sgl-project/sglang · error · ValueError

No prompts found in file: {path}

Error message

No prompts found in file: {path}

What it means

The prompt file given to DiffusionGenerator.generate exists and was opened successfully, but contains no non-blank lines after stripping, so no prompts could be extracted. Rather than sending an empty batch, the loader raises this ValueError.

Source

Thrown at python/sglang/multimodal_gen/runtime/entrypoints/diffusion_generator.py:489

            raise RuntimeError(output_batch.error)
        if output_batch.output is None:
            raise RuntimeError("action policy returned no output")
        return output_batch.output[0]

    def _resolve_prompts(
        self,
        prompt: str | list[str] | None,
        prompt_path: str | None = None,
    ) -> list[str]:
        """Collect prompts from the argument or from a prompt file."""
        path = prompt_path or self.server_args.prompt_file_path
        if path is not None:
            if not os.path.exists(path):
                raise FileNotFoundError(f"Prompt text file not found: {path}")
            with open(path, encoding="utf-8") as f:
                prompts = [line.strip() for line in f if line.strip()]
            if not prompts:
                raise ValueError(f"No prompts found in file: {path}")
            logger.info("Found %d prompts in %s", len(prompts), path)
            return prompts

        if prompt is None:
            return [" "]
        if isinstance(prompt, str):
            return [prompt]
        return list(prompt)

    def _log_summary(self, results: list[GenerationResult]) -> None:
        if not results:
            return
        if self.server_args.warmup_mode != "off":
            total_duration_ms = results[0].metrics.get("total_duration_ms", 0)
            logger.info(
                f"Warmed-up request processed in {GREEN}%.2f{RESET} seconds (with warmup excluded)",
                total_duration_ms / 1000.0,
            )

View on GitHub (pinned to 0132848349)

Solutions

1. Open the file and confirm it has at least one non-empty line
2. Regenerate the prompt file if it was truncated by an upstream step
3. Guard your pipeline: skip generation when the prompt file is empty
4. Check the file encoding is UTF-8 text and not, e.g., binary or UTF-16 read as one line

Example fix

# before
open("prompts.txt","w").close()  # accidentally empty
generator.generate(prompt_path="prompts.txt")
# after
with open("prompts.txt","w") as f:
    f.write("a red panda\n")
generator.generate(prompt_path="prompts.txt")

def load_prompts(path):
    lines = [l.strip() for l in open(path, encoding="utf-8") if l.strip()]
    if not lines:
        raise SystemExit(f"prompt file {path} has no prompts")
    return lines

def has_prompts(path) -> bool:
    return any(l.strip() for l in open(path, encoding="utf-8"))

When it happens

Trigger: generate(prompt_path=...) where the file is empty, contains only whitespace/newlines, or only blank lines; a server launched with --prompt-file-path pointing at a truncated/placeholder file.

Common situations: Downstream step of a pipeline wrote an empty prompt file; encoding mismatch making every line unreadable; placeholder file committed by mistake; file truncated by a previous crash.

Related errors

• Prompt text file not found: {path}
• metal_shader_sources is empty; nothing to compile
• unsupported input for causal Conv3D cat/pad CUDA
• unsupported input for usp_merge_heads CUDA
• unsupported input for modulate_scale_shift CUDA

AI-assisted analysis of sgl-project/sglang@0132848349 (2026-08-28). Data as JSON: /api/errors/ab8c42e2f970350d. Report an issue: GitHub.