invoke-ai/InvokeAI · error · ValueError

Unsupported IP-Adapter type: {type(self.ip_adapter)}

Error message

Unsupported IP-Adapter type: {type(self.ip_adapter)}

What it means

FluxDenoise._normalize_ip_adapter_fields expects the ip_adapter input to be either a single IPAdapterField, a list of them, or None. Any other Python type (str, dict, wrong field class) reaching _run_diffusion raises this ValueError because FLUX cannot interpret the IP-Adapter input shape.

Source

Thrown at invokeai/app/invocations/flux_denoise.py:903

        # Prepare mask conditioning.
        mask = mask[:, 0, :, :]
        # Rearrange mask to a 16-channel representation that matches the shape of the VAE-encoded latent space.
        mask = einops.rearrange(mask, "b (h ph) (w pw) -> b (ph pw) h w", ph=8, pw=8)
        mask = pack(mask)

        # Merge image and mask conditioning.
        img_cond = torch.cat((cond_img, mask), dim=-1)
        return img_cond

    def _normalize_ip_adapter_fields(self) -> list[IPAdapterField]:
        if self.ip_adapter is None:
            return []
        elif isinstance(self.ip_adapter, IPAdapterField):
            return [self.ip_adapter]
        elif isinstance(self.ip_adapter, list):
            return self.ip_adapter
        else:
            raise ValueError(f"Unsupported IP-Adapter type: {type(self.ip_adapter)}")

    def _prep_ip_adapter_image_prompt_clip_embeds(
        self,
        ip_adapter_fields: list[IPAdapterField],
        context: InvocationContext,
        device: torch.device,
    ) -> tuple[list[torch.Tensor], list[torch.Tensor]]:
        """Run the IPAdapter CLIPVisionModel, returning image prompt embeddings."""
        clip_image_processor = CLIPImageProcessor()

        pos_image_prompt_clip_embeds: list[torch.Tensor] = []
        neg_image_prompt_clip_embeds: list[torch.Tensor] = []
        for ip_adapter_field in ip_adapter_fields:
            # `ip_adapter_field.image` could be a list or a single ImageField. Normalize to a list here.
            ipa_image_fields: list[ImageField]
            if isinstance(ip_adapter_field.image, ImageField):
                ipa_image_fields = [ip_adapter_field.image]
            elif isinstance(ip_adapter_field.image, list):

View on GitHub (pinned to 0b6a024f2f)

Solutions

  1. Connect an IP-Adapter invocation (which outputs IPAdapterField) or a list of them to the flux_denoise ip_adapter input.
  2. Re-create the workflow in the current InvokeAI version instead of editing an old exported graph JSON.
  3. If writing a custom node, declare the output type as IPAdapterField and return IPAdapterField or list[IPAdapterField].

Example fix

// before: ip_adapter input wired from a generic model loader output
// after
from invokeai.app.invocations.ip_adapter import IPAdapterInvocation
# wire IPAdapterInvocation.ip_adapter -> FluxDenoiseInvocation.ip_adapter
Defensive patterns

Strategy: type-guard

Validate before calling

from invokeai.app.invocations.ip_adapter import IPAdapterField
assert ip_adapter_input is None or isinstance(ip_adapter_input, (IPAdapterField, list)) and all(isinstance(f, IPAdapterField) for f in (ip_adapter_input if isinstance(ip_adapter_input, list) else [ip_adapter_input]))

Type guard

def is_valid_ip_adapter_input(v) -> bool:
    if v is None or isinstance(v, IPAdapterField):
        return True
    return isinstance(v, list) and all(isinstance(f, IPAdapterField) for f in v)

Try / catch

try:
    result = flux_denoise.invoke(context)
except ValueError as e:
    if str(e).startswith("Unsupported IP-Adapter type"):
        log.error("ip_adapter input must be IPAdapterField or list[IPAdapterField]")
    else:
        raise

Prevention

When it happens

Trigger: Graph wiring passes a non-IPAdapterField value (e.g. a raw model field, image field, or deprecated IP-Adapter invocation output) into the ip_adapter input of the FLUX Denoise invocation.

Common situations: Loading an old workflow JSON created before the ip_adapter field was typed as IPAdapterField; custom nodes emitting the wrong output type; manual graph edits connecting an incompatible output to ip_adapter.

Related errors

AI-assisted analysis of invoke-ai/InvokeAI@0b6a024f2f (2026-08-29). Data as JSON: /api/errors/cc364ddc6089f2be. Report an issue: GitHub.