michalznalezniak / contrastive-hierarchical-clustering Goto Github PK

thank you for your patience.
i want to know what is the role of memory_data?
And if I want to design a customised dataset where all the data is used for training and then all the data is clustered, how should I do it?

Dear author,

Thanks for your code. I want to present the result clearly like the figure in your paper, how can I get the hierarchical labels and the corresponding probabilities? This is my email address: [email protected]. I will be very appreciated if you can response to me.

Hello sir, there seems to be a problem with your requirement.txt, I can't install the version you have in it!

https://github.com/MichalZnalezniak/Contrastive-Hierarchical-Clustering/blob/2c418fcc763081e89073d83d648308d277b1fd10/tree_losses.py#L31C1-L31C28

Hello, I tried running the code but I get this error:

Train Epoch: [1/4] Loss: 5.3809: 100%|██████████| 195/195 [02:26<00:00,  1.34it/s]
Train Epoch: [2/4] Loss: 5.1769: 100%|██████████| 195/195 [02:31<00:00,  1.29it/s]
  0%|          | 0/195 [00:33<?, ?it/s]
Traceback (most recent call last):
  File "/Contrastive-Hierarchical-Clustering-main/main.py", line 195, in <module>
    total_loss, tree_loss_train, reg_loss_train, simclr_loss_train = train(model, train_loader, optimizer, epoch, cfg, device)
  File "/Contrastive-Hierarchical-Clustering-main/main.py", line 47, in train
    tree_loss_value = tree_loss(tree_output1, tree_output2, batch_size, net.masks_for_level, mean_of_probs_per_level_per_epoch, cfg.tree.tree_level, device)
  File "/Contrastive-Hierarchical-Clustering-main/tree_losses.py", line 36, in tree_loss
    labels = labels * ~mask_for_level
TypeError: bad operand type for unary ~: 'dict'

What is the purpose of this mask here? The mask is a dict, what is the intent?

Dear author,

There is an error when I evaluate the dataset cifar10. I run the command: python evaluate.py --save_point ./pre-trained/CIFAR10/models/ --dataset-name cifar10 # Generate metrics for CIFAR10 and then the error occurs: