tape/optimization.py

# coding=utf-8
# Copyright 2018 The Google AI Language Team Authors and The HuggingFace Inc. team.
# Modifications by Roshan Rao
#
# 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

import torch
from torch.optim import Optimizer  # type: ignore
from torch.optim.lr_scheduler import LambdaLR

logger = logging.getLogger(__name__)


class ConstantLRSchedule(LambdaLR):
    """ 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):
    """ 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):
    """ 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):
    """ 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)))


class WarmupCosineWithHardRestartsSchedule(LambdaLR):
    """ Linear warmup and then cosine cycles with hard restarts.
        Linearly increases learning rate from 0 to 1 over `warmup_steps` training steps.
        If `cycles` (default=1.) is different from default, learning rate follows `cycles` times
        a cosine decaying learning rate (with hard restarts).
    """
    def __init__(self, optimizer, warmup_steps, t_total, cycles=1., last_epoch=-1):
        self.warmup_steps = warmup_steps
        self.t_total = t_total
        self.cycles = cycles
        super(WarmupCosineWithHardRestartsSchedule, 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))
        # progress after warmup
        progress = float(step - self.warmup_steps) / float(
            max(1, self.t_total - self.warmup_steps))
        if progress >= 1.0:
            return 0.0
        return max(0.0, 0.5 * (1. + math.cos(
            math.pi * ((float(self.cycles) * progress) % 1.0))))


class AdamW(Optimizer):
    """ Implements Adam algorithm with weight decay fix.

    Parameters:
        lr (float): learning rate. Default 1e-3.
        betas (tuple of 2 floats): Adams beta parameters (b1, b2). Default: (0.9, 0.999)
        eps (float): Adams epsilon. Default: 1e-6
        weight_decay (float): Weight decay. Default: 0.0
        correct_bias (bool): can be set to False to avoid correcting bias in Adam
            (e.g. like in Bert TF repository). Default True.
    """
    def __init__(self,
                 params,
                 lr=1e-3,
                 betas=(0.9, 0.999),
                 eps=1e-6,
                 weight_decay=0.0,
                 correct_bias=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(f"Invalid beta parameter: {betas[0]} - should be in [0.0, 1.0)")
        if not 0.0 <= betas[1] < 1.0:
            raise ValueError(f"Invalid beta parameter: {betas[1]} - should be in [0.0, 1.0)")
        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(AdamW, 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('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_(1.0 - beta1, grad)
                exp_avg_sq.mul_(beta2).addcmul_(1.0 - beta2, grad, grad)
                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_(-step_size, exp_avg, denom)

                # 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_(-group['lr'] * group['weight_decay'], p.data)

        return loss