ask the two file data_splits.p corrupted_img_paths.pkl

[liujun0621]

excuse me,when run train.py as follows:
rank_zero_deprecation(
Traceback (most recent call last):
File "/root/Sybil-main/scripts/train.py", line 415, in
train(args)
File "/root/Sybil-main/scripts/train.py", line 359, in train
args, get_dataset(args.dataset, "train", args)
File "/root/Sybil-main/sybil/utils/helpers.py", line 18, in get_dataset
return SUPPORTED_DATASETS[dataset_name](args, split)
File "/root/Sybil-main/sybil/datasets/nlst.py", line 102, in init
self.dataset = self.create_dataset(split_group)
Connection to connect.cqa1.seetacloud.com closed by remote host.reate_dataset
Connection to connect.cqa1.seetacloud.com closed.paths"]
File "/root/Sybil-main/sybil/datasets/nlst.py", line 495, in CORRUPTED_PATHS
C:\Users\liujun>e.load(open(CORRUPTED_PATHS, "rb"))
EOFError: Ran out of input

where are the file from?
/Mounts/rbg-storage1/datasets/NLST/Shetty_et_al(Google)/data_splits.p
/Mounts/rbg-storage1/datasets/NLST/corrupted_img_paths.pkl

thanks you

[pgmikhael]

Hi,

If you're training on a different set of splits, then you can ignore the use of data_splits.p. This just needs some simple alteration of the code. Otherwise, I have uploaded the file here.

You should ignore the use of the corrupted_img_paths.pkl file -- this is a result of our own download/conversion pipeline of the data, which is not likely to be relevant. Again, you can comment / remove from the nlst.py file the lines referencing that file.

[liujun0621]

when i annotation code CORRUPTED_PATHS = "/Mounts/rbg-storage1/datasets/NLST/corrupted_img_paths.pkl"
there are a lots of error,what can i do next?thanks you very much

[pgmikhael]

Hi,

I'm not sure what "annotation code" means -- but you simply need to comment out the line that reads the file and those using it in the if-statement. This is all in the nlst.py file.

[liujun0621]

annotation code as # or '''

[liujun0621]

there are a lots of depends
as follows
Traceback (most recent call last):
File "/root/Sybil-main/scripts/train.py", line 415, in
train(args)
File "/root/Sybil-main/scripts/train.py", line 358, in train
train_dataset = loaders.get_train_dataset_loader(
File "/root/Sybil-main/sybil/utils/loading.py", line 103, in get_train_dataset_loader
weights=train_data.weights, num_samples=len(train_data), replacement=True
AttributeError: 'NLST_Survival_Dataset' object has no attribute 'weights'

[pgmikhael]

Hi,

Yes – there are many dependencies within the code. For "weights" specifically, it is utilized with the --class_bal flag is set to True.