huggingface/smolagents · error · ValueError

This tool does not implement a default checkpoint, you need

Error message

This tool does not implement a default checkpoint, you need to pass one.

What it means

A PipelineTool needs a model checkpoint. If `model` is not passed and the subclass does not define `default_checkpoint`, __init__ cannot determine what to load and raises ValueError immediately.

Source

Thrown at src/smolagents/tools.py:1237

    def __init__(
        self,
        model=None,
        pre_processor=None,
        post_processor=None,
        device=None,
        device_map=None,
        model_kwargs=None,
        token=None,
        **hub_kwargs,
    ):
        if not _is_package_available("accelerate") or not _is_package_available("torch"):
            raise ModuleNotFoundError(
                "Please install 'transformers' extra to use a PipelineTool: `pip install 'smolagents[transformers]'`"
            )

        if model is None:
            if self.default_checkpoint is None:
                raise ValueError("This tool does not implement a default checkpoint, you need to pass one.")
            model = self.default_checkpoint
        if pre_processor is None:
            pre_processor = model

        self.model = model
        self.pre_processor = pre_processor
        self.post_processor = post_processor
        self.device = device
        self.device_map = device_map
        self.model_kwargs = {} if model_kwargs is None else model_kwargs
        if device_map is not None:
            self.model_kwargs["device_map"] = device_map
        self.hub_kwargs = hub_kwargs
        self.hub_kwargs["token"] = token

        super().__init__()

    def setup(self):

View on GitHub (pinned to 30bb116109)

Solutions

1. Pass an explicit checkpoint: `MyTool(model='repo/model-name')`
2. Define `default_checkpoint = 'org/model-name'` as a class attribute on the subclass
3. If the tool is intentionally checkpoint-less, reconsider subclassing PipelineTool vs plain Tool

Example fix

# before
class MyTranslator(PipelineTool):
    pass  # no default_checkpoint
MyTranslator()
# after
class MyTranslator(PipelineTool):
    default_checkpoint = "Helsinki-NLP/opus-mt-en-fr"
MyTranslator()

if getattr(MyPipelineTool, "default_checkpoint", None) is None and checkpoint is None:
    checkpoint = "org/default-model"  # your fallback
my_tool = MyPipelineTool(model=checkpoint)

def has_checkpoint(tool_cls, model=None) -> bool:
    return model is not None or getattr(tool_cls, "default_checkpoint", None) is not None

try:
    tool = MyPipelineTool()
except ValueError as e:
    if "default checkpoint" in str(e):
        tool = MyPipelineTool(model="org/default-model")
    else:
        raise

When it happens

Trigger: Creating a PipelineTool subclass whose class body sets `default_checkpoint = None` (or never defines it) without passing `model=...`, e.g. `MyPipeTool()` with no arguments.

Common situations: Writing custom PipelineTool subclasses and forgetting the default_checkpoint class attribute; refactoring a subclass and deleting the default; expecting the base class to auto-select a checkpoint.

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

• Only 'markdown' is supported for a string argument to `code_
• Unsupported executor type: {self.executor_type}

AI-assisted analysis of huggingface/smolagents@30bb116109 (2026-08-28). Data as JSON: /api/errors/e0b5e0287598574d. Report an issue: GitHub.