Constants API Reference

This page provides the complete API reference for the batchtensor.constants module.

The constants module defines dimension indices used throughout the library to identify batch and sequence dimensions, ensuring consistency across all operations.

For more information about dimension conventions, see the Constants User Guide.

batchtensor.constants

Defines important constants for batch and sequence dimensions.

This module provides standardized dimension indices used throughout the batchtensor package. These constants ensure consistent dimension handling across all tensor operations.

Constants

BATCH_DIM: The batch dimension index (0). This is the first dimension of tensors and represents independent samples in a batch. For example, in a tensor of shape (batch_size, seq_len, feature_dim), this refers to the batch_size dimension. SEQ_DIM: The sequence dimension index (1). This is the second dimension of tensors and represents sequential data within each batch item. For example, in a tensor of shape (batch_size, seq_len, feature_dim), this refers to the seq_len dimension.

Example

>>> from batchtensor.constants import BATCH_DIM, SEQ_DIM
>>> import torch
>>> # Create a batch of 3 sequences, each of length 5
>>> tensor = torch.randn(3, 5, 10)  # (batch, seq, features)
>>> # BATCH_DIM=0 refers to the dimension with size 3
>>> # SEQ_DIM=1 refers to the dimension with size 5
>>> tensor.shape[BATCH_DIM]
3
>>> tensor.shape[SEQ_DIM]
5

batchtensor.constants.BATCH_DIM module-attribute

BATCH_DIM = 0

int: The index of the batch dimension in tensors.

This constant is used throughout batchtensor to identify the batch dimension, which is always assumed to be the first dimension (index 0) of tensors.

batchtensor.constants.SEQ_DIM module-attribute

SEQ_DIM = 1

int: The index of the sequence dimension in tensors.

This constant is used throughout batchtensor to identify the sequence dimension, which is always assumed to be the second dimension (index 1) of tensors.