[Bug]: Use v1 engine to load lora weights. If tp=1, the step of creating cudagraph will only use cpu. This causes this process to take a very long time. If tp>1, the gpu will be used normally for processing.

[piekey1994]

Your current environment

The output of `python collect_env.py`

INFO 04-17 18:59:57 [__init__.py:239] Automatically detected platform cuda.
Collecting environment information...
PyTorch version: 2.6.0+cu124
Is debug build: False
CUDA used to build PyTorch: 12.4
ROCM used to build PyTorch: N/A

OS: Ubuntu 22.04.4 LTS (x86_64)
GCC version: (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0
Clang version: Could not collect
CMake version: version 4.0.0
Libc version: glibc-2.35

Python version: 3.12.9 (main, Feb  5 2025, 08:49:00) [GCC 11.4.0] (64-bit runtime)
Python platform: Linux-5.4.119-19.0009.37-x86_64-with-glibc2.35
Is CUDA available: True
CUDA runtime version: 12.4.131
CUDA_MODULE_LOADING set to: LAZY
GPU models and configuration: GPU 0: NVIDIA H800
Nvidia driver version: 535.104.05
cuDNN version: Could not collect
HIP runtime version: N/A
MIOpen runtime version: N/A
Is XNNPACK available: True
Architecture:                    x86_64
CPU op-mode(s):                  32-bit, 64-bit
Address sizes:                   52 bits physical, 48 bits virtual
Byte Order:                      Little Endian
CPU(s):                          192
On-line CPU(s) list:             0-191
Vendor ID:                       GenuineIntel
Model name:                      Intel(R) Xeon(R) Platinum 8468
CPU family:                      6
Model:                           143
Thread(s) per core:              1
Core(s) per socket:              96
Socket(s):                       2
Stepping:                        6
BogoMIPS:                        4200.00
Flags:                           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 cpuid tsc_known_freq pni pclmulqdq ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch cpuid_fault invpcid_single ssbd ibrs ibpb ibrs_enhanced fsgsbase 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 avx512_bf16 wbnoinvd arat avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq movdiri movdir64b fsrm arch_capabilities

Hypervisor vendor:               KVM
Virtualization type:             full
L1d cache:                       4.5 MiB (96 instances)
L1i cache:                       3 MiB (96 instances)
L2 cache:                        192 MiB (96 instances)
L3 cache:                        210 MiB (2 instances)
NUMA node(s):                    2
NUMA node0 CPU(s):               0-95
NUMA node1 CPU(s):               96-191
Vulnerability Itlb multihit:     Not affected
Vulnerability L1tf:              Not affected
Vulnerability Mds:               Not affected
Vulnerability Meltdown:          Not affected
Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl and seccomp
Vulnerability Spectre v1:        Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:        Mitigation; Enhanced IBRS, IBPB conditional, RSB filling
Vulnerability Srbds:             Not affected
Vulnerability Tsx async abort:   Not affected
Versions of relevant libraries:
[pip3] flashinfer-python==0.2.1.post2+cu124torch2.6
[pip3] numpy==1.26.4
[pip3] nvidia-cublas-cu12==12.4.5.8
[pip3] nvidia-cuda-cupti-cu12==12.4.127
[pip3] nvidia-cuda-nvrtc-cu12==12.4.127
[pip3] nvidia-cuda-runtime-cu12==12.4.127
[pip3] nvidia-cudnn-cu12==9.1.0.70
[pip3] nvidia-cufft-cu12==11.2.1.3
[pip3] nvidia-curand-cu12==10.3.5.147
[pip3] nvidia-cusolver-cu12==11.6.1.9
[pip3] nvidia-cusparse-cu12==12.3.1.170
[pip3] nvidia-cusparselt-cu12==0.6.2
[pip3] nvidia-ml-py==12.570.86
[pip3] nvidia-nccl-cu12==2.21.5
[pip3] nvidia-nvjitlink-cu12==12.4.127
[pip3] nvidia-nvtx-cu12==12.4.127
[pip3] pynvml==12.0.0
[pip3] pyzmq==26.4.0
[pip3] torch==2.6.0
[pip3] torchaudio==2.6.0
[pip3] torchvision==0.21.0
[pip3] transformers==4.51.0
[pip3] triton==3.2.0
[conda] Could not collect
ROCM Version: Could not collect
Neuron SDK Version: N/A
vLLM Version: 0.8.3
vLLM Build Flags:
CUDA Archs: Not Set; ROCm: Disabled; Neuron: Disabled
GPU Topology:
�[4mGPU0	CPU Affinity	NUMA Affinity	GPU NUMA ID�[0m
GPU0	 X 	0-95	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

NVIDIA_VISIBLE_DEVICES=none
NVIDIA_REQUIRE_CUDA=cuda>=12.4 brand=tesla,driver>=470,driver<471 brand=unknown,driver>=470,driver<471 brand=nvidia,driver>=470,driver<471 brand=nvidiartx,driver>=470,driver<471 brand=geforce,driver>=470,driver<471 brand=geforcertx,driver>=470,driver<471 brand=quadro,driver>=470,driver<471 brand=quadrortx,driver>=470,driver<471 brand=titan,driver>=470,driver<471 brand=titanrtx,driver>=470,driver<471 brand=tesla,driver>=525,driver<526 brand=unknown,driver>=525,driver<526 brand=nvidia,driver>=525,driver<526 brand=nvidiartx,driver>=525,driver<526 brand=geforce,driver>=525,driver<526 brand=geforcertx,driver>=525,driver<526 brand=quadro,driver>=525,driver<526 brand=quadrortx,driver>=525,driver<526 brand=titan,driver>=525,driver<526 brand=titanrtx,driver>=525,driver<526 brand=tesla,driver>=535,driver<536 brand=unknown,driver>=535,driver<536 brand=nvidia,driver>=535,driver<536 brand=nvidiartx,driver>=535,driver<536 brand=geforce,driver>=535,driver<536 brand=geforcertx,driver>=535,driver<536 brand=quadro,driver>=535,driver<536 brand=quadrortx,driver>=535,driver<536 brand=titan,driver>=535,driver<536 brand=titanrtx,driver>=535,driver<536
NCCL_VERSION=2.20.5-1
VLLM_CACHE_ROOT=/data/tmp/.cache/vllm
VLLM_CONFIG_ROOT=/data/tmp/.config/vllm
NVIDIA_DRIVER_CAPABILITIES=compute,utility
NVIDIA_PRODUCT_NAME=CUDA
VLLM_USAGE_SOURCE=production-docker-image
CUDA_VERSION=12.4.0
LD_LIBRARY_PATH=/usr/local/nvidia/lib:/usr/local/nvidia/lib64
VLLM_USE_V1=0
NCCL_CUMEM_ENABLE=0
PYTORCH_NVML_BASED_CUDA_CHECK=1
TORCHINDUCTOR_COMPILE_THREADS=1
CUDA_MODULE_LOADING=LAZY

🐛 Describe the bug

See this issue for details: https://github.com/vllm-project/vllm/issues/16961

Why does v1 engine only use cpu to create cudagraph when only one gpu is used? Is there something wrong with my graphics card or driver? This phenomenon always exists from 0.8.3 to 0.8.5.

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.

[jeejeelee]

Could you describe it in more detail?

[cjackal]

Could you describe it in more detail?

It's a reminder of #16961 , details are there. I couldn't reproduce the bug with 1 lora 1 tp on H100 though.

[jeejeelee]

@cjackal It might be an environment issue. Could you please provide your environment information? Thank you very much.

[lhcavalcanti]

I'm having similar problem, when starting vllm (0.8.5, engine v1) with Phi-4-MM, it keeps hanging after kv_cache_utils logger:

INFO 05-01 00:07:00 [backends.py:430] Dynamo bytecode transform time: 9.88 s
INFO 05-01 00:07:03 [backends.py:136] Cache the graph of shape None for later use
INFO 05-01 00:07:03 [backends.py:136] Cache the graph of shape None for later use
INFO 05-01 00:07:03 [backends.py:136] Cache the graph of shape None for later use
INFO 05-01 00:07:03 [backends.py:136] Cache the graph of shape None for later use
INFO 05-01 00:07:40 [backends.py:148] Compiling a graph for general shape takes 39.42 s
INFO 05-01 00:07:40 [backends.py:148] Compiling a graph for general shape takes 41.35 s
INFO 05-01 00:07:40 [backends.py:148] Compiling a graph for general shape takes 39.91 s
INFO 05-01 00:07:40 [backends.py:148] Compiling a graph for general shape takes 40.34 s
INFO 05-01 00:08:06 [monitor.py:33] torch.compile takes 49.93 s in total
INFO 05-01 00:08:06 [monitor.py:33] torch.compile takes 53.46 s in total
INFO 05-01 00:08:06 [monitor.py:33] torch.compile takes 49.31 s in total
INFO 05-01 00:08:06 [monitor.py:33] torch.compile takes 50.45 s in total
INFO 05-01 00:08:06 [kv_cache_utils.py:634] GPU KV cache size: 481,312 tokens
INFO 05-01 00:08:06 [kv_cache_utils.py:637] Maximum concurrency for 131,072 tokens per request: 3.67x
INFO 05-01 00:08:06 [kv_cache_utils.py:634] GPU KV cache size: 481,312 tokens
INFO 05-01 00:08:06 [kv_cache_utils.py:637] Maximum concurrency for 131,072 tokens per request: 3.67x
INFO 05-01 00:08:07 [kv_cache_utils.py:634] GPU KV cache size: 481,312 tokens
INFO 05-01 00:08:07 [kv_cache_utils.py:637] Maximum concurrency for 131,072 tokens per request: 3.67x
INFO 05-01 00:08:07 [kv_cache_utils.py:634] GPU KV cache size: 481,312 tokens
INFO 05-01 00:08:07 [kv_cache_utils.py:637] Maximum concurrency for 131,072 tokens per request: 3.67x

It keeps stuck there for > 20 minutes. CPU load is 75+ for 96 cores.

[houseroad]

cc: @frank-wei, we need to look into this.

[piekey1994]

export OMP_NUM_THREADS=2 just ok!