sgl-project/sglang · error · ValueError

Error while unloading LRU LoRA adapter '{lru_lora_name}': {u

Error message

Error while unloading LRU LoRA adapter '{lru_lora_name}': {unload_result.error_message}

What it means

During LRU eviction inside load_lora_adapter, the attempt to unload the chosen adapter failed and returned an error message, so the load aborts.

Source

Thrown at python/sglang/srt/managers/tokenizer_control_mixin.py:669

                            exclude_pinned=True
                        )
                        if lru_lora_name is None:
                            raise ValueError(
                                "Didn't find any LoRA adapters when trying to evict LRU LoRA adapter. "
                                f"LoRA registry is: {self.lora_registry._registry}"
                            )

                        logger.info(
                            f"Unloading least recently used LoRA adapter '{lru_lora_name}' "
                            f"(current number of adapters: {self.lora_registry.num_registered_loras}, "
                            f"max allowed: {self.server_args.max_loaded_loras})"
                        )

                        unload_result = await self._unload_lora_adapter_locked(
                            UnloadLoRAAdapterReqInput(lora_name=lru_lora_name)
                        )
                        if not unload_result.success:
                            raise ValueError(
                                f"Error while unloading LRU LoRA adapter '{lru_lora_name}': "
                                f"{unload_result.error_message}"
                            )
                        del result.loaded_adapters[lru_lora_name]

                return result
        except ValueError as e:
            return LoadLoRAAdapterReqOutput(
                success=False,
                error_message=str(e),
            )

    async def load_lora_adapter_from_tensors(
        self: TokenizerManager,
        obj: LoadLoRAAdapterFromTensorsReqInput,
        _: Optional[fastapi.Request] = None,
    ) -> LoadLoRAAdapterFromTensorsReqOutput:
        self.auto_create_handle_loop()

View on GitHub (pinned to 0132848349)

Solutions

1. Retry the load after in-flight requests using that adapter finish
2. Inspect server logs for the underlying unload error message included in the exception
3. Unoad the adapter manually via /unload_lora_adapter and inspect the error

for attempt in range(3):
    try:
        return await client.load_lora_adapter(req)
    except ValueError as e:
        if 'unloading LRU' in str(e) and attempt < 2:
            await asyncio.sleep(2); continue
        raise

When it happens

Trigger: _unload_lora_adapter_locked for the LRU adapter returns success=False (scheduler-side unload failure), e.g. adapter in use by in-flight requests or backend unload error.

Common situations: High concurrency with adapters hot in flight; LoRA backend errors during weight unloading.

Related errors

• LoRA with name {lora_name} does not exist. Loaded LoRAs: {se
• Didn't find any LoRA adapters when trying to evict LRU LoRA
• Got LoRA adapter that has never been loaded: {lora_path}\nAl
• LoRA batch_info must provide max_len or seg_lens.
• LoRA batch_info must provide max_len or seg_lens.

AI-assisted analysis of sgl-project/sglang@0132848349 (2026-08-28). Data as JSON: /api/errors/989d0b51ccd197db. Report an issue: GitHub.