nn.utils.rnn.pack_sequence¶

lucid.nn.utils.rnn.pack_sequence(sequences: Iterable[Tensor], enforce_sorted: bool = True) → PackedSequence¶

Function Signature¶

def pack_sequence(
    sequences: Iterable[Tensor],
    enforce_sorted: bool = True,
) -> PackedSequence

Parameters¶

sequences (Iterable[Tensor]): Iterable of tensors, each shaped (seq_len, feature) with matching trailing dimensions.
enforce_sorted (bool, optional): If True, the sequences must be sorted by length in decreasing order. If False, the function will sort internally.

Return Value¶

PackedSequence: Packed representation of the input sequences.

Behavior¶

Internally calls pad_sequence and pack_padded_sequence.

Examples¶

import lucid
import lucid.nn as nn

a = lucid.randn(3, 5)
b = lucid.randn(1, 5)

packed = nn.utils.rnn.pack_sequence([a, b], enforce_sorted=False)