vllm-project/vllm · error · ValueError

enable_expert_parallel must be True to use EPLB.

Error message

enable_expert_parallel must be True to use EPLB.

What it means

EPLB rebalances experts across the expert-parallel (EP) group, so it presupposes expert parallelism. ParallelConfig raises when enable_eplb=True while enable_expert_parallel=False, since there is no EP group to balance over.

Source

Thrown at vllm/config/parallel.py:500

            raise ValueError(
                "data_parallel_external_lb can only be set when data_parallel_size > 1"
            )

        if not self.numa_bind and (
            self.numa_bind_nodes is not None or self.numa_bind_cpus is not None
        ):
            raise ValueError(
                "numa_bind_nodes and numa_bind_cpus require numa_bind=True."
            )

        if self.enable_eplb:
            if not current_platform.is_cuda_alike():
                raise ValueError(
                    "Expert parallelism load balancing is only supported on "
                    "CUDA devices or ROCm devices now."
                )
            if not self.enable_expert_parallel:
                raise ValueError("enable_expert_parallel must be True to use EPLB.")
            # The EP group spans the TP x PCP x DP ranks. EPLB therefore needs
            # TP, PCP, or DP > 1.
            if (
                self.tensor_parallel_size
                * self.prefill_context_parallel_size
                * self.data_parallel_size
                <= 1
            ):
                raise ValueError(
                    "EPLB requires tensor, prefill-context, or data parallelism, "
                    f"but got TP={self.tensor_parallel_size}, "
                    f"PCP={self.prefill_context_parallel_size}, "
                    f"DP={self.data_parallel_size}."
                )
        else:
            if self.eplb_config.num_redundant_experts != 0:
                raise ValueError(
                    "num_redundant_experts is set to "

View on GitHub (pinned to c794754062)

Solutions

1. Add --enable-expert-parallel together with --enable-eplb.
2. If expert parallelism is not wanted, remove --enable-eplb and any eplb_config.

Example fix

# before
vllm serve Mixtral-8x7B --enable-eplb
# after
vllm serve Mixtral-8x7B --enable-expert-parallel --enable-eplb

def eplb_flags_valid(enable_eplb: bool, enable_expert_parallel: bool) -> bool:
    return not enable_eplb or enable_expert_parallel

assert eplb_flags_valid(True, True)

When it happens

Trigger: Launching a MoE model with --enable-eplb but without --enable-expert-parallel.

Common situations: Assuming EPLB turns on expert parallelism implicitly; enabling EPLB for its redundant-expert feature while forgetting the base EP flag in a long CLI invocation.

Related errors

• EPLB requires tensor, prefill-context, or data parallelism,
• Expert parallelism load balancing is only supported on CUDA
• num_redundant_experts is set to {self.eplb_config.num_redund
• Elastic EP is only supported with enable_eplb=True.
• multimodal preprocessing error: {0}

AI-assisted analysis of vllm-project/vllm@c794754062 (2026-08-14). Data as JSON: /api/errors/42e377012128a0d6. Report an issue: GitHub.