriate schedule.
"""
def lr_lambda(current_step):
if current_step < num_warmup_steps:
return float(current_step) / float(max(1, num_warmup_steps))
progress = float(current_step - num_warmup_steps) / float(max(1, num_training_steps - num_warmup_steps))
return max(0.0, 0.5 * (1.0 + math.cos(math.pi * float(num_cycles) * 2.0 * progress)))
return LambdaLR(optimizer, lr_lambda, last_epoch)
[docs]def get_cosine_with_hard_restarts_schedule_with_warmup(
optimizer: Optimizer, num_warmup_steps: int, num_training_steps: int, num_cycles: int = 1, last_epoch: int = -1
):
"""
Create a schedule with a learning rate that decreases following the values of the cosine function between the
initial lr set in the optimizer to 0, with several hard restarts, after a warmup period during which it increases
linearly between 0 and the initial lr set in the optimizer.
Args:
optimizer (:class:`~torch.optim.Optimizer`):
The optimizer for which to schedule the learning rate.
num_warmup_steps (:obj:`int`):
The number of steps for the warmup phase.
num_training_steps (:obj:`int`):
The total number of training steps.
num_cycles (:obj:`int`, `optional`, defaults to 1):
The number of hard restarts to use.
last_epoch (:obj:`int`, `optional`, defaults to -1):
The index of the last epoch when resuming training.
Return:
:obj:`torch.optim.lr_scheduler.LambdaLR` with the appropriate schedule.
"""
def lr_lambda(current_step):
if current_step < num_warmup_steps:
return float(current_step) / float(max(1, num_warmup_steps))
progress = float(current_step - num_warmup_steps) / float(max(1, num_training_steps - num_warmup_steps))
if progress >= 1.0:
return 0.0
return max(0.0, 0.5 * (1.0 + math.cos(math.pi * ((float(num_cycles) * progress) % 1.0))))
return LambdaLR(optimizer, lr_lambda, last_epoch)
[docs]class AdamW(Optimizer):
"""
Implements Adam algorithm with weight decay fix as introduced in
`Decoupled Weight Decay Regularization <https://arxiv.org/abs/1711.05101>`__.
Parameters:
params (:obj:`Iterable[torch.nn.parameter.Parameter]`):
Iterable of parameters to optimize or dictionaries defining parameter groups.
lr (:obj:`float`, `optional`, defaults to 1e-3):
The learning rate to use.
betas (:obj:`Tuple[float,float]`, `optional`, defaults to (0.9, 0.999)):
Adam's betas parameters (b1, b2).
eps (:obj:`float`, `optional`, defaults to 1e-6):
Adam's epsilon for numerical stability.
weight_decay (:obj:`float`, `optional`, defaults to 0):
Decoupled weight decay to apply.
correct_bias (:obj:`bool`, `optional`, defaults to `True`):
Whether ot not to correct bias in Adam (for instance, in Bert TF repository they use :obj:`False`).
"""
def __init__(
self,
params: Iterable[torch.nn.parameter.Parameter],
lr: float = 1e-3,
betas: Tuple[float, float] = (0.9, 0.999),
eps: float = 1e-6,
weight_decay: float = 0.0,
correct_bias: bool = True,
):
if lr < 0.0:
raise ValueError("Invalid learning rate: {} - should be >= 0.0".format(lr))
if not 0.0 <= betas[0] < 1.0:
raise ValueError("Invalid beta parameter: {} - should be in [0.0, 1.0[".format(betas[0]))
if not 0.0 <= betas[1] < 1.0:
raise ValueError("Invalid beta parameter: {} - should be in [0.0, 1.0[".format(betas[1]))
if not 0.0 <= eps:
raise ValueError("Invalid epsilon value: {} - should be >= 0.0".format(eps))
defaults = dict(lr=lr, betas=betas, eps=eps, weight_decay=weight_decay, correct_bias=correct_bias)
super().__init__(params, defaults)
[docs] def step(self, closure: Callable = None):
"""
Performs a single optimization step.
Arguments:
closure (:obj:`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("Adam does not support sparse gradients, please consider SparseAdam instead")
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
# In-place operations to update the averages at the same time
exp_avg.mul_(beta1).add_(grad, alpha=1.0 - beta1)
exp_avg_sq.mul_(beta2).addcmul_(grad, grad, value=1.0 - beta2)
denom = exp_avg_sq.sqrt().add_(group["eps"])
step_size = group["lr"]
if group["correct_bias"]: # No bias correction for Bert
bias_correction1 = 1.0 - beta1 ** state["step"]
bias_correction2 = 1.0 - beta2 ** state["step"]
step_size = step_size * math.sqrt(bias_correction2) / bias_correction1
p.data.addcdiv_(exp_avg, denom, value=-step_size)
# Just adding the square of the weights to the loss function is *not*
# the correct way of using L2 regularization/weight decay with Adam,
# since that will interact with the m and v parameters in strange ways.
#
# Instead we want to decay the weights in a manner that doesn't interact
# with the m/v parameters. This is equivalent to adding the square
# of the weights to the loss with plain (non-momentum) SGD.
# Add weight decay at the end (fixed version)
if group["weight_decay"] > 0.0:
p.data.add_(p.data, alpha=-group["lr"] * group["weight_decay"])
return loss