fairseq/fairseq/optim/adamax.py

# Copyright (c) Facebook, Inc. and its affiliates.
#
# This source code is licensed under the MIT license found in the
# LICENSE file in the root directory of this source tree.

import torch
import torch.optim

from . import LegacyFairseqOptimizer, register_optimizer


@register_optimizer("adamax")
class FairseqAdamax(LegacyFairseqOptimizer):
    def __init__(self, args, params):
        super().__init__(args)
        self._optimizer = Adamax(params, **self.optimizer_config)

    @staticmethod
    def add_args(parser):
        """Add optimizer-specific arguments to the parser."""
        # fmt: off
        parser.add_argument('--adamax-betas', default='(0.9, 0.999)', metavar='B',
                            help='betas for Adam optimizer')
        parser.add_argument('--adamax-eps', type=float, default=1e-8, metavar='D',
                            help='epsilon for Adam optimizer')
        parser.add_argument('--weight-decay', '--wd', default=0.0, type=float, metavar='WD',
                            help='weight decay')
        parser.add_argument('--no-bias-correction', default=False, action='store_true',
                            help='disable bias correction')
        # fmt: on

    @property
    def optimizer_config(self):
        """
        Return a kwarg dictionary that will be used to override optimizer
        args stored in checkpoints. This allows us to load a checkpoint and
        resume training using a different set of optimizer args, e.g., with a
        different learning rate.
        """
        return {
            "lr": self.args.lr[0],
            "betas": eval(self.args.adamax_betas),
            "eps": self.args.adamax_eps,
            "weight_decay": self.args.weight_decay,
            "bias_correction": not self.args.no_bias_correction,
        }


class Adamax(torch.optim.Optimizer):
    """Implements Adamax algorithm (a variant of Adam based on infinity norm).

    It has been proposed in `Adam: A Method for Stochastic Optimization`__.

    Compared to the version in PyTorch, this version implements a fix for weight decay.

    Args:
        params (iterable): iterable of parameters to optimize or dicts defining
            parameter groups
        lr (float, optional): learning rate (default: 2e-3)
        betas (Tuple[float, float], optional): coefficients used for computing
            running averages of gradient and its square
        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)
        bias_correction (bool, optional): enable bias correction (default: True)

    __ https://arxiv.org/abs/1412.6980
    """

    def __init__(
        self,
        params,
        lr=2e-3,
        betas=(0.9, 0.999),
        eps=1e-8,
        weight_decay=0,
        bias_correction=True,
    ):
        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]))
        if not 0.0 <= weight_decay:
            raise ValueError("Invalid weight_decay value: {}".format(weight_decay))

        defaults = dict(
            lr=lr,
            betas=betas,
            eps=eps,
            weight_decay=weight_decay,
            bias_correction=bias_correction,
        )
        super(Adamax, self).__init__(params, defaults)

    @property
    def supports_memory_efficient_fp16(self):
        return True

    @property
    def supports_flat_params(self):
        return True

    def step(self, closure=None):
        """Performs a single optimization step.

        Args:
            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.float()
                if grad.is_sparse:
                    raise RuntimeError("Adamax does not support sparse gradients")

                p_data_fp32 = p.data
                if p.data.dtype in {torch.float16, torch.bfloat16}:
                    p_data_fp32 = p_data_fp32.float()

                state = self.state[p]

                # State initialization
                if len(state) == 0:
                    state["step"] = 0
                    state["exp_avg"] = torch.zeros_like(p_data_fp32)
                    state["exp_inf"] = torch.zeros_like(p_data_fp32)
                else:
                    state["exp_avg"] = state["exp_avg"].to(p_data_fp32)
                    state["exp_inf"] = state["exp_inf"].to(p_data_fp32)

                exp_avg, exp_inf = state["exp_avg"], state["exp_inf"]
                beta1, beta2 = group["betas"]
                eps = group["eps"]

                state["step"] += 1

                # Update biased first moment estimate.
                exp_avg.mul_(beta1).add_(grad, alpha=1 - beta1)

                # Update the exponentially weighted infinity norm.
                torch.max(
                    exp_inf.mul_(beta2),
                    grad.abs_(),
                    out=exp_inf,
                )

                step_size = group["lr"]
                if group["bias_correction"]:
                    bias_correction = 1 - beta1 ** state["step"]
                    step_size /= bias_correction

                if group["weight_decay"] != 0:
                    p_data_fp32.add_(
                        p_data_fp32, alpha=-group["weight_decay"] * group["lr"]
                    )

                p_data_fp32.addcdiv_(exp_avg, exp_inf.add(eps), value=-step_size)

                if p.data.dtype in {torch.float16, torch.bfloat16}:
                    p.data.copy_(p_data_fp32)

        return loss