sgl-project/sglang · error · ValueError

{template_error}{suffix}

Error message

{template_error}{suffix}

What it means

While rendering an embedding input through the Jinja chat template, a jinja2.TemplateError (e.g. TemplateSyntaxError, TemplateNotFound) occurred. The embedding serving layer rethrows it as ValueError with the template name and line number so clients get a 400 rather than a 500.

Source

Thrown at python/sglang/srt/entrypoints/openai/serving_embedding.py:239

                template_content_format,
                image_data=[],
                video_data=[],
                audio_data=[],
                modalities=[],
            )
            try:
                prompt = self.tokenizer_manager.tokenizer.apply_chat_template(
                    [processed_msg],
                    tokenize=False,
                    add_generation_prompt=True,
                )
            except jinja2.TemplateError as template_error:
                location = getattr(template_error, "lineno", None)
                name = getattr(template_error, "name", None)
                suffix = ""
                if name or location:
                    suffix = f" (template={name or '<unknown>'}, line={location})"
                raise ValueError(f"{template_error}{suffix}") from template_error
            except (TypeError, KeyError, AttributeError) as template_error:
                raise ValueError(
                    f"Failed to render chat template for embedding input: {template_error}"
                ) from template_error
            prompts.append(prompt)

        return prompts

    async def _handle_non_streaming_request(
        self,
        adapted_request: EmbeddingReqInput,
        request: EmbeddingRequest,
        raw_request: Request,
    ) -> Union[EmbeddingResponse, ErrorResponse, ORJSONResponse]:
        """Handle the embedding request"""
        try:
            ret = await self.tokenizer_manager.generate_request(
                adapted_request, raw_request

View on GitHub (pinned to 0132848349)

Solutions

1. Check the reported (template=..., line=...) in the message and fix the template syntax at that line
2. Validate the template renders standalone with jinja2 before passing it to the server
3. Use the tokenizer's built-in chat template for embedding models like Qwen3-Embedding

from jinja2 import Environment; Environment().parse(open(template_path).read())  # syntax pre-check

except ValueError as e: report template name/line from message back to user

When it happens

Trigger: POST /v1/embeddings with chat-template input (e.g. messages or a --chat-template for embedding models) where the template itself is malformed or references an undefined construct that raises TemplateError.

Common situations: Supplying a broken custom --chat-template for an embedding/reranker model; template syntax errors introduced when hand-editing; template referencing variables not present in embedding requests.

Related errors

• Failed to render chat template for embedding input: {templat
• {template_error}
• Invalid content format: {content_format}
• Invalid thinking_mode `{thinking_mode}`
• Unknown role: {role}

AI-assisted analysis of sgl-project/sglang@0132848349 (2026-08-28). Data as JSON: /api/errors/4d64dbfd47329976. Report an issue: GitHub.