Comfy-Org/ComfyUI · error · ValueError

Invalid causality_axis: {self.causality_axis}

Error message

Invalid causality_axis: {self.causality_axis}

What it means

Raised in CausalConv2d.forward's padding-trim match when self.causality_axis does not match any known CausalityAxis case. It mirrors the init-time guard (error 184) and fires at inference time if the axis was set to an unexpected value after construction.

Source

Thrown at comfy/ldm/lightricks/vae/causal_audio_autoencoder.py:248

            # 2: [0,0,1]
            # 3: [0,1,1]
            # 4: [1,1,2]
            # 5: [1,2,2]
            # Notice that the first and second elements in the output rely only on the first element in the input,
            # while all other elements rely on two elements in the input.
            # So we can drop the first element to undo the padding (rather than the last element).
            # This is a no-op for non-causal convolutions.
            match self.causality_axis:
                case CausalityAxis.NONE:
                    pass  # x remains unchanged
                case CausalityAxis.HEIGHT:
                    x = x[:, :, 1:, :]
                case CausalityAxis.WIDTH:
                    x = x[:, :, :, 1:]
                case CausalityAxis.WIDTH_COMPATIBILITY:
                    pass  # x remains unchanged
                case _:
                    raise ValueError(f"Invalid causality_axis: {self.causality_axis}")

        return x


class Downsample(nn.Module):
    """
    A downsampling layer that can use either a strided convolution
    or average pooling. Supports standard and causal padding for the
    convolutional mode.
    """

    def __init__(self, in_channels, with_conv, causality_axis: CausalityAxis = CausalityAxis.WIDTH):
        super().__init__()
        self.with_conv = with_conv
        self.causality_axis = causality_axis

        if self.causality_axis != CausalityAxis.NONE and not self.with_conv:
            raise ValueError("causality is only supported when `with_conv=True`.")

View on GitHub (pinned to 1c6d8d45b3)

Solutions

1. Re-set the attribute to a valid member: conv.causality_axis = CausalityAxis.WIDTH
2. Fix the loading code so enum-typed attributes are converted through the enum, not assigned raw
3. Reconstruct the module via its normal __init__ with a validated causality_axis

Example fix

# before
conv.causality_axis = "width"  # raw string assigned post-init

# after
conv.causality_axis = CausalityAxis.WIDTH

assert isinstance(conv.causality_axis, CausalityAxis), "causality_axis must be a CausalityAxis member"

def is_causality_axis(v) -> bool:
    return isinstance(v, CausalityAxis)

When it happens

Trigger: Assigning an invalid value to conv.causality_axis after construction, or constructing the module through a path that skips __init__ validation (e.g. custom unpickling or meta-device instantiation followed by manual field assignment).

Common situations: Checkpoint loading code that restores attributes from a state dict/config with foreign enum values; test code mutating module attributes directly; duplicated enum classes across module copies.

Related errors

• Invalid causality_axis: {causality_axis}
• {cls.__name__} does not have a NONE member to map None to
• Invalid {cls.__name__} string: '{value}'. Valid values are:
• Cannot convert type {type(value).__name__} to {cls.__name__}
• Unknown attention type: {attn_type}

AI-assisted analysis of Comfy-Org/ComfyUI@1c6d8d45b3 (2026-08-14). Data as JSON: /api/errors/7907fc24285f27ba. Report an issue: GitHub.