invoke-ai/InvokeAI · error · ValueError

encoder_hid_dim_type: {encoder_hid_dim_type} must be None, '

Error message

encoder_hid_dim_type: {encoder_hid_dim_type} must be None, 'text_proj' or 'text_image_proj'.

What it means

The model constructor only supports three values for `encoder_hid_dim_type`: None, 'text_proj', and 'text_image_proj'. Any other string falls through all supported branches and hits this ValueError. The library enumerates its supported encoder-projection backends explicitly, so an unsupported/typo'd value cannot be silently ignored.

Source

Thrown at invokeai/backend/util/hotfixes.py:223

            raise ValueError(
                f"`encoder_hid_dim` has to be defined when `encoder_hid_dim_type` is set to {encoder_hid_dim_type}."
            )

        if encoder_hid_dim_type == "text_proj":
            self.encoder_hid_proj = nn.Linear(encoder_hid_dim, cross_attention_dim)
        elif encoder_hid_dim_type == "text_image_proj":
            # image_embed_dim DOESN'T have to be `cross_attention_dim`. To not clutter the __init__ too much
            # they are set to `cross_attention_dim` here as this is exactly the required dimension ...
            # for the currently only use
            # case when `addition_embed_type == "text_image_proj"` (Kadinsky 2.1)`
            self.encoder_hid_proj = TextImageProjection(
                text_embed_dim=encoder_hid_dim,
                image_embed_dim=cross_attention_dim,
                cross_attention_dim=cross_attention_dim,
            )

        elif encoder_hid_dim_type is not None:
            raise ValueError(
                f"encoder_hid_dim_type: {encoder_hid_dim_type} must be None, 'text_proj' or 'text_image_proj'."
            )
        else:
            self.encoder_hid_proj = None

        # class embedding
        if class_embed_type is None and num_class_embeds is not None:
            self.class_embedding = nn.Embedding(num_class_embeds, time_embed_dim)
        elif class_embed_type == "timestep":
            self.class_embedding = TimestepEmbedding(timestep_input_dim, time_embed_dim)
        elif class_embed_type == "identity":
            self.class_embedding = nn.Identity(time_embed_dim, time_embed_dim)
        elif class_embed_type == "projection":
            if projection_class_embeddings_input_dim is None:
                raise ValueError(
                    "`class_embed_type`: 'projection' requires `projection_class_embeddings_input_dim` be set"
                )
            # The projection `class_embed_type` is the same as the timestep `class_embed_type` except

View on GitHub (pinned to 0b6a024f2f)

Solutions

1. Set `encoder_hid_dim_type` to one of: None, 'text_proj', or 'text_image_proj' (check exact spelling/casing).
2. If you need 'text_image' style projection, use 'text_image_proj' and supply both `encoder_hid_dim` and `cross_attention_dim`.
3. If the value comes from a checkpoint config for a feature this vendored class lacks, upgrade InvokeAI/diffusers or use the stock diffusers class instead of the hotfix copy.
4. Print/inspect the offending config value before constructing to catch stray whitespace or case mismatches.

Example fix

// before
model = ControlNetModel2_5(
    encoder_hid_dim=1024,
    encoder_hid_dim_type="image_proj",
)
// after
model = ControlNetModel2_5(
    encoder_hid_dim=1024,
    encoder_hid_dim_type="text_image_proj",
    cross_attention_dim=1024,
)

ALLOWED = {None, "text_proj", "text_image_proj"}
t = model_config.get("encoder_hid_dim_type")
if t not in ALLOWED:
    raise ValueError(f"encoder_hid_dim_type={t!r} not in {sorted(str(a) for a in ALLOWED)}")

from typing import Optional, Literal
def is_valid_encoder_hid_dim_type(v: Optional[str]) -> bool:
    return v in (None, "text_proj", "text_image_proj")

try:
    model = ControlNetModel2_5(**cfg)
except ValueError as e:
    if "encoder_hid_dim_type" in str(e):
        cfg["encoder_hid_dim_type"] = None
        model = ControlNetModel2_5(**cfg)
    else:
        raise

When it happens

Trigger: Passing `encoder_hid_dim_type` with a value such as 'image_proj', 'ip_adapter', 'text', or a misspelling like 'text-proj'/'TextProj' to the model constructor or in a loaded config.json. Also occurs when a config from a newer diffusers release uses an encoder_hid_dim_type variant this vendored class does not implement.

Common situations: Typos in hand-written configs; copying an encoder_hid_dim_type from a different architecture (e.g. UNet variants that support more types); running an older vendored hotfix copy of the diffusers class against a newer checkpoint config.

Related errors

• `encoder_hid_dim` has to be defined when `encoder_hid_dim_ty
• `class_embed_type`: 'projection' requires `projection_class_
• addition_embed_type: {addition_embed_type} must be None, 'te
• User not found or inactive
• Missing authentication credentials

AI-assisted analysis of invoke-ai/InvokeAI@0b6a024f2f (2026-08-29). Data as JSON: /api/errors/a1bd72d48fe3f1d6. Report an issue: GitHub.