invoke-ai/InvokeAI · error · ValueError

size {size} has to be smaller or equal to {dim}.

Error message

size {size} has to be smaller or equal to {dim}.

What it means

Each per-layer attention slice size must not exceed that layer's head dimension (sliceable_head_dims[i]). This guard rejects any slice entry larger than the corresponding layer's dim, since slicing above the dimension size is meaningless and would misconfigure chunked attention.

Source

Thrown at invokeai/backend/util/hotfixes.py:550

            # speed and memory
            slice_size = [dim // 2 for dim in sliceable_head_dims]
        elif slice_size == "max":
            # make smallest slice possible
            slice_size = num_sliceable_layers * [1]

        slice_size = num_sliceable_layers * [slice_size] if not isinstance(slice_size, list) else slice_size

        if len(slice_size) != len(sliceable_head_dims):
            raise ValueError(
                f"You have provided {len(slice_size)}, but {self.config} has {len(sliceable_head_dims)} different"
                f" attention layers. Make sure to match `len(slice_size)` to be {len(sliceable_head_dims)}."
            )

        for i in range(len(slice_size)):
            size = slice_size[i]
            dim = sliceable_head_dims[i]
            if size is not None and size > dim:
                raise ValueError(f"size {size} has to be smaller or equal to {dim}.")

        # Recursively walk through all the children.
        # Any children which exposes the set_attention_slice method
        # gets the message
        def fn_recursive_set_attention_slice(module: torch.nn.Module, slice_size: List[int]):
            if hasattr(module, "set_attention_slice"):
                module.set_attention_slice(slice_size.pop())

            for child in module.children():
                fn_recursive_set_attention_slice(child, slice_size)

        reversed_slice_size = list(reversed(slice_size))
        for module in self.children():
            fn_recursive_set_attention_slice(module, reversed_slice_size)

    def _set_gradient_checkpointing(self, module, value=False):
        if isinstance(module, (CrossAttnDownBlock2D, DownBlock2D)):
            module.gradient_checkpointing = value

View on GitHub (pinned to 0b6a024f2f)

Solutions

1. Lower each slice_size entry so it is <= the corresponding layer's dim
2. Use 'auto' or a small power-of-two int (e.g. 2, 4, 8) instead of a custom list
3. Verify sliceable_head_dims in the model config and size the list to it

Example fix

// before
model.set_attention_slice([128, 128])
// after
model.set_attention_slice([64, 64])  # each <= matching layer dim, or use "auto"

dims = model.config.attention_head_dim  # list of per-layer dims
if isinstance(slice_size, list):
    slice_size = [min(s, d) if s is not None else None for s, d in zip(slice_size, dims)]

def slice_sizes_in_range(slice_size, dims):
    return all(s is None or s <= d for s, d in zip(slice_size, dims))

try:
    model.set_attention_slice(slice_size)
except ValueError:
    model.set_attention_slice("auto")

When it happens

Trigger: Calling set_attention_slice with a list containing an entry greater than the matching layer's dim, e.g. set_attention_slice([128]) on a layer with dim 64, or reusing a slice list tuned for a different model.

Common situations: Copy-pasting slice sizes between SD 1.x/2.x/SDXL models whose attention head dims differ; hand-tuning VRAM-saving slice values; config drift after model or library upgrades.

Related errors

• You have provided {len(slice_size)}, but {self.config} has {
• Invalid or expired token
• User not found or inactive
• Missing authentication credentials
• Invalid or expired authentication token

AI-assisted analysis of invoke-ai/InvokeAI@0b6a024f2f (2026-08-29). Data as JSON: /api/errors/d2d577379d0a9d1b. Report an issue: GitHub.