h5datamodule.yaml

_target_: src.datamodules.datamodules.SingleDataModule

defaults:
  - _self_
  - loaders: default.yaml
  - transforms: default.yaml

# Choose labels tensor type in dataset based on loss type you use, e.g.:
# - torch.nn.BCEWithLogitsLoss requires torch.FloatTensor
# - torch.nn.CrossEntropyLoss requires torch.LongTensor

datasets:
  train:
    _target_: src.datamodules.datasets.ClassificationDataset
    json_path: ${paths.data_dir}/train/data.json
    data_path: ${paths.data_dir}/train/data.h5
    label_type: "torch.FloatTensor"

  valid:
    _target_: src.datamodules.datasets.ClassificationDataset
    json_path: ${paths.data_dir}/valid/data.json
    data_path: ${paths.data_dir}/valid/data.h5
    label_type: "torch.FloatTensor"

  test:
    _target_: src.datamodules.datasets.ClassificationDataset
    json_path: ${paths.data_dir}/test/data.json
    data_path: ${paths.data_dir}/test/data.h5
    label_type: "torch.FloatTensor"

  predict:
    dataset1:
      _target_: src.datamodules.datasets.ClassificationDataset
      json_path: ${paths.data_dir}/predict/data1.json
      data_path: ${paths.data_dir}/predict/data1.h5
      label_type: "torch.FloatTensor"
    dataset2:
      _target_: src.datamodules.datasets.ClassificationDataset
      json_path: ${paths.data_dir}/predict/data2.json
      data_path: ${paths.data_dir}/predict/data2.h5
      label_type: "torch.FloatTensor"