sgl-project/sglang · error · ValueError

Failed to find config.json for {model_name_or_path} after fa

Error message

Failed to find config.json for {model_name_or_path} after failing to find model_index.jsonYou might be looking for models ending with '-Diffusers'

What it means

When a repo has no model_index.json, maybe_download_model_index falls back to treating it as a single model and looks for config.json in the downloaded snapshot. If that's also missing, it raises ValueError with a hint that the user may want the '-Diffusers' variant of the model.

Source

Thrown at python/sglang/multimodal_gen/runtime/utils/hf_diffusers_utils.py:849

        # Add the pipeline name for downstream use
        config["pipeline_name"] = config["_class_name"]

        logger.debug(
            "Resolved model_index.json for %s, pipeline: %s",
            model_name_or_path,
            config["_class_name"],
        )
        return config
    except EntryNotFoundError:
        logger.debug(
            "model_index.json not found for %s. Assuming it is a single model and downloading it.",
            model_name_or_path,
        )
        local_path = maybe_download_model(model_name_or_path)
        config_path = os.path.join(local_path, "config.json")
        if not os.path.exists(config_path):
            raise ValueError(
                f"Failed to find config.json for {model_name_or_path} after failing to find model_index.json"
                f"You might be looking for models ending with '-Diffusers'"
            )
        with open(config_path) as f:
            config = json.load(f)
        return config
    except Exception as e:
        raise ValueError(
            f"Failed to download or parse model_index.json for {model_name_or_path}: {e}"
        ) from e


def maybe_download_model(
    model_name_or_path: str,
    local_dir: str | None = None,
    download: bool = True,
    is_lora: bool = False,
    allow_patterns: list[str] | None = None,

View on GitHub (pinned to 0132848349)

Solutions

1. Switch to the '-Diffusers' suffixed repo for that model
2. Verify the repo ID spelling and that the repo is public / you are authenticated (huggingface-cli login)
3. For private/gated models, set HF_TOKEN

Example fix

# before
--model stabilityai/stable-diffusion-3.5-large
# after
--model stabilityai/stable-diffusion-3.5-large-diffusers

import os
assert os.path.exists(f'{p}/model_index.json') or os.path.exists(f'{p}/config.json'), 'neither pipeline nor single model'

try:
    info = get_model_info(name)
except ValueError:
    info = get_model_info(name + '-Diffusers')

When it happens

Trigger: Calling get_model_info / _get_config_info on a repo that is neither a diffusers pipeline (no model_index.json) nor a single HF model (no config.json) — e.g. the original non-diffusers checkpoint repo.

Common situations: Passing the base checkpoint repo (e.g. 'black-forest-labs/FLUX.1-dev' vs '...-Diffusers'); empty or private-without-auth repos; typo'd repo IDs.

Related errors

• Z-Image transformer has no `rotary_emb`. It likely loaded vi
• Model path '{model_path}' is already registered for pipeline
• --diffusers-kwargs must be valid JSON. Got: {args.diffusers_
• Model config does not contain a _class_name attribute. Only
• Model config does not contain a _class_name attribute. Only

AI-assisted analysis of sgl-project/sglang@0132848349 (2026-08-28). Data as JSON: /api/errors/8b8b3c4929944fb7. Report an issue: GitHub.