reference/code/Fast-dLLM/third_party/lmflow/optim/adamax.py

#!/usr/bin/env python
# -*- coding: utf-8 -*-

import torch
from torch.optim.optimizer import Optimizer

class Adamax(Optimizer):
    def __init__(self, params, lr=2e-3, betas=(0.9, 0.999), eps=1e-8, weight_decay=0):
        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)
        super(Adamax, self).__init__(params, defaults)

    def __setstate__(self, state):
        super(Adamax, 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('Adamax does not support sparse gradients')
                state = self.state[p]

                # State initialization
                if len(state) == 0:
                    state['step'] = 0
                    state['exp_avg'] = torch.zeros_like(p.data)
                    state['exp_inf'] = torch.zeros_like(p.data)

                exp_avg, exp_inf = state['exp_avg'], state['exp_inf']
                beta1, beta2 = group['betas']

                state['step'] += 1

                if group['weight_decay'] != 0:
                    grad = grad.add(group['weight_decay'], p.data)

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

                # Update the exponentially weighted infinity norm
                norm_buf = torch.cat([
                    exp_inf.mul_(beta2).unsqueeze(0),
                    grad.abs().unsqueeze_(0)
                ], 0)
                torch.max(norm_buf, 0, keepdim=False, out=(exp_inf, exp_inf.new().long()))

                bias_correction = 1 - beta1 ** state['step']
                clr = group['lr'] / bias_correction

                p.data.addcdiv_(-clr, exp_avg, exp_inf + group['eps'])

        return loss