sgl-project/sglang · error · ValueError

timestep must be contiguous

Error message

timestep must be contiguous

What it means

The Triton kernel for ltx2_ada_values9 assumes a contiguous timestep tensor so it can compute flat row indices. Non-contiguous inputs (transposed, sliced, or expanded views) are rejected before launch.

Source

Thrown at python/sglang/kernels/ops/diffusion/modulate/ltx2_ada_values_triton.py:147

    tl.store(out2_ptr + base, (table2 + temb2).to(tl.bfloat16), mask=mask)
    tl.store(out3_ptr + base, (table3 + temb3).to(tl.bfloat16), mask=mask)
    tl.store(out4_ptr + base, (table4 + temb4).to(tl.bfloat16), mask=mask)
    tl.store(out5_ptr + base, (table5 + temb5).to(tl.bfloat16), mask=mask)
    tl.store(out6_ptr + base, (table6 + temb6).to(tl.bfloat16), mask=mask)
    tl.store(out7_ptr + base, (table7 + temb7).to(tl.bfloat16), mask=mask)
    tl.store(out8_ptr + base, (table8 + temb8).to(tl.bfloat16), mask=mask)


def ltx2_ada_values9(
    scale_shift_table: torch.Tensor,
    timestep: torch.Tensor,
) -> tuple[torch.Tensor, ...]:
    if timestep.ndim != 3:
        raise ValueError("timestep must have shape [B, S, 9 * D]")
    if not timestep.is_cuda or timestep.dtype != torch.bfloat16:
        raise ValueError("timestep must be a CUDA bfloat16 tensor")
    if not timestep.is_contiguous():
        raise ValueError("timestep must be contiguous")
    if scale_shift_table.ndim != 2 or scale_shift_table.shape[0] != 9:
        raise ValueError("scale_shift_table must have shape [9, D]")
    if (
        not scale_shift_table.is_cuda
        or scale_shift_table.dtype not in (torch.bfloat16, torch.float32)
        or scale_shift_table.stride(-1) != 1
    ):
        raise ValueError(
            "scale_shift_table must be CUDA, bf16/fp32, last-dim contiguous"
        )

    total_params = int(scale_shift_table.shape[0])
    hidden = int(scale_shift_table.shape[1])
    if hidden <= 0 or timestep.shape[-1] != total_params * hidden:
        raise ValueError("timestep last dim must equal 9 * hidden")
    if hidden % 256 != 0 or hidden > 8192:
        raise ValueError("hidden size is outside the supported LTX2 fast-path range")

View on GitHub (pinned to 0132848349)

Solutions

1. Call .contiguous() on timestep before the call
2. Materialize expanded views: emb.expand(B, S, D).contiguous()
3. Use .reshape instead of views/slices that yield non-contiguous results

Example fix

# before
emb = emb[:, None, :].expand(B, S, 9*D)
vals = ltx2_ada_values9(table, emb)
# after
emb = emb[:, None, :].expand(B, S, 9*D).contiguous()
vals = ltx2_ada_values9(table, emb)

if not timestep.is_contiguous():
    timestep = timestep.contiguous()

def contiguous_or_fix(t):
    return t if t.is_contiguous() else t.contiguous()

When it happens

Trigger: Passing an expanded view (timestep[:, None, :].expand(...)) without materializing, a transposed tensor, or a slice along batch/seq that breaks contiguity.

Common situations: Broadcasting a [B, 1, 9*D] embedding across sequence positions with expand (stride-0 seq dim); memory-layout tricks from a checkpoint; sliced batch views in pipeline parallelism.

Related errors

• `mixed_qkv` must be contiguous in the last dim.
• q, k, and v must be contiguous in head_size
• scale_shift_table must be CUDA, bf16/fp32, last-dim contiguo
• hidden size is outside the supported LTX2 fast-path range
• kv-canary: {name} must be contiguous

AI-assisted analysis of sgl-project/sglang@0132848349 (2026-08-28). Data as JSON: /api/errors/61b1eb93923fef31. Report an issue: GitHub.