vllm version error about Ascend NPU

[xinyubai1209]

System Info

i try to run vllm+npu training script: examples/sglang_multiturn/run_qwen2.5-3b_gsm8k_multiturn_vllm_fsdp.sh

# run on Ascend 910
# make sure your current working directory is the root of the project

set -x
ulimit -n 65535

#set vllm v1 env
export VLLM_USE_V1=1

PROJECT_DIR="$(pwd)"
CONFIG_PATH="$PROJECT_DIR/examples/sglang_multiturn/config"

TRAIN_BATCH_SIZE=32
MICRO_BATCH_SIZE=8

python3 -m verl.trainer.main_ppo \
    --config-path="$CONFIG_PATH" \
    --config-name='gsm8k_multiturn_grpo' \
    actor_rollout_ref.rollout.name=vllm \
    algorithm.adv_estimator=grpo \
    data.train_batch_size=${TRAIN_BATCH_SIZE} \
    data.max_prompt_length=1024 \
    data.max_response_length=1024 \
    data.filter_overlong_prompts=True \
    data.truncation='error' \
    data.return_raw_chat=True \
    actor_rollout_ref.model.path="/home/Qwen2.5-3B-Instruct" \
    actor_rollout_ref.actor.optim.lr=1e-6 \
    actor_rollout_ref.actor.ppo_mini_batch_size=${TRAIN_BATCH_SIZE} \
    actor_rollout_ref.actor.ppo_micro_batch_size_per_gpu=${MICRO_BATCH_SIZE} \
    actor_rollout_ref.actor.use_kl_loss=True \
    actor_rollout_ref.actor.kl_loss_coef=0.001 \
    actor_rollout_ref.actor.kl_loss_type=low_var_kl \
    actor_rollout_ref.actor.entropy_coeff=0 \
    actor_rollout_ref.rollout.log_prob_micro_batch_size_per_gpu=${MICRO_BATCH_SIZE} \
    actor_rollout_ref.rollout.tensor_model_parallel_size=2 \
    actor_rollout_ref.rollout.n=8 \
    actor_rollout_ref.rollout.gpu_memory_utilization=0.9\
    actor_rollout_ref.ref.log_prob_micro_batch_size_per_gpu=${MICRO_BATCH_SIZE} \
    algorithm.use_kl_in_reward=False \
    trainer.critic_warmup=0 \
    trainer.project_name='gsm8k_async_rl' \
    trainer.experiment_name='qwen2.5-3b_function_rm-gsm8k-sgl-multi-w-tool-verify-n16' \
    trainer.device=npu \
    trainer.n_gpus_per_node=4 \
    trainer.nnodes=1 \
    trainer.save_freq=-1 \
    trainer.test_freq=20 \
    trainer.logger='["console"]' \
    data.train_files=/home/data/gsm8k/train.parquet \
    data.val_files=/home/data/gsm8k/test.parquet \
    trainer.total_epochs=15 \
    actor_rollout_ref.rollout.update_weights_bucket_megabytes=512 \
    actor_rollout_ref.rollout.trace.token2text=False \
    actor_rollout_ref.rollout.mode=async \
    actor_rollout_ref.rollout.multi_turn.enable=true \
    actor_rollout_ref.rollout.enforce_eager=True \
    actor_rollout_ref.actor.use_torch_compile=False \
    actor_rollout_ref.rollout.multi_turn.tool_config_path="$PROJECT_DIR/examples/sglang_multiturn/config/tool_config/gsm8k_tool_config.yaml" \
    actor_rollout_ref.rollout.free_cache_engine=True

Information

• [x] The official example scripts
• [ ] My own modified scripts

Tasks

• [x] An officially supported task in the examples folder (such as GLUE/SQuAD, ...)
• [ ] My own task or dataset (give details below)

Reproduction

the error information is:

Traceback (most recent call last):
  File "/verl/verl/trainer/main_ppo.py", line 42, in main
    run_ppo(config)
  File "/verl/verl/trainer/main_ppo.py", line 96, in run_ppo
    ray.get(runner.run.remote(config))
  File "/usr/local/python3.11.13/lib/python3.11/site-packages/ray/_private/auto_init_hook.py", line 21, in auto_init_wrapper
    return fn(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^
  File "/usr/local/python3.11.13/lib/python3.11/site-packages/ray/_private/client_mode_hook.py", line 103, in wrapper
    return func(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/python3.11.13/lib/python3.11/site-packages/ray/_private/worker.py", line 2822, in get
    values, debugger_breakpoint = worker.get_objects(object_refs, timeout=timeout)
                                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/python3.11.13/lib/python3.11/site-packages/ray/_private/worker.py", line 930, in get_objects
    raise value.as_instanceof_cause()
ray.exceptions.RayTaskError(AttributeError): ray::TaskRunner.run() (pid=312910, ip=10.1.68.108, actor_id=1d2b384df857f2b3e9ff165001000000, repr=<main_ppo.TaskRunner object at 0xfffc01112cd0>)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/verl/verl/trainer/main_ppo.py", line 339, in run
    trainer.init_workers()
  File "/verl/verl/trainer/ppo/ray_trainer.py", line 774, in init_workers
    self.async_rollout_manager = AgentLoopManager(
                                 ^^^^^^^^^^^^^^^^^
  File "/verl/verl/experimental/agent_loop/agent_loop.py", line 688, in __init__
    self.sleep()
  File "/verl/verl/experimental/agent_loop/agent_loop.py", line 801, in sleep
    self._run_all([replica.sleep() for replica in self.rollout_replicas])
  File "/verl/verl/experimental/agent_loop/agent_loop.py", line 807, in _run_all
    asyncio.run(run_all())
  File "/usr/local/python3.11.13/lib/python3.11/asyncio/runners.py", line 190, in run
    return runner.run(main)
           ^^^^^^^^^^^^^^^^
  File "/usr/local/python3.11.13/lib/python3.11/asyncio/runners.py", line 118, in run
    return self._loop.run_until_complete(task)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "uvloop/loop.pyx", line 1518, in uvloop.loop.Loop.run_until_complete
  File "/verl/verl/experimental/agent_loop/agent_loop.py", line 805, in run_all
    await asyncio.gather(*tasks)
  File "/verl/verl/workers/rollout/vllm_rollout/vllm_async_server.py", line 549, in sleep
    await self.servers[0].wait_for_requests_to_drain.remote()
ray.exceptions.RayTaskError(AttributeError): ray::vLLMHttpServer.wait_for_requests_to_drain() (pid=324161, ip=10.1.68.108, actor_id=7455a93360f01a31e1cf66ac01000000, repr=<verl.workers.rollout.vllm_rollout.vllm_async_server.vLLMHttpServer object at 0xffffb18750d0>)
  File "/usr/local/python3.11.13/lib/python3.11/concurrent/futures/_base.py", line 449, in result
    return self.__get_result()
           ^^^^^^^^^^^^^^^^^^^
  File "/usr/local/python3.11.13/lib/python3.11/concurrent/futures/_base.py", line 401, in __get_result
    raise self._exception
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/verl/verl/workers/rollout/vllm_rollout/vllm_async_server.py", line 431, in wait_for_requests_to_drain
    await self.engine.wait_for_requests_to_drain()
          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
AttributeError: 'AsyncLLM' object has no attribute 'wait_for_requests_to_drain'

Expected behavior

official examples scripts should run successfully

[1k77]

Please provide detailed environment information.