sgl-project/sglang · error · ValueError

server_args is required to resolve Ideogram4 NVFP4 paths

Error message

server_args is required to resolve Ideogram4 NVFP4 paths

What it means

Raised by Ideogram4Pipeline._get_model_resolution when the NVFP4 model-path resolution has not been cached yet and the caller passes server_args=None. The resolution needs ServerArgs (quantization overrides, model paths) to locate the Ideogram4 NVFP4 weights, so it cannot run lazily without them.

Source

Thrown at python/sglang/multimodal_gen/runtime/pipelines/ideogram.py:199

        )
        self.add_stage_factory(
            RoleType.DECODER,
            lambda: Ideogram4DecodingStage(vae=self.get_module("vae")),
            "ideogram4_decoding_stage",
        )


class Ideogram4Nvfp4Pipeline(Ideogram4Pipeline):
    pipeline_name = "Ideogram4Nvfp4Pipeline"
    _model_resolution: Ideogram4Nvfp4ModelResolution | None = None

    def _get_model_resolution(
        self,
        server_args: ServerArgs | None = None,
    ) -> Ideogram4Nvfp4ModelResolution:
        if self._model_resolution is None:
            if server_args is None:
                raise ValueError(
                    "server_args is required to resolve Ideogram4 NVFP4 paths"
                )
            self._model_resolution = resolve_ideogram4_nvfp4_model(
                server_args,
                self.model_path,
            )
        return self._model_resolution

    def _load_config(self) -> dict[str, Any]:
        model_resolution = self._get_model_resolution(self.server_args)
        logger.info("Model path: %s", self.model_path)
        logger.info(
            "Using base model '%s' at %s for config and non-transformer components",
            model_resolution.base_model_name,
            model_resolution.base_model_path,
        )
        config = verify_model_config_and_directory(model_resolution.base_model_path)
        return cast(dict[str, Any], config)

View on GitHub (pinned to 0132848349)

Solutions

1. Pass the real ServerArgs through to _get_model_resolution(server_args=server_args) on the first invocation so the resolution is cached
2. Fix the caller chain (_load_config / _resolve_component_path / load_modules overrides) to forward server_args instead of None
3. In tests, call _get_model_resolution(server_args=build_server_args()) once up front to prime _model_resolution

Example fix

// before
resolution = pipeline._get_model_resolution()  # cache empty -> ValueError

// after
resolution = pipeline._get_model_resolution(server_args=server_args)

if pipeline._model_resolution is None:
    if server_args is None:
        raise SystemExit('ServerArgs required before resolving Ideogram4 NVFP4 paths')
    pipeline._get_model_resolution(server_args=server_args)

When it happens

Trigger: Calling _get_model_resolution() with no argument (or None) before any prior call that passed server_args; calling helpers that depend on it — _load_config, _resolve_component_path, load_modules — on a pipeline instance whose _model_resolution cache is empty and whose code path forwards server_args=None.

Common situations: Custom loaders or tests that construct the pipeline and call config-loading helpers without threading ServerArgs through; refactors that dropped the server_args parameter from an internal call chain.

Understand the failure class

Background: "missing required argument" and "the following required arguments were not provided": what required-argument errors mean and how to fix them — this error's family across 20 libraries.

Related errors

• Usage: sglang serve --model-path <model-name-or-path> [addit
• Error: --model-path is required. Please provide the path to
• v_cache must be provided
• q must be provided unless qv is provided with only_qv=True
• MiniMax-H3 NVFP4 metadata requires checkpoint files

AI-assisted analysis of sgl-project/sglang@0132848349 (2026-08-28). Data as JSON: /api/errors/6f42867d9410d662. Report an issue: GitHub.