Function accuracy

train_utils.py:263–292

Computes the accuracy over the k top predictions for the specified values of k Args output: logits or probs (num of batch, num of classes) target: (num of batch, 1) or (num of batch, ) topk: list of returned k refer: https://github.com/pytorch/examples/

(output, target, topk=(1,))

def accuracy(output, target, topk=(1,)):
    """
    Computes the accuracy over the k top predictions for the specified values of k
    
    Args
        output: logits or probs (num of batch, num of classes)
        target: (num of batch, 1) or (num of batch, )
        topk: list of returned k
    
    refer: https://github.com/pytorch/examples/blob/master/imagenet/main.py
    """

    with torch.no_grad():
        maxk = max(topk)  # get k in top-k
        batch_size = target.size(0)  # get batch size of target

        # torch.topk(input, k, dim=None, largest=True, sorted=True, out=None)
        # return: value, index
        _, pred = output.topk(k=maxk, dim=1, largest=True, sorted=True)  # pred: [num of batch, k]
        pred = pred.t()  # pred: [k, num of batch]

        # [1, num of batch] -> [k, num_of_batch] : bool
        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))
        # np.shape(res): [k, 1]
        return res


def ce_loss(logits, targets, use_hard_labels=True, reduction='none'):