[Bug]: PyNcclCommunicator error when inferencing #8420
Description
Your current environment
The output of `python collect_env.py`
Collecting environment information...
PyTorch version: 2.4.0
Is debug build: False
CUDA used to build PyTorch: 12.1
ROCM used to build PyTorch: N/A
OS: Ubuntu 18.04.6 LTS (x86_64)
GCC version: (Ubuntu 7.5.0-3ubuntu1~18.04) 7.5.0
Clang version: Could not collect
CMake version: Could not collect
Libc version: glibc-2.27
Python version: 3.10.14 (main, May 6 2024, 19:42:50) [GCC 11.2.0] (64-bit runtime)
Python platform: Linux-5.4.0-150-generic-x86_64-with-glibc2.27
Is CUDA available: True
CUDA runtime version: Could not collect
CUDA_MODULE_LOADING set to: LAZY
GPU models and configuration:
GPU 0: NVIDIA GeForce RTX 4090
GPU 1: NVIDIA GeForce RTX 4090
GPU 2: NVIDIA GeForce RTX 4090
GPU 3: NVIDIA GeForce RTX 4090
Nvidia driver version: 535.54.03
cuDNN version: Probably one of the following:
/usr/local/cuda-11.7/targets/x86_64-linux/lib/libcudnn.so.8
/usr/local/cuda-11.7/targets/x86_64-linux/lib/libcudnn_adv_infer.so.8
/usr/local/cuda-11.7/targets/x86_64-linux/lib/libcudnn_adv_train.so.8
/usr/local/cuda-11.7/targets/x86_64-linux/lib/libcudnn_cnn_infer.so.8
/usr/local/cuda-11.7/targets/x86_64-linux/lib/libcudnn_cnn_train.so.8
/usr/local/cuda-11.7/targets/x86_64-linux/lib/libcudnn_ops_infer.so.8
/usr/local/cuda-11.7/targets/x86_64-linux/lib/libcudnn_ops_train.so.8
/usr/local/cuda-12.0/targets/x86_64-linux/lib/libcudnn.so.8.8.0
/usr/local/cuda-12.0/targets/x86_64-linux/lib/libcudnn_adv_infer.so.8.8.0
/usr/local/cuda-12.0/targets/x86_64-linux/lib/libcudnn_adv_train.so.8.8.0
/usr/local/cuda-12.0/targets/x86_64-linux/lib/libcudnn_cnn_infer.so.8.8.0
/usr/local/cuda-12.0/targets/x86_64-linux/lib/libcudnn_cnn_train.so.8.8.0
/usr/local/cuda-12.0/targets/x86_64-linux/lib/libcudnn_ops_infer.so.8.8.0
/usr/local/cuda-12.0/targets/x86_64-linux/lib/libcudnn_ops_train.so.8.8.0
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
CPU(s): 72
On-line CPU(s) list: 0-71
Thread(s) per core: 2
Core(s) per socket: 18
Socket(s): 2
NUMA node(s): 2
Vendor ID: GenuineIntel
CPU family: 6
Model: 85
Model name: Intel(R) Xeon(R) Gold 6240 CPU @ 2.60GHz
Stepping: 7
CPU MHz: 3300.000
CPU max MHz: 3900.0000
CPU min MHz: 1000.0000
BogoMIPS: 5200.00
Virtualization: VT-x
L1d cache: 32K
L1i cache: 32K
L2 cache: 1024K
L3 cache: 25344K
NUMA node0 CPU(s): 0-17,36-53
NUMA node1 CPU(s): 18-35,54-71
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 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 cdp_l3 invpcid_single intel_ppin 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 mpx rdt_a avx512f avx512dq rdseed adx smap clflushopt clwb intel_pt avx512cd avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local dtherm ida arat pln pts pku ospke avx512_vnni md_clear flush_l1d arch_capabilities
Versions of relevant libraries:
[pip3] numpy==1.26.4
[pip3] nvidia-ml-py==12.560.30
[pip3] pyzmq==26.2.0
[pip3] torch==2.4.0
[pip3] torchaudio==2.4.0
[pip3] torchvision==0.19.0
[pip3] transformers==4.44.2
[pip3] triton==3.0.0
[conda] blas 1.0 mkl defaults
[conda] cuda-cudart 12.1.105 0 nvidia
[conda] cuda-cupti 12.1.105 0 nvidia
[conda] cuda-libraries 12.1.0 0 nvidia
[conda] cuda-nvrtc 12.1.105 0 nvidia
[conda] cuda-nvtx 12.1.105 0 nvidia
[conda] cuda-opencl 12.6.68 0 nvidia
[conda] cuda-runtime 12.1.0 0 nvidia
[conda] cuda-version 12.6 3 nvidia
[conda] ffmpeg 4.3 hf484d3e_0 pytorch
[conda] libcublas 12.1.0.26 0 nvidia
[conda] libcufft 11.0.2.4 0 nvidia
[conda] libcufile 1.11.1.6 0 nvidia
[conda] libcurand 10.3.7.68 0 nvidia
[conda] libcusolver 11.4.4.55 0 nvidia
[conda] libcusparse 12.0.2.55 0 nvidia
[conda] libjpeg-turbo 2.0.0 h9bf148f_0 pytorch
[conda] libnpp 12.0.2.50 0 nvidia
[conda] libnvjitlink 12.1.105 0 nvidia
[conda] libnvjpeg 12.1.1.14 0 nvidia
[conda] mkl 2023.1.0 h213fc3f_46344 defaults
[conda] mkl-service 2.4.0 py310h5eee18b_1 defaults
[conda] mkl_fft 1.3.10 py310h5eee18b_0 defaults
[conda] mkl_random 1.2.7 py310h1128e8f_0 defaults
[conda] numpy 1.26.4 pypi_0 pypi
[conda] nvidia-ml-py 12.560.30 pypi_0 pypi
[conda] pytorch 2.4.0 py3.10_cuda12.1_cudnn9.1.0_0 pytorch
[conda] pytorch-cuda 12.1 ha16c6d3_5 pytorch
[conda] pytorch-mutex 1.0 cuda pytorch
[conda] pyzmq 26.2.0 pypi_0 pypi
[conda] torchaudio 2.4.0 py310_cu121 pytorch
[conda] torchtriton 3.0.0 py310 pytorch
[conda] torchvision 0.19.0 py310_cu121 pytorch
[conda] transformers 4.44.2 pypi_0 pypi
ROCM Version: Could not collect
Neuron SDK Version: N/A
vLLM Version: 0.6.0@32e7db25365415841ebc7c4215851743fbb1bad1
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 NODE SYS SYS 0-17,36-53 0 N/A
GPU1 NODE X SYS SYS 0-17,36-53 0 N/A
GPU2 SYS SYS X NODE 18-35,54-71 1 N/A
GPU3 SYS SYS NODE X 18-35,54-71 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
Model Input Dumps
No response
🐛 Describe the bug
When I started creating an instance from the LLM class, an error occurred.
It says AttributeError: 'PyNcclCommunicator' object has no attribute 'device', I don't know what to do then.
The code to test it:
from vllm import LLM
def init_model() -> LLM:
llm = LLM(
model="Qwen/Qwen2-7B-Instruct",
tokenizer_mode="auto",
trust_remote_code=True,
download_dir="./.cache",
tensor_parallel_size=2, # How many GPUs to use
gpu_memory_utilization=0.85,
pipeline_parallel_size=1,
dtype="bfloat16",
# max_model_len=20480, # Model context length
enable_prefix_caching=True,
enable_chunked_prefill=False,
num_scheduler_steps=8,
)
return llm
if __name__ == "__main__":
llm = init_model()
print(llm.generate("Hello, world!"))
INFO 09-13 00:13:27 config.py:890] Defaulting to use mp for distributed inference
WARNING 09-13 00:13:27 arg_utils.py:880] Enabled BlockSpaceManagerV2 because it is required for multi-step (--num-scheduler-steps > 1)
INFO 09-13 00:13:27 llm_engine.py:213] Initializing an LLM engine (v0.6.0) with config: model='Qwen/Qwen2-7B-Instruct', speculative_config=None, tokenizer='Qwen/Qwen2-7B-Instruct', skip_tokenizer_init=False, tokenizer_mode=auto, revision=None, override_neuron_config=None, rope_scaling=None, rope_theta=None, tokenizer_revision=None, trust_remote_code=True, dtype=torch.bfloat16, max_seq_len=32768, download_dir='./.cache', load_format=LoadFormat.AUTO, tensor_parallel_size=2, pipeline_parallel_size=1, disable_custom_all_reduce=False, quantization=None, enforce_eager=False, kv_cache_dtype=auto, quantization_param_path=None, device_config=cuda, decoding_config=DecodingConfig(guided_decoding_backend='outlines'), observability_config=ObservabilityConfig(otlp_traces_endpoint=None, collect_model_forward_time=False, collect_model_execute_time=False), seed=0, served_model_name=Qwen/Qwen2-7B-Instruct, use_v2_block_manager=True, num_scheduler_steps=8, enable_prefix_caching=True, use_async_output_proc=True)
WARNING 09-13 00:13:28 multiproc_gpu_executor.py:56] Reducing Torch parallelism from 36 threads to 1 to avoid unnecessary CPU contention. Set OMP_NUM_THREADS in the external environment to tune this value as needed.
INFO 09-13 00:13:29 custom_cache_manager.py:17] Setting Triton cache manager to: vllm.triton_utils.custom_cache_manager:CustomCacheManager
WARNING 09-13 00:13:29 registry.py:190] `mm_limits` has already been set for model=Qwen/Qwen2-7B-Instruct, and will be overwritten by the new values.
(VllmWorkerProcess pid=18643) WARNING 09-13 00:13:29 registry.py:190] `mm_limits` has already been set for model=Qwen/Qwen2-7B-Instruct, and will be overwritten by the new values.
(VllmWorkerProcess pid=18643) INFO 09-13 00:13:29 multiproc_worker_utils.py:215] Worker ready; awaiting tasks
INFO 09-13 00:13:29 utils.py:977] Found nccl from library libnccl.so.2
(VllmWorkerProcess pid=18643) INFO 09-13 00:13:29 utils.py:977] Found nccl from library libnccl.so.2
ERROR 09-13 00:13:29 pynccl_wrapper.py:196] Failed to load NCCL library from libnccl.so.2 .It is expected if you are not running on NVIDIA/AMD GPUs.Otherwise, the nccl library might not exist, be corrupted or it does not support the current platform Linux-5.4.0-150-generic-x86_64-with-glibc2.27.If you already have the library, please set the environment variable VLLM_NCCL_SO_PATH to point to the correct nccl library path.
(VllmWorkerProcess pid=18643) ERROR 09-13 00:13:29 pynccl_wrapper.py:196] Failed to load NCCL library from libnccl.so.2 .It is expected if you are not running on NVIDIA/AMD GPUs.Otherwise, the nccl library might not exist, be corrupted or it does not support the current platform Linux-5.4.0-150-generic-x86_64-with-glibc2.27.If you already have the library, please set the environment variable VLLM_NCCL_SO_PATH to point to the correct nccl library path.
(VllmWorkerProcess pid=18643) INFO 09-13 00:13:29 custom_all_reduce_utils.py:242] reading GPU P2P access cache from ~/.cache/vllm/gpu_p2p_access_cache_for_0,1.json
INFO 09-13 00:13:29 custom_all_reduce_utils.py:242] reading GPU P2P access cache from ~/.cache/vllm/gpu_p2p_access_cache_for_0,1.json
WARNING 09-13 00:13:29 custom_all_reduce.py:131] Custom allreduce is disabled because your platform lacks GPU P2P capability or P2P test failed. To silence this warning, specify disable_custom_all_reduce=True explicitly.
(VllmWorkerProcess pid=18643) WARNING 09-13 00:13:29 custom_all_reduce.py:131] Custom allreduce is disabled because your platform lacks GPU P2P capability or P2P test failed. To silence this warning, specify disable_custom_all_reduce=True explicitly.
INFO 09-13 00:13:29 shm_broadcast.py:235] vLLM message queue communication handle: Handle(connect_ip='127.0.0.1', local_reader_ranks=[1], buffer=<vllm.distributed.device_communicators.shm_broadcast.ShmRingBuffer object at 0x7fac3ec98730>, local_subscribe_port=46869, remote_subscribe_port=None)
INFO 09-13 00:13:29 model_runner.py:915] Starting to load model Qwen/Qwen2-7B-Instruct...
(VllmWorkerProcess pid=18643) INFO 09-13 00:13:29 model_runner.py:915] Starting to load model Qwen/Qwen2-7B-Instruct...
INFO 09-13 00:13:30 weight_utils.py:236] Using model weights format ['*.safetensors']
(VllmWorkerProcess pid=18643) INFO 09-13 00:13:31 weight_utils.py:236] Using model weights format ['*.safetensors']
Loading safetensors checkpoint shards: 0% Completed | 0/4 [00:00<?, ?it/s]
Loading safetensors checkpoint shards: 25% Completed | 1/4 [00:01<00:04, 1.54s/it]
Loading safetensors checkpoint shards: 50% Completed | 2/4 [00:03<00:03, 1.71s/it]
Loading safetensors checkpoint shards: 75% Completed | 3/4 [00:05<00:01, 1.75s/it]
(VllmWorkerProcess pid=18643) INFO 09-13 00:13:38 model_runner.py:926] Loading model weights took 7.1216 GB
Loading safetensors checkpoint shards: 100% Completed | 4/4 [00:07<00:00, 1.80s/it]
Loading safetensors checkpoint shards: 100% Completed | 4/4 [00:07<00:00, 1.76s/it]
INFO 09-13 00:13:38 model_runner.py:926] Loading model weights took 7.1216 GB
INFO 09-13 00:13:44 distributed_gpu_executor.py:57] # GPU blocks: 22639, # CPU blocks: 9362
INFO 09-13 00:13:50 model_runner.py:1217] Capturing the model for CUDA graphs. This may lead to unexpected consequences if the model is not static. To run the model in eager mode, set 'enforce_eager=True' or use '--enforce-eager' in the CLI.
INFO 09-13 00:13:50 model_runner.py:1221] CUDA graphs can take additional 1~3 GiB memory per GPU. If you are running out of memory, consider decreasing `gpu_memory_utilization` or enforcing eager mode. You can also reduce the `max_num_seqs` as needed to decrease memory usage.
(VllmWorkerProcess pid=18643) INFO 09-13 00:13:50 model_runner.py:1217] Capturing the model for CUDA graphs. This may lead to unexpected consequences if the model is not static. To run the model in eager mode, set 'enforce_eager=True' or use '--enforce-eager' in the CLI.
(VllmWorkerProcess pid=18643) INFO 09-13 00:13:50 model_runner.py:1221] CUDA graphs can take additional 1~3 GiB memory per GPU. If you are running out of memory, consider decreasing `gpu_memory_utilization` or enforcing eager mode. You can also reduce the `max_num_seqs` as needed to decrease memory usage.
[rank0]: Traceback (most recent call last):
[rank0]: File "~/psp/Reasoning-Carefully/test.py", line 25, in <module>
[rank0]: llm = init_model()
[rank0]: File "~/psp/Reasoning-Carefully/test.py", line 6, in init_model
[rank0]: llm = LLM(
[rank0]: File "~/anaconda3/envs/psp/lib/python3.10/site-packages/vllm/entrypoints/llm.py", line 177, in __init__
[rank0]: self.llm_engine = LLMEngine.from_engine_args(
[rank0]: File "~/anaconda3/envs/psp/lib/python3.10/site-packages/vllm/engine/llm_engine.py", line 538, in from_engine_args
[rank0]: engine = cls(
[rank0]: File "~/anaconda3/envs/psp/lib/python3.10/site-packages/vllm/engine/llm_engine.py", line 319, in __init__
[rank0]: self._initialize_kv_caches()
[rank0]: File "~/anaconda3/envs/psp/lib/python3.10/site-packages/vllm/engine/llm_engine.py", line 461, in _initialize_kv_caches
[rank0]: self.model_executor.initialize_cache(num_gpu_blocks, num_cpu_blocks)
[rank0]: File "~/anaconda3/envs/psp/lib/python3.10/site-packages/vllm/executor/distributed_gpu_executor.py", line 63, in initialize_cache
[rank0]: self._run_workers("initialize_cache",
[rank0]: File "~/anaconda3/envs/psp/lib/python3.10/site-packages/vllm/executor/multiproc_gpu_executor.py", line 199, in _run_workers
[rank0]: driver_worker_output = driver_worker_method(*args, **kwargs)
[rank0]: File "~/anaconda3/envs/psp/lib/python3.10/site-packages/vllm/worker/worker.py", line 265, in initialize_cache
[rank0]: self._warm_up_model()
[rank0]: File "~/anaconda3/envs/psp/lib/python3.10/site-packages/vllm/worker/worker.py", line 281, in _warm_up_model
[rank0]: self.model_runner.capture_model(self.gpu_cache)
[rank0]: File "~/anaconda3/envs/psp/lib/python3.10/site-packages/vllm/worker/multi_step_model_runner.py", line 543, in capture_model
[rank0]: return self._base_model_runner.capture_model(kv_caches)
[rank0]: File "~/anaconda3/envs/psp/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 116, in decorate_context
[rank0]: return func(*args, **kwargs)
[rank0]: File "~/anaconda3/envs/psp/lib/python3.10/site-packages/vllm/worker/model_runner.py", line 1327, in capture_model
[rank0]: graph_runner.capture(**capture_inputs)
[rank0]: File "~/anaconda3/envs/psp/lib/python3.10/site-packages/vllm/worker/model_runner.py", line 1569, in capture
[rank0]: self.model(
[rank0]: File "~/anaconda3/envs/psp/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1553, in _wrapped_call_impl
[rank0]: return self._call_impl(*args, **kwargs)
[rank0]: File "~/anaconda3/envs/psp/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1562, in _call_impl
[rank0]: return forward_call(*args, **kwargs)
[rank0]: File "~/anaconda3/envs/psp/lib/python3.10/site-packages/vllm/model_executor/models/qwen2.py", line 361, in forward
[rank0]: hidden_states = self.model(input_ids, positions, kv_caches,
[rank0]: File "~/anaconda3/envs/psp/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1553, in _wrapped_call_impl
[rank0]: return self._call_impl(*args, **kwargs)
[rank0]: File "~/anaconda3/envs/psp/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1562, in _call_impl
[rank0]: return forward_call(*args, **kwargs)
[rank0]: File "~/anaconda3/envs/psp/lib/python3.10/site-packages/vllm/model_executor/models/qwen2.py", line 269, in forward
[rank0]: hidden_states = self.embed_tokens(input_ids)
[rank0]: File "~/anaconda3/envs/psp/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1553, in _wrapped_call_impl
[rank0]: return self._call_impl(*args, **kwargs)
[rank0]: File "~/anaconda3/envs/psp/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1562, in _call_impl
[rank0]: return forward_call(*args, **kwargs)
[rank0]: File "~/anaconda3/envs/psp/lib/python3.10/site-packages/vllm/model_executor/layers/vocab_parallel_embedding.py", line 406, in forward
[rank0]: output = tensor_model_parallel_all_reduce(output_parallel)
[rank0]: File "~/anaconda3/envs/psp/lib/python3.10/site-packages/vllm/distributed/communication_op.py", line 11, in tensor_model_parallel_all_reduce
[rank0]: return get_tp_group().all_reduce(input_)
[rank0]: File "~/anaconda3/envs/psp/lib/python3.10/site-packages/vllm/distributed/parallel_state.py", line 288, in all_reduce
[rank0]: pynccl_comm.all_reduce(input_)
[rank0]: File "~/anaconda3/envs/psp/lib/python3.10/site-packages/vllm/distributed/device_communicators/pynccl.py", line 113, in all_reduce
[rank0]: assert tensor.device == self.device, (
[rank0]: AttributeError: 'PyNcclCommunicator' object has no attribute 'device'
INFO 09-13 00:13:54 multiproc_worker_utils.py:123] Killing local vLLM worker processes
Fatal Python error: _enter_buffered_busy: could not acquire lock for <_io.BufferedWriter name='<stdout>'> at interpreter shutdown, possibly due to daemon threads
Python runtime state: finalizing (tstate=0x0000000000fadf60)
Current thread 0x00007facf08ba100 (most recent call first):
<no Python frame>
Extension modules: numpy.core._multiarray_umath, numpy.core._multiarray_tests, numpy.linalg._umath_linalg, numpy.fft._pocketfft_internal, numpy.random._common, numpy.random.bit_generator, numpy.random._bounded_integers, numpy.random._mt19937, numpy.random.mtrand, numpy.random._philox, numpy.random._pcg64, numpy.random._sfc64, numpy.random._generator, torch._C, torch._C._fft, torch._C._linalg, torch._C._nested, torch._C._nn, torch._C._sparse, torch._C._special, _brotli, yaml._yaml, msgspec._core, psutil._psutil_linux, psutil._psutil_posix, sentencepiece._sentencepiece, PIL._imaging, PIL._imagingft, gmpy2.gmpy2, regex._regex, msgpack._cmsgpack, google._upb._message, setproctitle, uvloop.loop, ray._raylet, multidict._multidict, yarl._helpers_c, yarl._quoting_c, aiohttp._helpers, aiohttp._http_writer, aiohttp._http_parser, aiohttp._websocket, frozenlist._frozenlist, zmq.backend.cython._zmq (total: 44)
~/anaconda3/envs/psp/lib/python3.10/multiprocessing/resource_tracker.py:224: UserWarning: resource_tracker: There appear to be 1 leaked shared_memory objects to clean up at shutdown
warnings.warn('resource_tracker: There appear to be %d '
Aborted (core dumped)
Before 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.
