huggingface/smolagents · warning · FutureWarning

The 'model_id' parameter will be required in version 2.0.0.

Error message

The 'model_id' parameter will be required in version 2.0.0. Please update your code to pass this parameter to avoid future errors. For now, it defaults to 'HuggingFaceTB/SmolLM2-1.7B-Instruct'.

What it means

TransformersModel.__init__ warns (FutureWarning) when model_id is not passed; it currently falls back to 'HuggingFaceTB/SmolLM2-1.7B-Instruct' but model_id becomes required in version 2.0.0. The warning is the migration signal before the hard failure.

Source

Thrown at src/smolagents/models.py:933

        apply_chat_template_kwargs: dict[str, Any] | None = None,
        **kwargs,
    ):
        try:
            import torch
            from transformers import (
                AutoModelForCausalLM,
                AutoModelForImageTextToText,
                AutoProcessor,
                AutoTokenizer,
                TextIteratorStreamer,
            )
        except ModuleNotFoundError:
            raise ModuleNotFoundError(
                "Please install 'transformers' extra to use 'TransformersModel': `pip install 'smolagents[transformers]'`"
            )

        if not model_id:
            warnings.warn(
                "The 'model_id' parameter will be required in version 2.0.0. "
                "Please update your code to pass this parameter to avoid future errors. "
                "For now, it defaults to 'HuggingFaceTB/SmolLM2-1.7B-Instruct'.",
                FutureWarning,
            )
            model_id = "HuggingFaceTB/SmolLM2-1.7B-Instruct"

        max_new_tokens = max_tokens if max_tokens is not None else max_new_tokens

        if device_map is None:
            device_map = "cuda" if torch.cuda.is_available() else "cpu"
        logger.info(f"Using device: {device_map}")
        self._is_vlm = False
        self.model_kwargs = model_kwargs or {}
        self.apply_chat_template_kwargs = apply_chat_template_kwargs or {}
        try:
            self.model = AutoModelForImageTextToText.from_pretrained(
                model_id,

View on GitHub (pinned to 30bb116109)

Solutions

Pass an explicit model_id matching your use case
If the default was intentional, pass it explicitly to silence the warning
Audit configs/scripts for TransformersModel( constructions without model_id before upgrading to 2.x

Example fix

# before
model = TransformersModel()

# after
model = TransformersModel(model_id='Qwen/Qwen2.5-7B-Instruct')

Defensive patterns

Strategy: validation

Validate before calling

model = TransformersModel(model_id=os.environ['MODEL_ID'])

Try / catch

import warnings
with warnings.catch_warnings():
    warnings.simplefilter('ignore', FutureWarning)
    model = TransformersModel()  # temporary, will break in 2.0

Prevention

Always pass model_id explicitly
Grep code for TransformersModel( without model_id before upgrading
Store model ids in config/env, not implicit defaults

When it happens

Trigger: Constructing TransformersModel() with no model_id (relying on the old default) after upgrading smolagents; copying legacy examples that omit model_id.

Common situations: Version upgrades where code silently used the default SmolLM2 model; teams surprised their 'unconfigured' model was actually a small default model.

Related errors

AI-assisted analysis of huggingface/smolagents@30bb116109 (2026-08-28). Data as JSON: /api/errors/8a54e48894219faf. Report an issue: GitHub.