| |
| |
|
|
| import torch |
| import math |
|
|
| class NAdam(torch.optim.Optimizer): |
| def __init__(self, params, lr=2e-3, betas=(0.9, 0.999), eps=1e-8, weight_decay=0, momentum_decay=4e-3): |
| 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)) |
| if not 0.0 <= momentum_decay: |
| raise ValueError("Invalid momentum_decay value: {}".format(momentum_decay)) |
| defaults = dict(lr=lr, betas=betas, eps=eps, weight_decay=weight_decay, momentum_decay=momentum_decay) |
| super(NAdam, self).__init__(params, defaults) |
|
|
| def __setstate__(self, state): |
| super(NAdam, self).__setstate__(state) |
|
|
| 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 |
| if grad.is_sparse: |
| raise RuntimeError('NAdam does not support sparse gradients') |
| state = self.state[p] |
|
|
| |
| if len(state) == 0: |
| state['step'] = 0 |
| state['m_prev'] = torch.zeros_like(p.data) |
| state['v'] = torch.zeros_like(p.data) |
|
|
| m_prev, v = state['m_prev'], state['v'] |
| beta1, beta2 = group['betas'] |
|
|
| state['step'] += 1 |
| bias_correction1 = 1 - beta1 ** state['step'] |
| bias_correction2 = 1 - beta2 ** state['step'] |
|
|
| if group['weight_decay'] != 0: |
| grad = grad.add(group['weight_decay'], p.data) |
|
|
| m = beta1 * m_prev + (1 - beta1) * grad |
| v.mul_(beta2).addcmul_(1 - beta2, grad, grad) |
|
|
| m_hat = m / bias_correction1 |
| v_hat = v / bias_correction2 |
|
|
| denom = v_hat.sqrt().add_(group['eps']) |
|
|
| momentum_decay = group['momentum_decay'] |
| m_prev.mul_(beta1).add_(1 - beta1, grad) |
| m_prev_hat = m_prev / bias_correction1 |
|
|
| step_size = group['lr'] * math.sqrt(bias_correction2) / bias_correction1 |
|
|
| p.data.addcdiv_(-step_size, m_hat + momentum_decay * m_prev_hat, denom) |
|
|
| return loss |
