datasets.stocks package

Submodules

datasets.stocks.get_data module

Implements dataloaders for the robotics datasets.

class datasets.stocks.get_data.Grouping(n_groups)

Bases: Module

Module to collate stock data.

__init__(n_groups)

Instantiate grouper. n_groups determines the number of groups.

forward(x)

Apply grouper to input data.

Parameters:

x (torch.Tensor) – Input data.

Returns:

List of outputs

Return type:

list

datasets.stocks.get_data.get_dataloader(stocks, input_stocks, output_stocks, batch_size=16, train_shuffle=True, start_date=datetime.datetime(2000, 6, 1, 0, 0), end_date=datetime.datetime(2021, 2, 28, 0, 0), window_size=500, val_split=3200, test_split=3700, modality_first=True, cuda=True)

Generate dataloader for stock data.

Parameters:

  • stocks (list) – List of strings of stocks to grab data for.

  • input_stocks (list) – List of strings of input stocks

  • output_stocks (list) – List of strings of output stocks

  • batch_size (int, optional) – Batchsize. Defaults to 16.

  • train_shuffle (bool, optional) – Whether to shuffle training dataloader or not. Defaults to True.

  • start_date (datetime, optional) – Start-date to grab data from. Defaults to datetime.datetime(2000, 6, 1).

  • end_date (datetime, optional) – End-date to grab data from. Defaults to datetime.datetime(2021, 2, 28).

  • window_size (int, optional) – Window size. Defaults to 500.

  • val_split (int, optional) – Number of samples in validation split. Defaults to 3200.

  • test_split (int, optional) – Number of samples in test split. Defaults to 3700.

  • modality_first (bool, optional) – Whether to make modality the first index or not. Defaults to True.

  • cuda (bool, optional) – Whether to load data to cuda objects or not. Defaults to True.

Returns:

Tuple of training data-loader, test data-loader, and validation data-loader.

Return type:

tuple

Module contents