Source code for cooper.optim.torch_optimizers.extragradient

# Copyright (C) 2025 The Cooper Developers.
# Licensed under the MIT License.

"""Optimizers based on extra-gradient."""

import math
from collections.abc import Callable, Iterable
from typing import NoReturn, Optional

import torch

# -----------------------------------------------------------------------------
# Implementation of ExtraOptimizers contains minor edits on source code from:
# https://github.com/GauthierGidel/Variational-Inequality-GAN/blob/master/optim/extragradient.py
# * We add a `maximize` flag to the `ExtraSGD` and `ExtraAdam` classes to allow for
# maximization steps.
# * We slightly modify the docstrings to comply with our style guide.
#
# TODO(juan43ramirez): The implementations below "manually" apply SGD and Adam updates.
# Alternatively, we could carry out updates using functional implementations of SGD and
# Adam from `torch.optim`. This way, we can easily stay up-to-date with the community
# approved implementations of these optimizers.


# -----------------------------------------------------------------------------

#  MIT License

# Copyright (c) Facebook, Inc. and its affiliates.

# 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.

# written by Hugo Berard (berard.hugo@gmail.com) while at Facebook.


[docs] class ExtragradientOptimizer(torch.optim.Optimizer): r"""Base class for :class:`torch.optim.Optimizer`\s with an ``extrapolation`` step. Args: params: an iterable of :class:`torch.Tensor`\s or :class:`dict`\s. Specifies what Tensors should be optimized. defaults: a dict containing default values of optimization options (used when a parameter group doesn't specify them). """ def __init__(self, params: Iterable, defaults: dict) -> None: super().__init__(params, defaults) self.params_copy: list[torch.nn.Parameter] = [] def update(self, p: torch.Tensor, group: dict) -> NoReturn: raise NotImplementedError
[docs] def extrapolation(self) -> None: """Performs the extrapolation step and saves a copy of the current parameters for the update step. """ # Check if a copy of the parameters was already made. is_empty = len(self.params_copy) == 0 for group in self.param_groups: for p in group["params"]: u = self.update(p, group) if is_empty: # Save the current parameters for the update step. Several # extrapolation step can be made before each update but only # the parameters before the first extrapolation step are # saved. self.params_copy.append(p.data.clone()) if u is None: continue # Update the current parameters p.data.add_(u)
[docs] def step(self, closure: Optional[Callable] = None) -> Optional[torch.Tensor]: """Performs a single optimization step. Args: closure: A closure that reevaluates the model and returns the loss. """ if len(self.params_copy) == 0: raise RuntimeError("Need to call extrapolation before calling step.") loss = None if closure is not None: with torch.enable_grad(): loss = closure() i = -1 for group in self.param_groups: for p in group["params"]: i += 1 u = self.update(p, group) if u is None: continue # Update the parameters saved during the extrapolation step p.data = self.params_copy[i].add_(u) # Free the old parameters self.params_copy = [] return loss
[docs] class ExtraSGD(ExtragradientOptimizer): r"""Extrapolation-compatible implementation of SGD with momentum. .. note:: The implementation of SGD with Momentum/Nesterov subtly differs from :cite:p:`sutskever2013initialization` and implementations in some other frameworks. Considering the specific case of Momentum, the update can be written as: .. math:: \vv_{t+1} = \rho \cdot \vv_t + \nabla_{\vtheta} L(\vtheta_t) \\ \vtheta_{t+1} = \vtheta_t - \eta \cdot \vv_{t+1}, where :math:`\vtheta`, :math:`\vv`, :math:`\nabla_{\vtheta} L` and :math:`\rho` denote the parameters, velocity, gradient and momentum respectively. This is in contrast to :cite:p:`sutskever2013initialization` and other frameworks which employ an update of the form: .. math:: \vv_{t+1} &= \rho \cdot \vv_t + \eta \cdot \nabla_{\vtheta} L(\vtheta_t) \\ \vtheta_{t+1} &= \vtheta_t - \vv_{t+1}. The Nesterov version is modified analogously. Args: params: Iterable of parameters to optimize or dicts defining parameter groups. lr: Learning rate. momentum: Momentum factor. weight_decay: Weight decay (L2 penalty). dampening: Dampening for momentum. nesterov: If ``True``, enables Nesterov momentum. Raises: ValueError: If the learning rate, momentum, or weight decay are negative. ValueError: If Nesterov momentum is enabled while momentum is set to zero or dampening is not zero. """ def __init__( self, params: Iterable, lr: float = 1e-3, momentum: float = 0, dampening: float = 0, weight_decay: float = 0, nesterov: bool = False, maximize: bool = False, ) -> None: if lr < 0.0: raise ValueError(f"Invalid learning rate: {lr}") if momentum < 0.0: raise ValueError(f"Invalid momentum value: {momentum}") if weight_decay < 0.0: raise ValueError(f"Invalid weight_decay value: {weight_decay}") defaults = { "lr": lr, "momentum": momentum, "dampening": dampening, "weight_decay": weight_decay, "nesterov": nesterov, "maximize": maximize, } if nesterov and (momentum == 0 or dampening != 0): raise ValueError("Nesterov momentum requires a momentum and zero dampening") super().__init__(params, defaults) def __setstate__(self, state: dict) -> None: super(torch.optim.SGD, self).__setstate__(state) for group in self.param_groups: group.setdefault("nesterov", False) def update(self, p: torch.Tensor, group: dict) -> Optional[torch.Tensor]: weight_decay = group["weight_decay"] momentum = group["momentum"] dampening = group["dampening"] nesterov = group["nesterov"] maximize = group["maximize"] if p.grad is None: return None d_p = p.grad.data if maximize: d_p = -d_p if weight_decay != 0: d_p.add_(weight_decay, p.data) if momentum != 0: param_state = self.state[p] if "momentum_buffer" not in param_state: buf = param_state["momentum_buffer"] = torch.zeros_like(p.data) buf.mul_(momentum).add_(d_p) else: buf = param_state["momentum_buffer"] buf.mul_(momentum).add_(d_p, alpha=1 - dampening) d_p = d_p.add(momentum, buf) if nesterov else buf return -group["lr"] * d_p
[docs] class ExtraAdam(ExtragradientOptimizer): """Implements the Adam algorithm with an extrapolation step. Args: params: Iterable of parameters to optimize or dicts defining parameter groups. lr : Learning rate. betas: Coefficients used for computing running averages of gradient and its square. eps : Term added to the denominator to improve numerical stability. weight_decay: Weight decay (L2 penalty). amsgrad: Flag to use the AMSGrad variant of this algorithm from :cite:p:`reddi2018amsgrad`. Raises: ValueError: If the learning rate or epsilon value is negative. ValueError: If the beta parameters are not in the range [0, 1). """ def __init__( self, params: Iterable, lr: float = 1e-3, betas: tuple[float, float] = (0.9, 0.999), eps: float = 1e-8, weight_decay: float = 0, amsgrad: bool = False, maximize: bool = False, ) -> None: if lr < 0.0: raise ValueError(f"Invalid learning rate: {lr}") if eps < 0.0: raise ValueError(f"Invalid epsilon value: {eps}") if not 0.0 <= betas[0] < 1.0: raise ValueError(f"Invalid beta parameter at index 0: {betas[0]}") if not 0.0 <= betas[1] < 1.0: raise ValueError(f"Invalid beta parameter at index 1: {betas[1]}") defaults = { "lr": lr, "betas": betas, "eps": eps, "weight_decay": weight_decay, "amsgrad": amsgrad, "maximize": maximize, } super().__init__(params, defaults) def __setstate__(self, state: dict) -> None: super().__setstate__(state) for group in self.param_groups: group.setdefault("amsgrad", False) def update(self, p: torch.Tensor, group: dict) -> Optional[torch.Tensor]: if p.grad is None: return None grad = p.grad.data if group["maximize"]: grad = -grad if grad.is_sparse: raise RuntimeError("Adam does not support sparse gradients, please consider SparseAdam instead") amsgrad = group["amsgrad"] 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) if amsgrad: # Maintains max of all exp. moving avg. of sq. grad. values state["max_exp_avg_sq"] = torch.zeros_like(p.data) exp_avg, exp_avg_sq = state["exp_avg"], state["exp_avg_sq"] if amsgrad: max_exp_avg_sq = state["max_exp_avg_sq"] beta1, beta2 = group["betas"] state["step"] += 1 if group["weight_decay"] != 0: grad = grad.add(group["weight_decay"], p.data) # Decay the first and second moment running average coefficient exp_avg.mul_(beta1).add_(grad, alpha=1 - beta1) exp_avg_sq.mul_(beta2).addcmul_(grad, grad, value=1 - beta2) if amsgrad: # Maintains the maximum of all 2nd moment running avg. till now torch.max(max_exp_avg_sq, exp_avg_sq, out=max_exp_avg_sq) # Use the max. for normalizing running avg. of gradient denom = max_exp_avg_sq.sqrt().add_(group["eps"]) else: denom = exp_avg_sq.sqrt().add_(group["eps"]) bias_correction1 = 1 - beta1 ** state["step"] bias_correction2 = 1 - beta2 ** state["step"] step_size = group["lr"] * math.sqrt(bias_correction2) / bias_correction1 return -step_size * exp_avg / denom