util.py

from __future__ import print_function

import math
from typing import Sized, Iterator

import numpy as np
import torch
from torch import Tensor
import torch.optim as optim


class TwoCropTransform:
    """Create two crops of the same image"""

    def __init__(self, transform):
        self.transform = transform

    def __call__(self, x):
        return [self.transform(x), self.transform(x)]


class AverageMeter(object):
    """Computes and stores the average and current value"""

    def __init__(self):
        self.reset()

    def reset(self):
        self.val = 0
        self.avg = 0
        self.sum = 0
        self.count = 0

    def update(self, val, n=1):
        self.val = val
        self.sum += val * n
        self.count += n
        self.avg = self.sum / self.count


def accuracy(output, target, topk=(1,)):
    """Computes the accuracy over the k top predictions for the specified values of k"""
    with torch.no_grad():
        maxk = max(topk)
        batch_size = target.size(0)

        _, pred = output.topk(maxk, 1, True, True)
        pred = pred.t()
        correct = pred.eq(target.view(1, -1).expand_as(pred))

        res = []
        for k in topk:
            correct_k = correct[:k].view(-1).float().sum(0, keepdim=True)
            res.append(correct_k.mul_(100.0 / batch_size))
        return res


def adjust_learning_rate(args, optimizer, epoch):
    lr = args.learning_rate
    if args.cosine:
        eta_min = lr * (args.lr_decay_rate ** 3)
        lr = eta_min + (lr - eta_min) * (
            1 + math.cos(math.pi * epoch / args.epochs)) / 2
    else:
        steps = np.sum(epoch > np.asarray(args.lr_decay_epochs))
        if steps > 0:
            lr = lr * (args.lr_decay_rate ** steps)

    for param_group in optimizer.param_groups:
        param_group['lr'] = lr


def warmup_learning_rate(args, epoch, batch_id, total_batches, optimizer):
    if args.warm and epoch <= args.warm_epochs:
        p = (batch_id + (epoch - 1) * total_batches) / \
            (args.warm_epochs * total_batches)
        lr = args.warmup_from + p * (args.warmup_to - args.warmup_from)

        for param_group in optimizer.param_groups:
            param_group['lr'] = lr


def set_optimizer(opt, model):
    optimizer = optim.SGD(model.parameters(),
                          lr=opt.learning_rate,
                          momentum=opt.momentum,
                          weight_decay=opt.weight_decay)
    return optimizer


def save_model(model, optimizer, opt, epoch, save_file):
    print('==> Saving...')
    state = {
        'opt': opt,
        'model': model.state_dict(),
        'optimizer': optimizer.state_dict(),
        'epoch': epoch,
    }
    torch.save(state, save_file)
    del state


from torch.utils.data.sampler import Sampler


class _InfiniteIterator(Iterator):

    def __init__(self, dataset_length, shuffle=True) -> None:
        self._dataset_length = dataset_length
        self._shuffle = shuffle
        self._regenerate_iter()

    def __iter__(self):
        return self.iterator

    def __next__(self):
        try:
            idx = next(self.iterator)
        except StopIteration:
            self._regenerate_iter()
            idx = next(self.iterator)
        return idx

    def _regenerate_iter(self):
        if self._shuffle:
            self.iterator = iter(torch.randperm(self._dataset_length).tolist())
        else:
            self.iterator = iter(
                torch.arange(start=0, end=self._dataset_length).tolist()
            )


class InfiniteSampler(Sampler):
    def __init__(self, data_source: Sized, shuffle=True) -> None:
        self._dataset = data_source
        self._shuffle = shuffle

    def __iter__(self):
        return _InfiniteIterator(self._dataset.__len__(), shuffle=self._shuffle)

    def __len__(self) -> int:
        return len(self._dataset)



if __name__ == '__main__':
    dataset = [1, 2, 3, 4, 5]
    sampler = InfiniteSampler(dataset, shuffle=False)
    for i in sampler:
        print(i)