torch. nn. utils. rnn. pad_ Detailed explanation of sequence() [pytoch getting started manual]

The function prototype

torch.nn.utils.rnn.pad_sequence(sequences, batch_first=False, padding_value=0.0)

The functionality

This function returns a size of T x B x * or B x T x * Tensor , among T Is the length of the longest sequence .

Parameters,

• sequences (list[Tensor])： A list of variable length sequences ,shape=[batch_size, N],N Different lengths .
• batch_first (bool, optional) ： Default batch_size In the first dimension
• padding_value (float, optional) ： Fill the value of the , The default is 0.

Example

from torch.nn.utils.rnn import pad_sequence

a = torch.ones(25, 300)
b = torch.ones(22, 300)
c = torch.ones(15, 300)
pad_sequence([a, b, c]).size()

Output

torch.Size([25, 3, 300])