Custom Dataset Class

Problem Statement

Create a custom Dataset class that generates data on-the-fly instead of storing it all in memory.

Background

For large datasets, storing everything in memory isn't practical. PyTorch's Dataset interface lets you generate or load data on demand by implementing three methods: init, len, and getitem.

Your Task

Create a class SequenceDataset that generates sequential data. For index i, it should return:

Feature: a tensor of [i, i+1, i+2]
Label: the sum of the feature values

The class must accept a size parameter that determines how many samples it contains.

Output Format

Each sample should be a tuple of (feature_tensor, label_tensor).

Example:

Input:

features=[[1.0, 2.0], [3.0, 4.0]], labels=[0, 1]

Output:

Dataset with 2 samples, dataset[0] returns (tensor([1., 2.]), tensor(0))

Reasoning:

We create a Dataset class that stores features and labels and returns them as tensors

Constraints:

Features and labels lists will have the same length
Features can be 1D or 2D lists
Labels are integers
Return tensors (not lists) from getitem