Skip to main content

model.dataloader

save_normalization​

def save_normalization(train_dir: str | Path, data_train: np.ndarray) -> dict

Compute the model input normalization from the training data and save it to seq_mean.npy and seq_std.npy in train_dir, overwriting existing files.

Returns

  • dict: Normalization statistics, with float values under the keys mean and std.

load_normalization​

def load_normalization(train_dir: str | Path) -> tuple[float, float]

Load the model input normalization (mean, std) saved with the training data. Training and inference both use these values.

SEQUENCE_DATASET Objects​

class SEQUENCE_DATASET(Dataset)

__init__​

def __init__(path_to_file: str,
             data: str,
             train: bool,
             temporal_window: int,
             samples_per_epoch: Optional[int] = None,
             **kwargs) -> None

Initialize the Sequence Dataset. Normalizes the data with the statistics saved by create_trainset at:

  • project_name/
  • data/
    • train/
      • seq_mean.npy
      • seq_std.npy

Parameters

  • path_to_file (str): Path to the dataset files.
  • data (str): Name of the data file.
  • train (bool): Flag indicating whether it's training data.
  • temporal_window (int): Size of the temporal window.

Returns

  • None

__len__​

def __len__() -> int

Return the number of data points.

Returns

  • int: Number of data points.

__getitem__​

def __getitem__(index: int) -> torch.Tensor

Get a normalized sequence at the specified index.

Parameters

  • index (int): Index of the item.

Returns

  • torch.Tensor: Normalized sequence data at the specified index.