vllm.v1.attention.backends.mamba2_attn ¶

def compute_varlen_chunk_metadata(
    query_start_loc: torch.Tensor,
    chunk_size: int,
) -> tuple[torch.Tensor, torch.Tensor, torch.Tensor]:
    """
    Build chunk-aligned, variable-length metadata used by Mamba2 SSD kernels.

    Given per-sequence cumulative token starts `query_start_loc` of shape [B+1]
    and a physical `chunk_size`, returns three tensors on the same device:
      - cu_chunk_seqlens:  (nchunks+1,) int32   exclusive prefix-sum of
        logical-chunk lengths (each logical chunk never crosses a sequence or
        physical-chunk boundary).
      - last_chunk_indices: (B,)       int32   index of the last logical chunk
        for each sequence (=-1 for empty sequences).
      - seq_idx_chunks:     (nchunks,) int32   sequence index for each logical
        chunk in order.

    This is intentionally lightweight and CPU-side; it mirrors the metadata
    produced by the V1 Mamba2 meta-data builder and is exported so tests
    (and other callers) can avoid duplicating the logic.
    """
    assert query_start_loc.ndim == 1, "query_start_loc must be 1-D [B+1]"
    assert int(query_start_loc[0].item()) == 0, "query_start_loc[0] must be 0"
    device = query_start_loc.device

    qsl64 = query_start_loc.to(torch.int64)
    starts = qsl64[:-1].tolist()
    ends = qsl64[1:].tolist()
    total = int(qsl64[-1].item())

    chunk_lens: list[int] = []
    seq_idx_chunks: list[int] = []
    last_chunk_indices: list[int] = [-1] * len(starts)

    for b, (s, e) in enumerate(zip(starts, ends)):
        if e <= s:
            # empty sequence
            continue
        pos = s
        while pos < e:
            # split at both sequence boundaries and physical chunk boundaries
            room = chunk_size - (pos % chunk_size)
            take = min(room, e - pos)
            chunk_lens.append(int(take))
            seq_idx_chunks.append(b)
            last_chunk_indices[b] = len(chunk_lens) - 1
            pos += take

    # Exclusive prefix sum over logical-chunk lengths
    if chunk_lens:
        cu_chunk_seqlens = torch.tensor(
            [0] + list(itertools.accumulate(chunk_lens)),
            device=device,
            dtype=torch.int32,
        )
        # Final boundary must equal total tokens
        assert int(cu_chunk_seqlens[-1].item()) == total
    else:
        cu_chunk_seqlens = torch.tensor([0], device=device, dtype=torch.int32)

    last_chunk_indices_t = (
        torch.tensor(last_chunk_indices, device=device, dtype=torch.int32)
        if len(starts) > 0
        else torch.empty((0,), device=device, dtype=torch.int32)
    )
    seq_idx_chunks_t = torch.tensor(seq_idx_chunks, device=device, dtype=torch.int32)
    return cu_chunk_seqlens, last_chunk_indices_t, seq_idx_chunks_t