ValueError: val_set must contain 'X_ori' for model validation.

[petteriTeikari]

1. System Info

pypots: 0.8.1

2. Information

• The official example scripts
• My own created scripts

3. Reproduction

Use dataset_val in addition to the dataset (train_set) in the usage instructions

4. Expected behavior

Not sure which one you meant to be the correct key, but your documentation [1] says that there should be X in the dictinary, whereas your code says there should be X_ori [2]

And I guess the validation split needs to contain both X () and X_ori (without the missing values) which is not given in the documentation [1]?

e.g.

dataset_val = {"X": X_val_gt_missing, "X_ori": X_val_gt}

[1] https://docs.pypots.com/en/latest/pypots.imputation.html#pypots.imputation.csdi.CSDI.fit

[2]

[WenjieDu]

Both X_ori and X can contain missing values. X_ori is the one holding ground truth values, and X is without them. X is for training, and X_ori is for the model to calculate training loss or validation error. If you still feel lost here, please read SAITS paper https://arxiv.org/abs/2202.08516 carefully