2noise/ChatTTS · error · ValueError

best_of must be greater than or equal to n, got n={self.n} a

Error message

best_of must be greater than or equal to n, got n={self.n} and best_of={self.best_of}.

What it means

SamplingParams._verify_args requires best_of >= n. best_of is the total number of sequences to generate (the top n of which are returned), so it cannot be smaller than the number of returned sequences.

Source

Thrown at ChatTTS/model/velocity/sampling_params.py:184

        self.logits_processors = logits_processors
        self.include_stop_str_in_output = include_stop_str_in_output
        self._verify_args()
        if self.use_beam_search:
            self._verify_beam_search()
        else:
            self._verify_non_beam_search()
            # if self.temperature < _SAMPLING_EPS:
            #     # Zero temperature means greedy sampling.
            #     self.top_p = 1.0
            #     self.top_k = -1
            #     self.min_p = 0.0
            #     self._verify_greedy_sampling()

    def _verify_args(self) -> None:
        if self.n < 1:
            raise ValueError(f"n must be at least 1, got {self.n}.")
        if self.best_of < self.n:
            raise ValueError(
                f"best_of must be greater than or equal to n, "
                f"got n={self.n} and best_of={self.best_of}."
            )
        if not -2.0 <= self.presence_penalty <= 2.0:
            raise ValueError(
                "presence_penalty must be in [-2, 2], got " f"{self.presence_penalty}."
            )
        if not -2.0 <= self.frequency_penalty <= 2.0:
            raise ValueError(
                "frequency_penalty must be in [-2, 2], got "
                f"{self.frequency_penalty}."
            )
        if not 0.0 < self.repetition_penalty <= 2.0:
            raise ValueError(
                "repetition_penalty must be in (0, 2], got "
                f"{self.repetition_penalty}."
            )
        # if self.temperature < 0.0:

View on GitHub (pinned to 77b89ee281)

Solutions

Set best_of equal to or greater than n (or omit best_of — it defaults to n)
If you only want n sequences returned, drop best_of entirely

Example fix

# before
params = SamplingParams(n=4, best_of=2)

# after
params = SamplingParams(n=4)  # best_of defaults to n

Defensive patterns

Strategy: validation

Validate before calling

def normalize(best_of, n):
    return max(best_of, n) if best_of is not None else n

Prevention

Omit best_of unless you specifically need extra candidates
Assert best_of >= n before constructing params

When it happens

Trigger: Constructing SamplingParams with best_of less than n, e.g. SamplingParams(n=4, best_of=2). Note best_of defaults to n, so this only fires when best_of is set explicitly below n.

Common situations: Copying OpenAI-style params where best_of/n semantics differ; tuning configs and shrinking best_of while forgetting n; swapping the two arguments.

Related errors

AI-assisted analysis of 2noise/ChatTTS@77b89ee281 (2026-08-26). Data as JSON: /api/errors/770bf8496f4aa44a. Report an issue: GitHub.