What causes "activation_chunk_size must be positive" in invoke-ai/InvokeAI?

Pass a positive activation_chunk_size (e.g. 8, 16, or 32) when enabling the optimization.

invoke-ai/InvokeAI · error · ValueError

activation_chunk_size must be positive

Error message

activation_chunk_size must be positive

What it means

wan_memory_optimization is a context manager that patches Wan transformer blocks to chunk pointwise activations in fixed-size groups, trading compute for peak memory. A non-positive activation_chunk_size (0 or negative) is meaningless for chunking and would break the internal batching logic, so the generator raises this ValueError before patching anything.

Source

Thrown at invokeai/backend/wan/memory_optimization.py:208

        )
        output[:, start:end].copy_(output_chunk)

    return output


@contextmanager
def wan_memory_optimization(
    transformer: torch.nn.Module,
    *,
    enabled: bool,
    activation_chunk_size: int = WAN_ACTIVATION_CHUNK_SIZE,
) -> Iterator[None]:
    """Temporarily chunk Wan transformer pointwise activations during inference."""
    if not enabled:
        yield
        return
    if activation_chunk_size <= 0:
        raise ValueError("activation_chunk_size must be positive")

    blocks: Any = getattr(transformer, "blocks", None)
    if blocks is None:
        raise TypeError(f"Expected a Wan transformer with blocks, got {type(transformer).__name__}.")
    blocks = list(blocks)
    if hasattr(transformer, "_invokeai_original_forward") or any(
        hasattr(block, "_invokeai_original_forward") for block in blocks
    ):
        raise RuntimeError("Wan memory optimization context cannot be nested.")

    patched_blocks: list[tuple[torch.nn.Module, Any, bool]] = []
    original_transformer_forward = transformer.forward
    transformer_had_instance_forward = "forward" in transformer.__dict__
    patch_transformer_forward = all(
        hasattr(transformer, name)
        for name in ("condition_embedder", "patch_embedding", "proj_out", "rope", "scale_shift_table")
    )
    try:

View on GitHub (pinned to 0b6a024f2f)

Solutions

Pass a positive activation_chunk_size (e.g. 8, 16, or 32) when enabling the optimization.
Set enabled=False instead of activation_chunk_size=0 to disable chunking.
Clamp user config: activation_chunk_size = max(1, int(cfg_value)).

Example fix

// before
with wan_memory_optimization(transformer, enabled=True, activation_chunk_size=0):
// after
with wan_memory_optimization(transformer, enabled=True, activation_chunk_size=16):

Defensive patterns

Strategy: validation

Validate before calling

chunk = max(1, int(config.get("activation_chunk_size", 16)))
with wan_memory_optimization(transformer, enabled=enable, activation_chunk_size=chunk):
    run_diffusion(...)

Try / catch

try:
    with wan_memory_optimization(transformer, True, activation_chunk_size=cfg_chunk):
        run_diffusion()
except ValueError as e:
    if "must be positive" in str(e):
        run_diffusion()  # proceed without chunking

Prevention

Treat 0 as 'disabled' by mapping it to enabled=False instead of chunk_size=0.
Clamp config values with max(1, value).
Validate chunk-size configs at startup, not at inference time.

When it happens

Trigger: Entering `with wan_memory_optimization(transformer, enabled=True, activation_chunk_size=0)` (or any value <= 0); e.g. chunk size loaded from config as 0 or computed as len//x when the denominator exceeds length.

Common situations: Config file with chunk_size: 0 intending 'auto', dividing to get zero, CLI flag defaulting to 0.

Related errors

AI-assisted analysis of invoke-ai/InvokeAI@0b6a024f2f (2026-08-29). Data as JSON: /api/errors/b6c3cb893014cc0b. Report an issue: GitHub.