unslothai/unsloth · error · RuntimeError

apply_chat_template_for_generation: no attempt produced a re

Error message

apply_chat_template_for_generation: no attempt produced a result

What it means

Raised by apply_chat_template_for_generation when its loop of candidate kwarg attempts (tools on/off, continue_final_message variants, etc.) produced no result AND no exception was recorded — every attempt was skipped via `continue` on TypeError without any call succeeding. It is a defensive invariant: the attempt list was empty or all attempts failed with TypeError.

Source

Thrown at studio/backend/core/inference/chat_template_helpers.py:2555

        boundary = _boundary_kwargs if boundary is None else boundary
        last_exc: Optional[Exception] = None
        for kwargs in attempts:
            try:
                return tokenizer.apply_chat_template(
                    _swept_for(kwargs, msgs),
                    tokenize = False,
                    **boundary,
                    **kwargs,
                )
            except TypeError as e:
                last_exc = e
                continue
            except Exception as e:
                last_exc = e
                break
        if last_exc is not None:
            raise last_exc
        raise RuntimeError("apply_chat_template_for_generation: no attempt produced a result")

    def _render_continuation_manually(msgs: list) -> str:
        """For tokenizers predating ``continue_final_message`` (TypeError above).

        Prefix and partial come from the SAME swept copy: an attempt that drops the tools
        kwarg re-sweeps for the default template, whose markup would otherwise survive raw.
        """
        for kwargs in attempts:
            swept = _swept_for(kwargs, msgs)
            try:
                prefix = tokenizer.apply_chat_template(
                    swept[:-1], tokenize = False, add_generation_prompt = True, **kwargs
                )
            except TypeError:
                continue
            partial = trailing_assistant_text(swept) or _continue_text
            return f"{strip_open_reasoning_prefill(prefix)}{partial}"
        raise TypeError("no attempt rendered the continuation prefix")

View on GitHub (pinned to 203007d190)

Solutions

Check the transformers version in use; upgrade/downgrade to a version whose apply_chat_template accepts add_generation_prompt and the swept kwargs
Log each attempt's kwargs and the TypeError to see which signature mismatch repeats, then add a compatible attempt
As a caller, fall back to rendering the prompt manually (join message contents with the tokenizer's known template) when this RuntimeError fires

Defensive patterns

Strategy: try-catch

Try / catch

try:
    prompt = apply_chat_template_for_generation(tokenizer, messages)
except RuntimeError as e:
    if "no attempt produced a result" in str(e):
        prompt = manual_prompt_render(messages)  # known-format fallback

Prevention

Pin a transformers version compatible with the models in use
Smoke-test apply_chat_template once at startup, not per request
Log each attempt's kwargs to extend the attempt list for new tokenizer signatures

When it happens

Trigger: A tokenizer whose apply_chat_template signature is incompatible with every attempt (very old transformers); an attempts list built empty because the swept message/kwarg combinations deduplicated to nothing; TypeError raised on the first attempt short-circuiting the loop via `break` with last_exc set is re-raised instead — this specific message means the no-exception no-result branch.

Common situations: transformers version far older/newer than the supported range so all kwarg combinations raise TypeError; a tokenizer class overriding apply_chat_template with an incompatible signature; none of the model card's template variants being applicable.

Related errors

AI-assisted analysis of unslothai/unsloth@203007d190 (2026-08-15). Data as JSON: /api/errors/ccb77fd9bd5d7e17. Report an issue: GitHub.