Skip to main content

​
Overview

Dataset class for LPTM models, supporting forecasting, imputation, detection, and classification tasks.

​
Class signature

class LPTMDataset(BaseDataset):
    def __init__(
        self,
        name: str = None,
        datetime_col: str = None,
        path: str = None,
        batchsize: int = 16,
        mode: str = "train",
        boundaries: list = [0, 0, 0],
        horizon: int = 0,
        task_name: str = "forecasting",
        label_col: str = None,
        stride: int = 10,
        seq_len: int = 512,
        **kwargs,
    )

​
Parameters

​
name
str
default:"None"
Name of the dataset. Used to locate data automatically.
​
datetime_col
str
default:"None"
Name of the datetime column in the CSV file.
​
path
str
default:"None"
Path to the dataset CSV file. If None, uses the loader function for name.
​
batchsize
int
default:"16"
Batch size for DataLoader.
​
mode
str
default:"train"
Mode of operation: 'train' or 'test'.
​
boundaries
list
default:"[0, 0, 0]"
Train/val/test split boundaries. Default splits as: 60% train, 20% val, 20% test.
​
horizon
int
default:"0"
Forecast horizon length.
​
task_name
str
default:"forecasting"
Task type: 'forecasting', 'imputation', 'forecasting2', 'detection', or 'classification'.
​
label_col
str
default:"None"
Column name for labels in classification tasks. Defaults to 'label' if not provided.
​
stride
int
default:"10"
Stride for windowing time series data.
​
seq_len
int
default:"512"
Sequence length for input windows.
​
kwargs
dict
Extra backend-specific options.

​
Methods

​
__len__()

Get the total number of data samples. 
def __len__(self) -> int
Returns: int - Number of samples available for iteration.

​
__getitem__(index)

Get a data sample by index. 
def __getitem__(self, index: int)
Parameters:
  • index (int): Index of the data sample.
Returns: Depending on the task, returns input sequences, masks, forecasts, labels, etc.

​
get_data_loader()

Get a DataLoader for the dataset. 
def get_data_loader()
Returns: DataLoader - PyTorch DataLoader for the dataset.

​
Example usage

from samay.dataset import LPTMDataset

# Forecasting task
dataset = LPTMDataset(
    path="data/timeseries.csv",
    datetime_col="date",
    task_name="forecasting",
    horizon=96,
    seq_len=512,
    mode="train"
)

loader = dataset.get_data_loader()
for batch in loader:
    input_seq, input_mask, forecast_seq = batch
    # Training logic here

​
Task-specific outputs

​
Forecasting

Returns: (input_seq, input_mask, forecast_seq)

​
Imputation

Returns: (input_seq, input_mask)

​
Detection

Returns: (input_seq, input_mask, labels)

​
Classification

Returns: (input_seq, input_mask, labels)

Build docs developers (and LLMs) love

Get started for freeTalk to us