Comfy-Org/ComfyUI · critical · ValueError

Got {params.axes_dim} but expected positional dim {pe_dim}

Error message

Got {params.axes_dim} but expected positional dim {pe_dim}

What it means

ChromaRadiance requires sum(axes_dim) == hidden_size // num_heads (pe_dim) for its rotational position embeddings, identical to the Chroma check. axes_dim distributes the per-head dimension across the positional axes (typically two spatial axes plus a time/extra axis in radiance models); an inconsistent sum means RoPE cannot be constructed over the per-head dim. The check runs right after the divisibility check in __init__.

Source

Thrown at comfy/ldm/chroma_radiance/model.py:65

    """

    def __init__(self, image_model=None, final_layer=True, dtype=None, device=None, operations=None, **kwargs):
        if operations is None:
            raise RuntimeError("Attempt to create ChromaRadiance object without setting operations")
        nn.Module.__init__(self)
        self.dtype = dtype
        params = ChromaRadianceParams(**kwargs)
        self.params = params
        self.patch_size = params.patch_size
        self.in_channels = params.in_channels
        self.out_channels = params.out_channels
        if params.hidden_size % params.num_heads != 0:
            raise ValueError(
                f"Hidden size {params.hidden_size} must be divisible by num_heads {params.num_heads}"
            )
        pe_dim = params.hidden_size // params.num_heads
        if sum(params.axes_dim) != pe_dim:
            raise ValueError(f"Got {params.axes_dim} but expected positional dim {pe_dim}")
        self.hidden_size = params.hidden_size
        self.num_heads = params.num_heads
        self.in_dim = params.in_dim
        self.out_dim = params.out_dim
        self.hidden_dim = params.hidden_dim
        self.n_layers = params.n_layers
        self.pe_embedder = EmbedND(dim=pe_dim, theta=params.theta, axes_dim=params.axes_dim)
        self.img_in_patch = operations.Conv2d(
            params.in_channels,
            params.hidden_size,
            kernel_size=params.patch_size,
            stride=params.patch_size,
            bias=True,
            dtype=dtype,
            device=device,
        )
        self.txt_in = operations.Linear(params.context_in_dim, self.hidden_size, dtype=dtype, device=device)
        # set as nn identity for now, will overwrite it later.

View on GitHub (pinned to 1c6d8d45b3)

Solutions

1. Set axes_dim entries so they sum to hidden_size // num_heads
2. Derive the last axis as pe_dim minus the fixed spatial axes instead of hard-coding all entries
3. Keep the checkpoint's original axes_dim; do not override it in transformer options

Example fix

# before
# hidden_size=3000, num_heads=24 -> pe_dim=125
axes_dim = [16, 56, 56]  # sum 128 != 125

# after
axes_dim = [125 - 56 - 56, 56, 56]  # [13, 56, 56], sums to 125

pe_dim = config["hidden_size"] // config["num_heads"]
if sum(config["axes_dim"]) != pe_dim:
    spatial = config["axes_dim"][1:]
    config["axes_dim"] = [pe_dim - sum(spatial)] + spatial

When it happens

Trigger: Constructing ChromaRadiance with an axes_dim list whose sum differs from hidden_size // num_heads, e.g. reusing Flux's [16, 56, 56] with a 125-dim head, or editing the spatial axis sizes for higher-resolution radiance output without rebalancing the total.

Common situations: Tuning axes_dim for custom resolutions/view counts in radiance generation; porting configs between Chroma and ChromaRadiance; hand-built configs for NeRF-style experiments.

Related errors

• Got {params.axes_dim} but expected positional dim {pe_dim}
• Hidden size {params.hidden_size} must be divisible by num_he
• Unsupported nerf_final_head_type {params.nerf_final_head_typ
• Unknown interpolation method {self.interpolation}
• Got {params.axes_dim} but expected positional dim {pe_dim}

AI-assisted analysis of Comfy-Org/ComfyUI@1c6d8d45b3 (2026-08-14). Data as JSON: /api/errors/5f0c558593badfe2. Report an issue: GitHub.