adamp.py

"""
AdamP
Copyright (c) 2020-present NAVER Corp.
MIT license

+ eps inside sqrt
+ init sq with ones instead of zeros
"""

import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.optim.optimizer import Optimizer, required
import math


class AdamP(Optimizer):
    def __init__(
        self, params, lr=1e-3, betas=(0.9, 0.999), eps=1e-10, weight_decay=0, delta=0.1, wd_ratio=0.1, nesterov=False
    ):
        defaults = dict(
            lr=lr, betas=betas, eps=eps, weight_decay=weight_decay, delta=delta, wd_ratio=wd_ratio, nesterov=nesterov
        )
        super(AdamP, self).__init__(params, defaults)

    def _channel_view(self, x):
        return x.view(x.size(0), -1)

    def _layer_view(self, x):
        return x.view(1, -1)

    def _cosine_similarity(self, x, y, eps, view_func):
        x = view_func(x)
        y = view_func(y)

        return F.cosine_similarity(x, y, dim=1, eps=eps).abs_()

    def _projection(self, p, grad, perturb, delta, wd_ratio, eps):
        wd = 1
        expand_size = [-1] + [1] * (len(p.shape) - 1)
        for view_func in [self._channel_view, self._layer_view]:

            cosine_sim = self._cosine_similarity(grad, p.data, eps, view_func)

            if cosine_sim.max() < delta / math.sqrt(view_func(p.data).size(1)):
                p_n = p.data / view_func(p.data).norm(dim=1).view(expand_size).add_(eps)
                perturb -= p_n * view_func(p_n * perturb).sum(dim=1).view(expand_size)
                wd = wd_ratio

                return perturb, wd

        return perturb, wd

    def step(self, closure=None):
        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
                beta1, beta2 = group["betas"]
                nesterov = group["nesterov"]

                state = self.state[p]

                # State initialization
                if len(state) == 0:
                    state["step"] = 0
                    state["exp_avg"] = torch.zeros_like(p.data)
                    # state['exp_avg_sq'] = torch.zeros_like(p.data) # original
                    state["exp_avg_sq"] = torch.ones_like(p.data)  # much better init

                # Adam
                exp_avg, exp_avg_sq = state["exp_avg"], state["exp_avg_sq"]

                state["step"] += 1
                bias_correction1 = 1 - beta1 ** state["step"]
                bias_correction2 = 1 - beta2 ** state["step"]

                exp_avg.mul_(beta1).add_(grad, alpha=1 - beta1)
                exp_avg_sq.mul_(beta2).addcmul_(grad, grad, value=1 - beta2)

                # denom = (exp_avg_sq.sqrt() / math.sqrt(bias_correction2)).add_(group['eps']) # original
                denom = exp_avg_sq.add_(group["eps"]).sqrt() / math.sqrt(bias_correction2)  # eps inside sqrt
                step_size = group["lr"] / bias_correction1

                if nesterov:
                    perturb = (beta1 * exp_avg + (1 - beta1) * grad) / denom
                else:
                    perturb = exp_avg / denom

                # Projection
                wd_ratio = 1
                if len(p.shape) > 1:
                    perturb, wd_ratio = self._projection(
                        p, grad, perturb, group["delta"], group["wd_ratio"], group["eps"]
                    )

                # Weight decay
                if group["weight_decay"] > 0:
                    p.data.mul_(1 - group["lr"] * group["weight_decay"] * wd_ratio)

                # Step
                p.data.add_(perturb, alpha=-step_size)

        return loss