sgl-project/sglang · error · ValueError

Missing tensor payload for module(s)

Error message

Missing tensor payload for module(s): {missing}. Provided modules: {list(named_tensors.keys())}

What it means

Raised by WeightsUpdater._resolve_module_payloads when named_tensors is a dict but one or more modules selected for update has no corresponding key in the dict. Each module being updated must have an entry in the payload mapping.

Solutions

1. Add the missing module's tensor payload under exactly the name listed in 'missing'
2. Or restrict target_modules to only the modules you actually have payloads for

Example fix

// before
updater.update_weights_from_tensor(named_tensors={"encoder": enc_tensors}, target_modules=["encoder","decoder"])
// after
updater.update_weights_from_tensor(named_tensors={"encoder": enc_tensors, "decoder": dec_tensors}, target_modules=["encoder","decoder"])

missing = [m for m in target_modules if m not in named_tensors]
if missing: raise KeyError(f"payload missing {missing}")

def has_all_payloads(named_tensors: dict, modules: list[str]) -> bool:
    return isinstance(named_tensors, dict) and all(m in named_tensors for m in modules)

When it happens

Trigger: Calling update_weights_from_tensor with named_tensors={'encoder': ...} while the updater resolves two modules (e.g. target_modules=['encoder','decoder']) — 'decoder' is missing so the error fires.

Common situations: Adding a new module to target_modules without extending the payload dict; key typos; partial checkpoint serialization that skipped one module's tensors.

Related errors

• Ambiguous tensor payload for multi-module update. Provide a…
• flattened_bucket payload must be a dict with…
• Module(s) requested for update not found in pipeline
• Unsupported module payload type for load_format=
• batching config rule requires max_batch_size

AI-assisted analysis of sgl-project/sglang@0132848349 (2026-08-28). Data as JSON: /api/errors/1785a962b1e96019. Report an issue: GitHub.

Appendix: source

Thrown at python/sglang/multimodal_gen/runtime/post_training/weights_updater.py:750

            )

        message = (
            f"Updated {updated} LoRA layers in {target_module} from IPC tensors "
            f"(skipped {skipped} unknown layers)."
        )
        logger.info(message)
        return True, message

    def _resolve_module_payloads(
        self,
        named_tensors: Any,
        modules_to_update: list[tuple[str, torch.nn.Module]],
    ) -> dict[str, Any]:
        module_names = [name for name, _ in modules_to_update]
        if isinstance(named_tensors, dict):
            missing = [name for name in module_names if name not in named_tensors]
            if missing:
                raise ValueError(
                    f"Missing tensor payload for module(s): {missing}. "
                    f"Provided modules: {list(named_tensors.keys())}"
                )
            return {name: named_tensors[name] for name in module_names}

        if len(module_names) == 1:
            return {module_names[0]: named_tensors}

        raise ValueError(
            "Ambiguous tensor payload for multi-module update. "
            "Provide a dict mapping module_name -> module payload, "
            f"requested modules: {module_names}."
        )

    def _materialize_weights_iter(self, module_payload: Any, load_format: str | None):
        if load_format == "flattened_bucket":
            if not isinstance(module_payload, dict):
                raise ValueError(

View on GitHub (pinned to 0132848349)