sgl-project/sglang · error · ValueError

`b` must be 2D (got b.ndim={b.ndim}).

Error message

`b` must be 2D (got b.ndim={b.ndim}).

What it means

The gate logit b for replaySSM decode must be 2D [num_tokens, HV]: one row of gate logits per token. Passing 1D (per-head only) or 3D tensors fails this precondition.

Source

Thrown at python/sglang/kernels/ops/attention/fla/fused_recurrent_linear_replayssm.py:482

        ``dt_bias``=[HV, K], ``g_cache``=[num_slots, HV, L, K].
    ``A_log`` is [HV] (per-head scalar) for both.

    Same call surface as the packed decode plus the three ring caches
    (``d_cache`` / ``k_cache`` / ``g_cache``) and the per-decode-row
    ``write_pos`` cursor.  ``initial_state`` is both the checkpoint read (h0)
    and the (flush-only) checkpoint write (ht), in place.

    Allocates nothing persistent: the caller owns the ring tensors and is
    responsible for advancing / resetting ``write_pos`` (e.g. ``(write_pos+1) %
    L`` after each step).  This is a STANDALONE kernel; the memory-pool / cache
    integration is a later phase.
    """
    if mixed_qkv.ndim != 2:
        raise ValueError(f"`mixed_qkv` must be 2D (got ndim={mixed_qkv.ndim}).")
    if mixed_qkv.stride(-1) != 1:
        raise ValueError("`mixed_qkv` must be contiguous in the last dim.")
    if b.ndim != 2:
        raise ValueError(f"`b` must be 2D (got b.ndim={b.ndim}).")
    if A_log.ndim != 1:
        raise ValueError("`A_log` must be a 1D tensor.")
    if initial_state.ndim != 4:
        raise ValueError(f"`initial_state` must be 4D (got ndim={initial_state.ndim}).")
    if not out.is_contiguous():
        raise ValueError("`out` must be contiguous.")
    if write_pos.ndim != 1 or write_pos.dtype != torch.int32:
        raise ValueError("`write_pos` must be a 1D int32 tensor.")
    if force_flush is not None and (
        force_flush.ndim != 1 or force_flush.dtype != torch.int32
    ):
        raise ValueError("`force_flush` must be a 1D int32 tensor or None.")

    B = mixed_qkv.shape[0]
    num_state_slots, HV, V, K = initial_state.shape
    qkv_dim = mixed_qkv.shape[1]
    q_dim = (qkv_dim - HV * V) // 2
    if q_dim <= 0 or q_dim % K != 0:

View on GitHub (pinned to 0132848349)

Solutions

1. Broadcast/reshape b to [num_tokens, HV] (e.g. b.expand(T, HV) if constant per token, or b.reshape(-1, HV))
2. Make sure num_tokens matches mixed_qkv.shape[0]

Example fix

// before
b = a_log[None]  # or per-head [HV]
// after
b = b_1d.unsqueeze(0).expand(num_tokens, HV).contiguous()  # [T, HV]

assert b.ndim == 2 and b.shape[0] == mixed_qkv.shape[0]

When it happens

Trigger: Passing b shaped [HV] (a single head-vector), [B, T, HV], or a transposed tensor to fused_recurrent_linear_replayssm_decode.

Common situations: Sharing a cached per-head b/a bias tensor across tokens; reshaping mismatches after switching from a chunked to decode API.

Related errors

• `mixed_qkv` must be 2D (got ndim={mixed_qkv.ndim}).
• `A_log` must be a 1D tensor.
• `initial_state` must be 4D (got ndim={initial_state.ndim}).
• `dt_bias` must have {HV * K} elements (got {dt_bias.numel()}
• Invalid packed Q size {q_dim}: must be divisible by K={K}. K

AI-assisted analysis of sgl-project/sglang@0132848349 (2026-08-28). Data as JSON: /api/errors/df7edcae540f008d. Report an issue: GitHub.