vllm: [Bug]: LLM is not getting loaded on multiple GPUs but works fine on a single GPU
Your current environment
PyTorch version: 2.1.2
Is debug build: False
CUDA used to build PyTorch: 12.1
ROCM used to build PyTorch: N/A
OS: Ubuntu 20.04.6 LTS (x86_64)
GCC version: (Ubuntu 9.4.0-1ubuntu1~20.04.2) 9.4.0
Clang version: Could not collect
CMake version: version 3.29.1
Libc version: glibc-2.31
Python version: 3.10.13 (main, Sep 11 2023, 13:44:35) [GCC 11.2.0] (64-bit runtime)
Python platform: Linux-5.15.0-94-generic-x86_64-with-glibc2.31
Is CUDA available: True
CUDA runtime version: 12.1.105
CUDA_MODULE_LOADING set to: LAZY
GPU models and configuration:
GPU 0: NVIDIA RTX 6000 Ada Generation
GPU 1: NVIDIA RTX 6000 Ada Generation
GPU 2: NVIDIA RTX 6000 Ada Generation
GPU 3: NVIDIA RTX 6000 Ada Generation
Nvidia driver version: 535.154.05
cuDNN version: Probably one of the following:
/usr/lib/x86_64-linux-gnu/libcudnn.so.8.9.0
/usr/lib/x86_64-linux-gnu/libcudnn_adv_infer.so.8.9.0
/usr/lib/x86_64-linux-gnu/libcudnn_adv_train.so.8.9.0
/usr/lib/x86_64-linux-gnu/libcudnn_cnn_infer.so.8.9.0
/usr/lib/x86_64-linux-gnu/libcudnn_cnn_train.so.8.9.0
/usr/lib/x86_64-linux-gnu/libcudnn_ops_infer.so.8.9.0
/usr/lib/x86_64-linux-gnu/libcudnn_ops_train.so.8.9.0
HIP runtime version: N/A
MIOpen runtime version: N/A
Is XNNPACK available: True
CPU:
Architecture: x86_64
CPU op-mode(s): 32-bit, 64-bit
Byte Order: Little Endian
Address sizes: 46 bits physical, 57 bits virtual
CPU(s): 96
On-line CPU(s) list: 0-95
Thread(s) per core: 2
Core(s) per socket: 24
Socket(s): 2
NUMA node(s): 2
Vendor ID: GenuineIntel
CPU family: 6
Model: 143
Model name: Intel(R) Xeon(R) Gold 5418Y
Stepping: 8
Frequency boost: enabled
CPU MHz: 989.502
CPU max MHz: 2001.0000
CPU min MHz: 800.0000
BogoMIPS: 4000.00
Virtualization: VT-x
L1d cache: 2.3 MiB
L1i cache: 1.5 MiB
L2 cache: 96 MiB
L3 cache: 90 MiB
NUMA node0 CPU(s): 0-23,48-71
NUMA node1 CPU(s): 24-47,72-95
Vulnerability Gather data sampling: Not affected
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 rstack overflow: 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, PBRSB-eIBRS SW sequence
Vulnerability Srbds: Not affected
Vulnerability Tsx async abort: Not affected
Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf tsc_known_freq pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid dca sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault epb cat_l3 cat_l2 cdp_l3 invpcid_single cdp_l2 ssbd mba ibrs ibpb stibp ibrs_enhanced tpr_shadow vnmi flexpriority ept vpid ept_ad fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid cqm rdt_a avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb intel_pt avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local split_lock_detect avx_vnni avx512_bf16 wbnoinvd dtherm ida arat pln pts avx512vbmi umip pku ospke waitpkg avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg tme avx512_vpopcntdq la57 rdpid bus_lock_detect cldemote movdiri movdir64b enqcmd fsrm md_clear serialize tsxldtrk pconfig arch_lbr amx_bf16 avx512_fp16 amx_tile amx_int8 flush_l1d arch_capabilities
Versions of relevant libraries:
[pip3] mypy-extensions==1.0.0
[pip3] numpy==1.26.2
[pip3] onnxruntime==1.17.1
[pip3] torch==2.1.2
[pip3] torchaudio==2.1.2
[pip3] torchelastic==0.2.2
[pip3] torchvision==0.16.2
[pip3] triton==2.1.0
[conda] blas 1.0 mkl
[conda] ffmpeg 4.3 hf484d3e_0 pytorch
[conda] libjpeg-turbo 2.0.0 h9bf148f_0 pytorch
[conda] mkl 2023.1.0 h213fc3f_46344
[conda] mkl-service 2.4.0 py310h5eee18b_1
[conda] mkl_fft 1.3.8 py310h5eee18b_0
[conda] mkl_random 1.2.4 py310hdb19cb5_0
[conda] numpy 1.26.2 py310h5f9d8c6_0
[conda] numpy-base 1.26.2 py310hb5e798b_0
[conda] pytorch 2.1.2 py3.10_cuda12.1_cudnn8.9.2_0 pytorch
[conda] pytorch-cuda 12.1 ha16c6d3_5 pytorch
[conda] pytorch-mutex 1.0 cuda pytorch
[conda] torchaudio 2.1.2 py310_cu121 pytorch
[conda] torchelastic 0.2.2 pypi_0 pypi
[conda] torchtriton 2.1.0 py310 pytorch
[conda] torchvision 0.16.2 py310_cu121 pytorchROCM Version: Could not collect
Neuron SDK Version: N/A
vLLM Version: 0.4.0.post1
vLLM Build Flags:
CUDA Archs: Not Set; ROCm: Disabled; Neuron: Disabled
GPU Topology:
GPU0 GPU1 GPU2 GPU3 CPU Affinity NUMA Affinity GPU NUMA ID
GPU0 X PIX SYS SYS 24-47,72-95 1 N/A
GPU1 PIX X SYS SYS 24-47,72-95 1 N/A
GPU2 SYS SYS X PIX 24-47,72-95 1 N/A
GPU3 SYS SYS PIX X 24-47,72-95 1 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
🐛 Describe the bug
When I try to load the model by using the following command
from vllm import LLM
llm = LLM(model="./mistral/models--mistralai--Mistral-7B-Instruct-v0.2/snapshots/", tensor_parallel_size=2)
The model is not loaded at all, I get the following information on the CLI and that’s it. The loading is never finished.
I can see that the 2 GPU devices are occupied while the above message is displayed, but nothing else. The line of code is never fully executed.
When I try to load the model using only one GPU, the loading process is smooth.
from vllm import LLM
llm = LLM(model="./mistral/models--mistralai--Mistral-7B-Instruct-v0.2/snapshots/", tensor_parallel_size=1)
Below is the screenshot of the successfull loading message:
The llm inference is quite fast and everyhting works as expected.
So the problem clearly lies with multiple GPUs. This issue happens with all the models and not particular to just one organisation. Can someone please help me in this regard? What am I doing wrong? Is it something due to nccl or is something mssing? Any help is appreciated, thanks 😃
About this issue
- Original URL
- State: closed
- Created 3 years ago
- Reactions: 1
- Comments: 20 (11 by maintainers)
Commits related to this issue
- [Bugfix] Fix pcie nvlink topology detection (#3974) — committed to agt/vllm by agt 3 years ago
- [Bugfix] Fix CustomAllreduce nvlink topology detection (#3974) [Bugfix] Fix CustomAllreduce pcie nvlink topology detection (#3974) (#4159) — committed to vllm-project/vllm by agt 3 years ago
- [Bugfix] Fix CustomAllreduce nvlink topology detection (#3974) [Bugfix] Fix CustomAllreduce pcie nvlink topology detection (#3974) (#4159) — committed to xjpang/vllm by agt 3 years ago
- [Bugfix] Fix CustomAllreduce nvlink topology detection (#3974) [Bugfix] Fix CustomAllreduce pcie nvlink topology detection (#3974) (#4159) — committed to neuralmagic/nm-vllm by agt 3 years ago
- [Bugfix] Fix CustomAllreduce nvlink topology detection (#3974) [Bugfix] Fix CustomAllreduce pcie nvlink topology detection (#3974) (#4159) — committed to neuralmagic/nm-vllm by agt 3 years ago
- [Bugfix] Fix CustomAllreduce nvlink topology detection (#3974) [Bugfix] Fix CustomAllreduce pcie nvlink topology detection (#3974) (#4159) — committed to alexeykondrat/ci-vllm by agt 3 years ago
- [Bugfix] Fix CustomAllreduce nvlink topology detection (#3974) [Bugfix] Fix CustomAllreduce pcie nvlink topology detection (#3974) (#4159) — committed to z103cb/opendatahub_vllm by agt 3 years ago
I just found a solution that works! I set the env variable as
export NCCL_P2P_DISABLE=1This works for me now 😃
I am using
microsoft/Phi-3-vision-128k-instructand it gives out of memory error. But if I usefacebook/opt-13b, then it works fine even if it much bigger model.Command with output
My environment
Hi @venki-lfc, sorry to hear that didn’t work! 0.4.1 will include an option to log all function calls, perhaps doing so will identify the culprit as it did for me. I’d be happy to review if you post that info in a new bug.
Good job! nccl is quite a black box, and we have a hard time with it 😦
I guess we can keep the issue open 😃 Obviously mine’s just a work around and doesn’t address the root cause of the issue.