[Bug]: AsyncLLMEngine stuck in V1

Open lullabies777 opened this issue 6 months ago • 0 comments

Your current environment

INFO 06-20 10:55:55 [init.py:244] Automatically detected platform cuda. Collecting environment information...

    System Info

============================== OS : Ubuntu 22.04.5 LTS (x86_64) GCC version : (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0 Clang version : Could not collect CMake version : version 4.0.2 Libc version : glibc-2.35

============================== PyTorch Info

PyTorch version : 2.7.0+cu126 Is debug build : False CUDA used to build PyTorch : 12.6 ROCM used to build PyTorch : N/A

============================== Python Environment

Python version : 3.12.10 | packaged by conda-forge | (main, Apr 10 2025, 22:21:13) [GCC 13.3.0] (64-bit runtime) Python platform : Linux-5.10.134-16.1.al8.x86_64-x86_64-with-glibc2.35

============================== CUDA / GPU Info

Is CUDA available : True CUDA runtime version : Could not collect CUDA_MODULE_LOADING set to : LAZY GPU models and configuration : GPU 0: NVIDIA H20 Nvidia driver version : 550.54.15 cuDNN version : Could not collect HIP runtime version : N/A MIOpen runtime version : N/A Is XNNPACK available : True

============================== CPU Info

架构： x86_64 CPU 运行模式： 32-bit, 64-bit Address sizes: 52 bits physical, 57 bits virtual 字节序： Little Endian CPU: 24 在线 CPU 列表： 0-23 厂商 ID： GenuineIntel 型号名称： Intel(R) Xeon(R) Platinum 8469C CPU 系列： 6 型号： 143 每个核的线程数： 2 每个座的核数： 12 座： 1 步进： 8 CPU 最大 MHz： 3800.0000 CPU 最小 MHz： 800.0000 BogoMIPS： 5200.00 标记： fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid aperfmperf tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pdcm pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch cpuid_fault invpcid_single ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx_vnni avx512_bf16 wbnoinvd ida arat hwp hwp_notify hwp_act_window hwp_epp hwp_pkg_req avx512vbmi umip pku ospke waitpkg avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid bus_lock_detect cldemote movdiri movdir64b enqcmd fsrm uintr md_clear serialize tsxldtrk amx_bf16 avx512_fp16 amx_tile amx_int8 arch_capabilities 超管理器厂商： KVM 虚拟化类型：完全 L1d 缓存： 576 KiB (12 instances) L1i 缓存： 384 KiB (12 instances) L2 缓存： 24 MiB (12 instances) L3 缓存： 97.5 MiB (1 instance) NUMA 节点： 1 NUMA 节点0 CPU： 0-23 Vulnerability Itlb multihit: Not affected Vulnerability L1tf: Not affected Vulnerability Mds: Not affected Vulnerability Meltdown: Not affected Vulnerability Mmio stale data: Not affected Vulnerability Retbleed: Not affected Vulnerability Spec store bypass: Vulnerable Vulnerability Spectre v1: Mitigation; usercopy/swapgs barriers and __user pointer sanitization Vulnerability Spectre v2: Mitigation; Enhanced IBRS, RSB filling, PBRSB-eIBRS SW sequence Vulnerability Srbds: Not affected Vulnerability Tsx async abort: Not affected

============================== Versions of relevant libraries

[pip3] numpy==1.26.4 [pip3] nvidia-cublas-cu12==12.6.4.1 [pip3] nvidia-cuda-cupti-cu12==12.6.80 [pip3] nvidia-cuda-nvrtc-cu12==12.6.77 [pip3] nvidia-cuda-runtime [pip3] nvidia-cudnn-cu12==9.5.1.17 [pip3] nvidia-cufft-cu12==11.3.0.4 [pip3] nvidia-cufile-cu12==1.11.1.6 [pip3] nvidia-curand-cu12==10.3.7.77 [pip3] nvidia-cusolver-cu12==11.7.1.2 [pip3] nvidia-cusparse-cu12==12.5.4.2 [pip3] nvidia-cusparselt-cu12==0.6.3 [pip3] nvidia-ml-py==12.575.51 [pip3] nvidia-nccl-cu12==2.26.2 [pip3] nvidia-nvjitlink-cu12==12.6.85 [pip3] nvidia-nvtx-cu12==12.6.77 [pip3] pytorch-lightning==2.5.1.post0 [pip3] pyzmq==26.2.0 [pip3] torch==2.7.0 [pip3] torchaudio==2.7.0 [pip3] torchmetrics==0.10.3 [pip3] torchvision==0.22.0 [pip3] transformers==4.52.4 [pip3] triton==3.3.0 [conda] numpy [conda] nvidia-cublas-cu12 [conda] nvidia-cuda-cupti-cu12 [conda] nvidia-cuda-nvrtc-cu12 [conda] nvidia-cuda-runtime-cu12 [conda] nvidia-cudnn-cu12 [conda] nvidia-cufft-cu12 [conda] nvidia-cufile-cu12 [conda] nvidia-curand-cu12 [conda] nvidia-cusolver-cu12 [conda] nvidia-cusparse-cu12 [conda] nvidia-cusparselt-cu12 [conda] nvidia-ml-py [conda] nvidia-nccl-cu12 [conda] nvidia-nvjitlink-cu12 [conda] nvidia-nvtx-cu12 [conda] pytorch-lightning [conda] pyzmq [conda] torch [conda] torchaudio [conda] torchmetrics [conda] torchvision [conda] transformers [conda] triton -cu12==12.6.77 1.26.4 pypi_0 pypi 12.6.4.1 pypi_0 pypi 12.6.80 pypi_0 pypi 12.6.77 pypi_0 pypi 12.6.77 pypi_0 pypi 9.5.1.17 pypi_0 pypi 11.3.0.4 pypi_0 pypi 1.11.1.6 pypi_0 pypi 10.3.7.77 pypi_0 pypi 11.7.1.2 pypi_0 pypi 12.5.4.2 pypi_0 pypi 0.6.3 pypi_0 pypi 12.575.51 pypi_0 pypi 2.26.2 pypi_0 pypi 12.6.85 pypi_0 pypi 12.6.77 pypi_0 pypi 2.5.1.post0 pypi_0 pypi 26.4.0 pypi_0 pypi 2.7.0 pypi_0 pypi 2.7.0 pypi_0 pypi 0.10.3 pypi_0 pypi 0.22.0 pypi_0 pypi 4.52.4 pypi_0 pypi 3.3.0 pypi_0 pypi

============================== vLLM Info

ROCM Version : Could not collect Neuron SDK Version : N/A vLLM Version : 0.9.1 vLLM Build Flags: CUDA Archs: Not Set; ROCm: Disabled; Neuron: Disabled GPU Topology: GPU0 CPU Affinity NUMA Affinity GPU NUMA ID GPU0 X 0-23 0 N/A

Legend:

X = Self SYS = Connection traversing PCIe as well as the SMP interconnect between NUMA nodes (e.g., QPI/UPI) NODE = Connection traversing PCIe as well as the interconnect between PCIe Host Bridges within a NUMA node PHB = Connection traversing PCIe as well as a PCIe Host Bridge (typically the CPU) PXB = Connection traversing multiple PCIe bridges (without traversing the PCIe Host Bridge) PIX = Connection traversing at most a single PCIe bridge NV# = Connection traversing a bonded set of # NVLinks

============================== Environment Variables

NCCL_CUMEM_ENABLE=0 PYTORCH_NVML_BASED_CUDA_CHECK=1 TORCHINDUCTOR_COMPILE_THREADS=1 CUDA_MODULE_LOADING=LAZY

🐛 Describe the bug

AsyncLLMEngine stucks in V1, but works in V0.

predict.py

from vllm import SamplingParams
from vllm.lora.request import LoRARequest
import time 
# from loguru import logger
import torch

async def async_gen(model, text, uid):
    results_generator = model.generate(
            text,
            SamplingParams(
                n=1,  # Number of output sequences to return for each prompt.
                top_k=1,  # Float that controls the cumulative probability of the top tokens to consider.
                temperature=0,  # randomness of the sampling
                seed=777, # Seed for reprodicibility
                skip_special_tokens=False,  # Whether to skip special tokens in the output.
                max_tokens=1,  # Maximum number of tokens to generate per output sequence.
                ],
                logprobs=25
            ),
            uid,
            lora_request=LoRARequest("clue_type", 1, "./checkpoint-300")
        )
    
    final_output = None

    async for request_output in results_generator:
        final_output = request_output
    
    return final_output

async def predict_clue_type(model, tokenizer, inputs, uid):
    text = inputs["text"]
    
    responses = await async_gen(model, text, uid)
  
    
    return {"success": True, "prob": prob, "message": ""}

main.py

@app.post("/v1/models/{model_name}:predict")
async def predict(request: Item): # async def predict(request: Item):
    inputs = request.dict()
    st = time.time()
    md5 = hashlib.md5()
    md5.update(json.dumps(inputs, indent=4, ensure_ascii=False).encode("utf-8"))
    uid = md5.hexdigest()
    random_uid = uuid.uuid4()
    
    
    try:
        res = await predict_clue_type(engine, tokenizer, inputs, random_uid)
    except:
        error_mess = json.dumps(
            { "uid": uid,
                "input": inputs,
                "error": traceback.format_exc()
            }, indent = 4)
        
        #logger.error(error_mess)
        print(error_mess)
        res = {"success": False, "message": error_mess, "prob": -1}
    finally:
        et = time.time()
        cost_time = round(et - st, 1)
        #logger.info(f"Predict {uid} costs : {cost_time}s")
        print(f"Predict {uid} costs : {cost_time}s")
    return res



@app.on_event("startup")
async def startup_event():
    st = time.time()
    global tokenizer, engine
    tokenizer = AutoTokenizer.from_pretrained('./Qwen3-8B')

    engine = AsyncLLMEngine.from_engine_args(
            AsyncEngineArgs(
                    model="./Qwen3-8B",
                    tensor_parallel_size=1,
                    gpu_memory_utilization=0.5, 
                    trust_remote_code=True,
                    dtype="bfloat16" if is_bfloat16_supported() else "float16", 
                    enforce_eager=True,
                    max_model_len=4096,
                    disable_log_stats=True,
                    max_logprobs=25,
                    enable_lora=True,
                    max_lora_rank=64,
                    disable_log_requests=True
            )
        )
    et = time.time()
    cost_time = round(et - st, 1)
    print(f"Initalization costs : {cost_time}s")

Before submitting a new issue...

[x] 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.

Jun 20 '25 02:06 lullabies777