csvsdataset 0.0.5

Memory frugal torch dataset from a csv collection

csvsdataset

csvsdataset is a Python library designed to simplify the process of working with multiple CSV files as a single dataset. The primary functionality is provided by the CsvsDataset class in the csvsdataset.py module.

Installation

To install the csvsdataset library, simply run:

pip install csvsdataset

Usage

    from csvsdataset.csvsdataset import CsvsDataset
    
    # Initialize the CsvsDataset instance
    dataset = CsvsDataset(folder_path="path/to/your/csv/folder",
                          file_pattern="*.csv",
                          x_columns=["column1", "column2"],
                          y_column="target_column")
    
    # Iterate over the dataset
    for x_data, y_data in dataset:
        # Your processing code here
        pass
    
    # Access a specific item in the dataset
    x_data, y_data = dataset[42]

Memory frugality

Only data from a small number of csv files is maintained in memory. The rest is discarded on a LRU basis. This class is intended for use when a very large number of data files exist which cannot be loaded into memory conveniently.