Stability-AI/generative-models · error · ValueError

unknown merge strategy {merge_strategy}

Error message

unknown merge strategy {merge_strategy}

What it means

The merge-strategy constructor in video_model.py (VideoResBlock path) accepts merge_strategy values including 'fixed_with_images'; anything else raises ValueError('unknown merge strategy {merge_strategy}'). The strategy determines how spatial and temporal features are blended via the mix_factor parameter.

Source

Thrown at sgm/modules/diffusionmodules/video_model.py:561

        time_embed_dim = self.in_channels * 4
        self.time_mix_time_embed = nn.Sequential(
            linear(self.in_channels, time_embed_dim),
            nn.SiLU(),
            linear(time_embed_dim, self.in_channels),
        )

        self.use_spatial_context = use_spatial_context

        if merge_strategy == "fixed":
            self.register_buffer("mix_factor", th.Tensor([merge_factor]))
        elif merge_strategy == "learned" or merge_strategy == "learned_with_images":
            self.register_parameter(
                "mix_factor", th.nn.Parameter(th.Tensor([merge_factor]))
            )
        elif merge_strategy == "fixed_with_images":
            self.mix_factor = None
        else:
            raise ValueError(f"unknown merge strategy {merge_strategy}")

        self.get_alpha_fn = functools.partial(
            get_alpha,
            merge_strategy,
            self.mix_factor,
            apply_sigmoid=apply_sigmoid_to_merge,
        )

    def forward(
        self,
        x: th.Tensor,
        context: Optional[th.Tensor] = None,
        # cam: Optional[th.Tensor] = None,
        time_context: Optional[th.Tensor] = None,
        timesteps: Optional[int] = None,
        image_only_indicator: Optional[th.Tensor] = None,
        conv_view: Optional[th.Tensor] = None,
        conv_motion: Optional[th.Tensor] = None,

View on GitHub (pinned to e8cd657656)

Solutions

1. Use a supported merge_strategy: 'fixed', 'learned', or 'fixed_with_images' (whichever set this class documents).
2. Check that the strategy string matches exactly (case and underscores) the values in the constructor.
3. Validate config with a schema or assert against the supported list before model construction.

Example fix

// before
block = VideoResBlock(..., merge_strategy="learned_with_images")
// after
block = VideoResBlock(..., merge_strategy="fixed_with_images")

VALID = {"fixed", "learned", "fixed_with_images"}
if merge_strategy not in VALID:
    raise ValueError(f"merge_strategy must be one of {VALID}, got {merge_strategy!r}")

def is_valid_video_merge_strategy(s) -> bool:
    return isinstance(s, str) and s in {"fixed", "learned", "fixed_with_images"}

try:
    block = VideoResBlock(..., merge_strategy=strategy)
except ValueError as e:
    logger.warning("%s; defaulting to 'fixed'", e)
    block = VideoResBlock(..., merge_strategy="fixed")

When it happens

Trigger: Creating a VideoResBlock (or similar) with merge_strategy not matching any known string — e.g. 'learned_with_images' when only 'fixed_with_images' plus the fixed/learned variants are supported at this site, or a typo like 'fixd'.

Common situations: Config values copied between AlphaBlender and VideoResBlock (their supported strategy sets differ), YAML typos, or overriding merge_strategy from the CLI with an invalid value.

Understand the failure class

Background: Invalid enum value errors: "Unknown type", "Invalid scope", "must be one of" — when a string is not on the library's allowed list — this error's family across 23 libraries.

Related errors

• unsupported dimensions: {dims}
• unknown merge strategy {self.merge_strategy}
• input has {x.ndim} dims but target_dims is {target_dims}, wh
• Model {model_id} not supported
• unknown discretization {params.discretization}

AI-assisted analysis of Stability-AI/generative-models@e8cd657656 (2026-08-29). Data as JSON: /api/errors/5c3f9b84a9951d89. Report an issue: GitHub.