diff --git a/litellm/router_python_core.py b/litellm/router_python_core.py index 49562ee7e35..52e9d1a96ef 100644 --- a/litellm/router_python_core.py +++ b/litellm/router_python_core.py @@ -14926,22 +14926,21 @@ class PythonCore(DeprecatedRouterMembers): ) raise ValueError("Either messages or input must be provided to count tokens") - def _longest_embedding_input_tokens(self, items: list[object], tokenizer_model: str) -> int | None: + def _longest_embedding_input_tokens(self, items: Sequence[object], tokenizer_model: str) -> int | None: """An embedding model's context window bounds each input on its own, so a batch is never summed.""" if not items: return None if all(isinstance(item, str) for item in items): count_tokens: Final = self._host.invoke("token_counter") return max(count_tokens(model=tokenizer_model, text=item) for item in items) + counting_disabled: Final = self._host.dynamic().disable_token_counter if all(isinstance(item, int) and not isinstance(item, bool) for item in items): - longest_token_ids = len(items) - else: - try: - token_id_batch: Final = _EMBEDDING_TOKEN_ID_BATCH_ADAPTER.validate_python(items, strict=True) - except ValidationError: - return None - longest_token_ids = max(len(token_ids) for token_ids in token_id_batch) - return 0 if self._host.dynamic().disable_token_counter else longest_token_ids + return 0 if counting_disabled else len(items) + try: + token_id_batch: Final = _EMBEDDING_TOKEN_ID_BATCH_ADAPTER.validate_python(items, strict=True) + except ValidationError: + return None + return 0 if counting_disabled else max(len(token_ids) for token_ids in token_id_batch) def _deployment_max_input_tokens(self, model: str, deployment: Mapping[str, object]) -> int | None: """The deployment's declared context window, or None when it declares none or cannot be resolved."""