[Bug]: AssertionError: Do not capture num_reqs > max_num_reqs for uniform batch #25494
Description
Your current environment
The output of python collect_env.py
Your output of `python collect_env.py` here
🐛 Describe the bug
vllm serve --model="deepseek-ai/DeepSeek-R1" --max-num-seqs 512 --data-parallel-size 8 --enable-expert-parallel --gpu-memory-utilization 0.9 --port 9256
vllm bench serve --model deepseek-ai/DeepSeek-R1 --dataset-name random --host 127.0.0.1 --port 9256 --random-input-len 130000 --random-output-len 1 --request-rate inf --num-prompts 1
Will meet an error:
(EngineCore_DP7 pid=3747941) File "/usr/lib/python3.12/multiprocessing/process.py", line 314, in _bootstrap
(EngineCore_DP7 pid=3747941) self.run()
(EngineCore_DP7 pid=3747941) File "/usr/lib/python3.12/multiprocessing/process.py", line 108, in run
(EngineCore_DP7 pid=3747941) self._target(*self._args, **self._kwargs)
(EngineCore_DP7 pid=3747941) File "/home/wentao/vllm-source/vllm/v1/engine/core.py", line 712, in run_engine_core
(EngineCore_DP7 pid=3747941) raise e
(EngineCore_DP7 pid=3747941) File "/home/wentao/vllm-source/vllm/v1/engine/core.py", line 701, in run_engine_core
(EngineCore_DP7 pid=3747941) engine_core.run_busy_loop()
(EngineCore_DP7 pid=3747941) File "/home/wentao/vllm-source/vllm/v1/engine/core.py", line 1056, in run_busy_loop
(EngineCore_DP7 pid=3747941) self.execute_dummy_batch()
(EngineCore_DP7 pid=3747941) File "/home/wentao/vllm-source/vllm/v1/engine/core.py", line 387, in execute_dummy_batch
(EngineCore_DP7 pid=3747941) self.model_executor.execute_dummy_batch()
(EngineCore_DP7 pid=3747941) File "/home/wentao/vllm-source/vllm/v1/executor/abstract.py", line 109, in execute_dummy_batch
(EngineCore_DP7 pid=3747941) self.collective_rpc("execute_dummy_batch")
(EngineCore_DP7 pid=3747941) File "/home/wentao/vllm-source/vllm/executor/uniproc_executor.py", line 83, in collective_rpc
(EngineCore_DP7 pid=3747941) return [run_method(self.driver_worker, method, args, kwargs)]
(EngineCore_DP7 pid=3747941) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(EngineCore_DP7 pid=3747941) File "/home/wentao/vllm-source/vllm/utils/__init__.py", line 3010, in run_method
(EngineCore_DP7 pid=3747941) return func(*args, **kwargs)
(EngineCore_DP7 pid=3747941) ^^^^^^^^^^^^^^^^^^^^^
(EngineCore_DP7 pid=3747941) File "/home/wentao/vllm-source/vllm/v1/worker/gpu_worker.py", line 490, in execute_dummy_batch
(EngineCore_DP7 pid=3747941) self.model_runner._dummy_run(1, uniform_decode=True)
(EngineCore_DP7 pid=3747941) File "/home/wentao/.venv/lib/python3.12/site-packages/torch/utils/_contextlib.py", line 116, in decorate_context
(EngineCore_DP7 pid=3747941) return func(*args, **kwargs)
(EngineCore_DP7 pid=3747941) ^^^^^^^^^^^^^^^^^^^^^
(EngineCore_DP7 pid=3747941) File "/home/wentao/vllm-source/vllm/v1/worker/gpu_model_runner.py", line 2918, in _dummy_run
(EngineCore_DP7 pid=3747941) assert num_reqs <= max_num_reqs, \
(EngineCore_DP7 pid=3747941) ^^^^^^^^^^^^^^^^^^^^^^^^
(EngineCore_DP7 pid=3747941) AssertionError: Do not capture num_reqs > max_num_reqs for uniform batchBefore submitting a new issue...
- Make sure you already searched for relevant issues, and asked the chatbot living at the bottom right corner of the documentation page, which can answer lots of frequently asked questions.