Source code for machina.optims.distributed_adamw

import math
import torch
from torch.optim.optimizer import Optimizer
import torch.distributed as dist

[docs]class DistributedAdamW(Optimizer): """Implements AdamW algorithm with distributed settings. It has been proposed in `Fixing Weight Decay Regularization in Adam`. Arguments: params (iterable): iterable of parameters to optimize or dicts defining parameter groups lr (float, optional): learning rate (default: 1e-3) betas (Tuple[float, float], optional): coefficients used for computing running averages of gradient and its square (default: (0.9, 0.999)) eps (float, optional): term added to the denominator to improve numerical stability (default: 1e-8) weight_decay (float, optional): weight decay (not L2 penalty) (default: 0) """ def __init__(self, params, local_rank, world_size, lr=1e-3, betas=(0.9, 0.999), eps=1e-8, weight_decay=0): defaults = dict(lr=lr, betas=betas, eps=eps, weight_decay=weight_decay) super(DistributedAdamW, self).__init__(params, defaults) self.local_rank = local_rank self.world_size = world_size
[docs] 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() grads = [] for group in self.param_groups: for p in group['params']: if p.grad is None: continue grads.append(p.grad) flat_grads = torch.nn.utils.parameters_to_vector(grads) dist.all_reduce_multigpu([flat_grads]) flat_grads /= self.world_size torch.nn.utils.vector_to_parameters(flat_grads, grads) for group in self.param_groups: for p in group['params']: if p.grad is None: continue grad = state = self.state[p] # State initialization if len(state) == 0: state['step'] = 0 # Exponential moving average of gradient values state['exp_avg'] = # Exponential moving average of squared gradient values state['exp_avg_sq'] = 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 exp_avg.mul_(beta1).add_(1 - beta1, grad) exp_avg_sq.mul_(beta2).addcmul_(1 - beta2, grad, grad) 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 if group['weight_decay'] != 0:['weight_decay'],, exp_avg, denom) params = [] for group in self.param_groups: for p in group['params']: if p.grad is None: continue params.append(p) params_vec = torch.nn.utils.parameters_to_vector(params) dist.broadcast_multigpu([params_vec], 0) torch.nn.utils.vector_to_parameters(params_vec, params) return loss