mynameischaos/gcc

Graph Contrastive Clustering (ICCV2021)

gcc's Issues

关于图构建时滑动平均的问题

注意到这几行代码：

    @torch.no_grad()
    def push(self, feats, indexes):
        for i in range(indexes.shape[0]):
            key = indexes[i].item()
            #data = feats[i][0].cpu().detach().numpy()
            data = feats[i].unsqueeze_(0)
            if key not in self.memory_dict.keys():
                self.memory_dict[key] = [data]
                #self.memory_dict[key] = data
            elif len(self.memory_dict[key]) == self.size:
                self.memory_dict[key].pop(0)
                self.memory_dict[key].append(data)
            else:
                self.memory_dict[key].append(data)
                #self.memory_dict[key] = self.alpha * self.memory_dict[key] + (1.0 - self.alpha) * data

模型在训练的时候好像并没有使用到滑动平均。想知道论文中提到的moving average提现在哪里？

感谢！

可以提供imagenet_dogs_data_list.txt和imagenet_dogs_label_list.txt

您好，非常感谢您的代码，很有帮助。但是Imagenet以及imagenet-dogs之前只使用过一个txt的，请问可以提供一下imagenet10_data_list.txt imagenet10_label_list.txt imagenet_dogs_data_list.txt imagenet_dogs_label_list.txt吗，非常感谢

About pretext documents and training time

Hello!
I have been unable to download pretexte files because it needs permission to access. Could you give me this permission because the training model is a waste of time?
And another question, how long have you been running 1000 epochs on cifar-10 or other datasets? It is not clearly stated in the paper?
thank you！

About simclr_train

could you upload the “simclr_train" function

Some question about data augmentation

Hi,

When you get the dataset, it looks like you've done the transformation in your code, but then you do the transformation again in the End2EndDataset.

Do I need to transform the data twice？

Thanks,

Zhaoyu.

没有simclr_train的问题

simclr.py文件中from utils.train_utils import simclr_train这句代码，utils.train_utils是没有simclr_train的，只有gcc_train，这个怎么说？

Can you provide the datasets of ImageNet-10, ImageNet-Dog-15 and Tiny-ImageNet-200?

Hello,
Thanks for your code. It helps me a lot but now I'm confused about how to get the ImageNet-10, ImageNet-Dog-15 and Tiny-ImageNet-200.

报错

请问 PermissionError: [Errno 13] Permission denied: '/gruntdata' 这个问题怎么解决？