Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

[Bug]: Llama 3.2 90b crash #10648

Open
1 task done
yessenzhar opened this issue Nov 26, 2024 · 10 comments
Open
1 task done

[Bug]: Llama 3.2 90b crash #10648

yessenzhar opened this issue Nov 26, 2024 · 10 comments
Labels
bug Something isn't working

Comments

@yessenzhar
Copy link

Your current environment

The output of `python collect_env.py` 
Collecting environment information...
PyTorch version: 2.5.0+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: Could not collect
Libc version: glibc-2.35

Python version: 3.12.7 (main, Oct  1 2024, 08:52:12) [GCC 11.4.0] (64-bit runtime)
Python platform: Linux-5.15.0-112-generic-x86_64-with-glibc2.35
Is CUDA available: False
CUDA runtime version: No CUDA
CUDA_MODULE_LOADING set to: N/A
GPU models and configuration: No CUDA
Nvidia driver version: No CUDA
cuDNN version: No CUDA
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
Address sizes:                      46 bits physical, 57 bits virtual
Byte Order:                         Little Endian
CPU(s):                             224
On-line CPU(s) list:                0-223
Vendor ID:                          GenuineIntel
Model name:                         Intel(R) Xeon(R) Platinum 8480+
CPU family:                         6
Model:                              143
Thread(s) per core:                 2
Core(s) per socket:                 56
Socket(s):                          2
Stepping:                           8
BogoMIPS:                           4000.00
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 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 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
L1d cache:                          5.3 MiB (112 instances)
L1i cache:                          3.5 MiB (112 instances)
L2 cache:                           224 MiB (112 instances)
L3 cache:                           210 MiB (2 instances)
NUMA node(s):                       2
NUMA node0 CPU(s):                  0,2,4,6,8,10,12,14,16,18,20,22,24,26,28,30,32,34,36,38,40,42,44,46,48,50,52,54,56,58,60,62,64,66,68,70,72,74,76,78,80,82,84,86,88,90,92,94,96,98,100,102,104,106,108,110,112,114,116,118,120,122,124,126,128,130,132,134,136,138,140,142,144,146,148,150,152,154,156,158,160,162,164,166,168,170,172,174,176,178,180,182,184,186,188,190,192,194,196,198,200,202,204,206,208,210,212,214,216,218,220,222
NUMA node1 CPU(s):                  1,3,5,7,9,11,13,15,17,19,21,23,25,27,29,31,33,35,37,39,41,43,45,47,49,51,53,55,57,59,61,63,65,67,69,71,73,75,77,79,81,83,85,87,89,91,93,95,97,99,101,103,105,107,109,111,113,115,117,119,121,123,125,127,129,131,133,135,137,139,141,143,145,147,149,151,153,155,157,159,161,163,165,167,169,171,173,175,177,179,181,183,185,187,189,191,193,195,197,199,201,203,205,207,209,211,213,215,217,219,221,223
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; BHI BHI_DIS_S
Vulnerability Srbds:                Not affected
Vulnerability Tsx async abort:      Not affected

Versions of relevant libraries:
[pip3] flashinfer==0.1.6+cu121torch2.4
[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-ml-py==12.560.30
[pip3] nvidia-nccl-cu12==2.21.5
[pip3] nvidia-nvjitlink-cu12==12.4.127
[pip3] nvidia-nvtx-cu12==12.4.127
[pip3] pyzmq==26.2.0
[pip3] torch==2.5.0
[pip3] torchvision==0.20.0
[pip3] transformers==4.46.1
[pip3] triton==3.1.0
[conda] Could not collect
ROCM Version: Could not collect
Neuron SDK Version: N/A
vLLM Version: 0.1.dev3243+g52666ce.d20241104 (git sha: 52666ce.d20241104
vLLM Build Flags:
CUDA Archs: Not Set; ROCm: Disabled; Neuron: Disabled
GPU Topology:
Could not collect

NVIDIA_VISIBLE_DEVICES=GPU-8c14074c-9922-6592-0c14-eec5bb4463e5,GPU-2492c185-f06f-4e97-4063-72067f27ee7b,GPU-882e27e1-4752-f6a0-bfc0-3e55ce274b8a,GPU-b49f0503-f439-921d-8124-e6029f30b83b
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
NVIDIA_DRIVER_CAPABILITIES=compute,utility
VLLM_USAGE_SOURCE=production-docker-image
CUDA_VERSION=12.4.1
LD_LIBRARY_PATH=/usr/local/lib/python3.12/dist-packages/cv2/../../lib64:/usr/local/nvidia/lib:/usr/local/nvidia/lib64

Model Input Dumps

No response

🐛 Describe the bug

When running 90b LLama 3.2 vision model using vllm, see command below:
python3 -m vllm.entrypoints.openai.api_server --served-model-name=meta-llama/Llama-3.2-90B-Vision-Instruct --model=/data/001 --tensor-parallel-size=4 --max-num-seqs=96 --max-log-len=0 --load-format=safetensors --host=0.0.0.0 --port=80 --max-num-seqs=64 --gpu-memory-utilization=0.95 --enforce-eager --max-model-len=32768

It occasionally crashes, see logs

INFO 11-25 16:02:04 logger.py:37] Received request cmpl-6ccba6262268459ab8a7317db4e6a4dd-0: prompt: '', params: SamplingParams(n=1, presence_penalty=0.0, frequency_penalty=0.0, repetition_penalty=1.0, temperature=0.7, top_p=0.9, top_k=10, min_p=0.0, seed=3497136763981421277, stop=['<|eot_id|>', '<|end_of_text|>', '<|eom_id|>'], stop_token_ids=[], bad_words=[], include_stop_str_in_output=False, ignore_eos=False, max_tokens=1, min_tokens=0, logprobs=None, prompt_logprobs=None, skip_special_tokens=True, spaces_between_special_tokens=True, truncate_prompt_tokens=None, guided_decoding=None), prompt_token_ids: [], lora_request: None, prompt_adapter_request: None.
INFO 11-25 16:02:04 logger.py:37] Received request chatcmpl-0f4e61f625404415ae19e6c65a805c48: prompt: '', params: SamplingParams(n=1, presence_penalty=0.0, frequency_penalty=0.0, repetition_penalty=1.0, temperature=0.3, top_p=1.0, top_k=-1, min_p=0.0, seed=None, stop=['<|eot_id|>', '<|end_of_text|>', '<|eom_id|>'], stop_token_ids=[], bad_words=[], include_stop_str_in_output=False, ignore_eos=False, max_tokens=750, min_tokens=0, logprobs=None, prompt_logprobs=None, skip_special_tokens=True, spaces_between_special_tokens=True, truncate_prompt_tokens=None, guided_decoding=None), prompt_token_ids: None, lora_request: None, prompt_adapter_request: None.
INFO 11-25 16:02:05 engine.py:275] Added request cmpl-6ccba6262268459ab8a7317db4e6a4dd-0.
INFO 11-25 16:02:05 engine.py:275] Added request chatcmpl-0f4e61f625404415ae19e6c65a805c48.
INFO 11-25 16:02:05 logger.py:37] Received request chatcmpl-8599dad45f1d4e808952b46daef59e09: prompt: '', params: SamplingParams(n=1, presence_penalty=0.0, frequency_penalty=0.0, repetition_penalty=1.0, temperature=0.3, top_p=1.0, top_k=-1, min_p=0.0, seed=None, stop=['<|eot_id|>', '<|end_of_text|>', '<|eom_id|>'], stop_token_ids=[], bad_words=[], include_stop_str_in_output=False, ignore_eos=False, max_tokens=750, min_tokens=0, logprobs=None, prompt_logprobs=None, skip_special_tokens=True, spaces_between_special_tokens=True, truncate_prompt_tokens=None, guided_decoding=None), prompt_token_ids: None, lora_request: None, prompt_adapter_request: None.
CRITICAL 11-25 16:02:05 launcher.py:99] MQLLMEngine is already dead, terminating server process
INFO: 10.244.54.175:43202 - "POST /v1/chat/completions HTTP/1.1" 500 Internal Server Error
CRITICAL 11-25 16:02:05 launcher.py:99] MQLLMEngine is already dead, terminating server process
INFO: 10.244.54.169:53022 - "POST /v1/chat/completions HTTP/1.1" 500 Internal Server Error
CRITICAL 11-25 16:02:05 launcher.py:99] MQLLMEngine is already dead, terminating server process
INFO: 10.244.51.218:57838 - "POST /v1/chat/completions HTTP/1.1" 500 Internal Server Error
CRITICAL 11-25 16:02:05 launcher.py:99] MQLLMEngine is already dead, terminating server process
INFO: 10.244.53.240:41778 - "POST /v1/chat/completions HTTP/1.1" 500 Internal Server Error
CRITICAL 11-25 16:02:05 launcher.py:99] MQLLMEngine is already dead, terminating server process
INFO: 10.244.10.36:41008 - "POST /v1/completions HTTP/1.1" 500 Internal Server Error
CRITICAL 11-25 16:02:05 launcher.py:99] MQLLMEngine is already dead, terminating server process
INFO: 10.244.54.171:36942 - "POST /v1/chat/completions HTTP/1.1" 500 Internal Server Error
CRITICAL 11-25 16:02:05 launcher.py:99] MQLLMEngine is already dead, terminating server process
INFO: 10.244.51.220:60982 - "POST /v1/chat/completions HTTP/1.1" 500 Internal Server Error
ERROR 11-25 16:02:05 engine.py:143] RuntimeError('CUDA error: invalid configuration argument\nCUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.\nFor debugging consider passing CUDA_LAUNCH_BLOCKING=1\nCompile with TORCH_USE_CUDA_DSA to enable device-side assertions.\n')
ERROR 11-25 16:02:05 engine.py:143] Traceback (most recent call last):
ERROR 11-25 16:02:05 engine.py:143] File "/usr/local/lib/python3.12/dist-packages/vllm/engine/multiprocessing/engine.py", line 141, in start
ERROR 11-25 16:02:05 engine.py:143] self.run_engine_loop()
ERROR 11-25 16:02:05 engine.py:143] File "/usr/local/lib/python3.12/dist-packages/vllm/engine/multiprocessing/engine.py", line 204, in run_engine_loop
ERROR 11-25 16:02:05 engine.py:143] request_outputs = self.engine_step()
ERROR 11-25 16:02:05 engine.py:143] ^^^^^^^^^^^^^^^^^^
ERROR 11-25 16:02:05 engine.py:143] File "/usr/local/lib/python3.12/dist-packages/vllm/engine/multiprocessing/engine.py", line 222, in engine_step
ERROR 11-25 16:02:05 engine.py:143] raise e
ERROR 11-25 16:02:05 engine.py:143] File "/usr/local/lib/python3.12/dist-packages/vllm/engine/multiprocessing/engine.py", line 213, in engine_step
ERROR 11-25 16:02:05 engine.py:143] return self.engine.step()
ERROR 11-25 16:02:05 engine.py:143] ^^^^^^^^^^^^^^^^^^
ERROR 11-25 16:02:05 engine.py:143] File "/usr/local/lib/python3.12/dist-packages/vllm/engine/llm_engine.py", line 1456, in step
ERROR 11-25 16:02:05 engine.py:143] outputs = self.model_executor.execute_model(
ERROR 11-25 16:02:05 engine.py:143] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ERROR 11-25 16:02:05 engine.py:143] File "/usr/local/lib/python3.12/dist-packages/vllm/executor/distributed_gpu_executor.py", line 82, in execute_model
ERROR 11-25 16:02:05 engine.py:143] driver_outputs = self._driver_execute_model(execute_model_req)
ERROR 11-25 16:02:05 engine.py:143] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ERROR 11-25 16:02:05 engine.py:143] File "/usr/local/lib/python3.12/dist-packages/vllm/executor/multiproc_gpu_executor.py", line 155, in _driver_execute_model
ERROR 11-25 16:02:05 engine.py:143] return self.driver_worker.execute_model(execute_model_req)
ERROR 11-25 16:02:05 engine.py:143] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ERROR 11-25 16:02:05 engine.py:143] File "/usr/local/lib/python3.12/dist-packages/vllm/worker/worker_base.py", line 343, in execute_model
ERROR 11-25 16:02:05 engine.py:143] output = self.model_runner.execute_model(
ERROR 11-25 16:02:05 engine.py:143] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ERROR 11-25 16:02:05 engine.py:143] File "/usr/local/lib/python3.12/dist-packages/torch/utils/_contextlib.py", line 116, in decorate_context
ERROR 11-25 16:02:05 engine.py:143] return func(*args, **kwargs)
ERROR 11-25 16:02:05 engine.py:143] ^^^^^^^^^^^^^^^^^^^^^
ERROR 11-25 16:02:05 engine.py:143] File "/usr/local/lib/python3.12/dist-packages/vllm/worker/enc_dec_model_runner.py", line 201, in execute_model
ERROR 11-25 16:02:05 engine.py:143] hidden_or_intermediate_states = model_executable(
ERROR 11-25 16:02:05 engine.py:143] ^^^^^^^^^^^^^^^^^
ERROR 11-25 16:02:05 engine.py:143] File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1736, in _wrapped_call_impl
ERROR 11-25 16:02:05 engine.py:143] return self._call_impl(*args, **kwargs)
ERROR 11-25 16:02:05 engine.py:143] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ERROR 11-25 16:02:05 engine.py:143] File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1747, in _call_impl
ERROR 11-25 16:02:05 engine.py:143] return forward_call(*args, **kwargs)
ERROR 11-25 16:02:05 engine.py:143] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ERROR 11-25 16:02:05 engine.py:143] File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/mllama.py", line 1363, in forward
ERROR 11-25 16:02:05 engine.py:143] outputs = self.language_model(
ERROR 11-25 16:02:05 engine.py:143] ^^^^^^^^^^^^^^^^^^^^
ERROR 11-25 16:02:05 engine.py:143] File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1736, in _wrapped_call_impl
ERROR 11-25 16:02:05 engine.py:143] return self._call_impl(*args, **kwargs)
ERROR 11-25 16:02:05 engine.py:143] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ERROR 11-25 16:02:05 engine.py:143] File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1747, in _call_impl
ERROR 11-25 16:02:05 engine.py:143] return forward_call(*args, **kwargs)
ERROR 11-25 16:02:05 engine.py:143] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ERROR 11-25 16:02:05 engine.py:143] File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/mllama.py", line 1030, in forward
ERROR 11-25 16:02:05 engine.py:143] hidden_states = self.model(
ERROR 11-25 16:02:05 engine.py:143] ^^^^^^^^^^^
ERROR 11-25 16:02:05 engine.py:143] File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1736, in _wrapped_call_impl
ERROR 11-25 16:02:05 engine.py:143] return self._call_impl(*args, **kwargs)
ERROR 11-25 16:02:05 engine.py:143] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ERROR 11-25 16:02:05 engine.py:143] File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1747, in _call_impl
ERROR 11-25 16:02:05 engine.py:143] return forward_call(*args, **kwargs)
ERROR 11-25 16:02:05 engine.py:143] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ERROR 11-25 16:02:05 engine.py:143] File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/mllama.py", line 962, in forward
ERROR 11-25 16:02:05 engine.py:143] hidden_states = decoder_layer(
ERROR 11-25 16:02:05 engine.py:143] ^^^^^^^^^^^^^^
ERROR 11-25 16:02:05 engine.py:143] File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1736, in _wrapped_call_impl
ERROR 11-25 16:02:05 engine.py:143] return self._call_impl(*args, **kwargs)
ERROR 11-25 16:02:05 engine.py:143] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ERROR 11-25 16:02:05 engine.py:143] File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1747, in _call_impl
ERROR 11-25 16:02:05 engine.py:143] return forward_call(*args, **kwargs)
ERROR 11-25 16:02:05 engine.py:143] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ERROR 11-25 16:02:05 engine.py:143] File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/mllama.py", line 880, in forward
ERROR 11-25 16:02:05 engine.py:143] hidden_states = self.cross_attn(
ERROR 11-25 16:02:05 engine.py:143] ^^^^^^^^^^^^^^^^
ERROR 11-25 16:02:05 engine.py:143] File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1736, in _wrapped_call_impl
ERROR 11-25 16:02:05 engine.py:143] return self._call_impl(*args, **kwargs)
ERROR 11-25 16:02:05 engine.py:143] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ERROR 11-25 16:02:05 engine.py:143] File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1747, in _call_impl
ERROR 11-25 16:02:05 engine.py:143] return forward_call(*args, **kwargs)
ERROR 11-25 16:02:05 engine.py:143] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ERROR 11-25 16:02:05 engine.py:143] File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/mllama.py", line 764, in forward
ERROR 11-25 16:02:05 engine.py:143] output = self.attention_with_mask(q, k, v, kv_cache,
ERROR 11-25 16:02:05 engine.py:143] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ERROR 11-25 16:02:05 engine.py:143] File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/mllama.py", line 807, in attention_with_mask
ERROR 11-25 16:02:05 engine.py:143] self.head_dim).contiguous()
ERROR 11-25 16:02:05 engine.py:143] ^^^^^^^^^^^^
ERROR 11-25 16:02:05 engine.py:143] RuntimeError: CUDA error: invalid configuration argument
ERROR 11-25 16:02:05 engine.py:143] CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
ERROR 11-25 16:02:05 engine.py:143] For debugging consider passing CUDA_LAUNCH_BLOCKING=1
ERROR 11-25 16:02:05 engine.py:143] Compile with TORCH_USE_CUDA_DSA to enable device-side assertions.
ERROR 11-25 16:02:05 engine.py:143]
�[1;36m(VllmWorkerProcess pid=352)�[0;0m ERROR 11-25 16:02:05 multiproc_worker_utils.py:229] Exception in worker VllmWorkerProcess while processing method start_worker_execution_loop.
�[1;36m(VllmWorkerProcess pid=352)�[0;0m ERROR 11-25 16:02:05 multiproc_worker_utils.py:229] Traceback (most recent call last):
�[1;36m(VllmWorkerProcess pid=352)�[0;0m ERROR 11-25 16:02:05 multiproc_worker_utils.py:229] File "/usr/local/lib/python3.12/dist-packages/vllm/executor/multiproc_worker_utils.py", line 223, in _run_worker_process
�[1;36m(VllmWorkerProcess pid=352)�[0;0m ERROR 11-25 16:02:05 multiproc_worker_utils.py:229] output = executor(*args, **kwargs)
�[1;36m(VllmWorkerProcess pid=352)�[0;0m ERROR 11-25 16:02:05 multiproc_worker_utils.py:229] ^^^^^^^^^^^^^^^^^^^^^^^^^
�[1;36m(VllmWorkerProcess pid=352)�[0;0m ERROR 11-25 16:02:05 multiproc_worker_utils.py:229] File "/usr/local/lib/python3.12/dist-packages/torch/utils/_contextlib.py", line 116, in decorate_context
�[1;36m(VllmWorkerProcess pid=352)�[0;0m ERROR 11-25 16:02:05 multiproc_worker_utils.py:229] return func(*args, **kwargs)
�[1;36m(VllmWorkerProcess pid=352)�[0;0m ERROR 11-25 16:02:05 multiproc_worker_utils.py:229] ^^^^^^^^^^^^^^^^^^^^^
�[1;36m(VllmWorkerProcess pid=352)�[0;0m ERROR 11-25 16:02:05 multiproc_worker_utils.py:229] File "/usr/local/lib/python3.12/dist-packages/vllm/worker/worker_base.py", line 85, in start_worker_execution_loop
�[1;36m(VllmWorkerProcess pid=352)�[0;0m ERROR 11-25 16:02:05 multiproc_worker_utils.py:229] output = self.execute_model(execute_model_req=None)
�[1;36m(VllmWorkerProcess pid=352)�[0;0m ERROR 11-25 16:02:05 multiproc_worker_utils.py:229] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
�[1;36m(VllmWorkerProcess pid=352)�[0;0m ERROR 11-25 16:02:05 multiproc_worker_utils.py:229] File "/usr/local/lib/python3.12/dist-packages/vllm/worker/worker_base.py", line 343, in execute_model
�[1;36m(VllmWorkerProcess pid=352)�[0;0m ERROR 11-25 16:02:05 multiproc_worker_utils.py:229] output = self.model_runner.execute_model(
�[1;36m(VllmWorkerProcess pid=352)�[0;0m ERROR 11-25 16:02:05 multiproc_worker_utils.py:229] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
�[1;36m(VllmWorkerProcess pid=352)�[0;0m ERROR 11-25 16:02:05 multiproc_worker_utils.py:229] File "/usr/local/lib/python3.12/dist-packages/torch/utils/_contextlib.py", line 116, in decorate_context
�[1;36m(VllmWorkerProcess pid=352)�[0;0m ERROR 11-25 16:02:05 multiproc_worker_utils.py:229] return func(*args, **kwargs)
�[1;36m(VllmWorkerProcess pid=352)�[0;0m ERROR 11-25 16:02:05 multiproc_worker_utils.py:229] ^^^^^^^^^^^^^^^^^^^^^
�[1;36m(VllmWorkerProcess pid=352)�[0;0m ERROR 11-25 16:02:05 multiproc_worker_utils.py:229] File "/usr/local/lib/python3.12/dist-packages/vllm/worker/enc_dec_model_runner.py", line 201, in execute_model
�[1;36m(VllmWorkerProcess pid=352)�[0;0m ERROR 11-25 16:02:05 multiproc_worker_utils.py:229] hidden_or_intermediate_states = model_executable(
�[1;36m(VllmWorkerProcess pid=352)�[0;0m ERROR 11-25 16:02:05 multiproc_worker_utils.py:229] ^^^^^^^^^^^^^^^^^
�[1;36m(VllmWorkerProcess pid=352)�[0;0m ERROR 11-25 16:02:05 multiproc_worker_utils.py:229] File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1736, in _wrapped_call_impl
�[1;36m(VllmWorkerProcess pid=352)�[0;0m ERROR 11-25 16:02:05 multiproc_worker_utils.py:229] return self._call_impl(*args, **kwargs)
�[1;36m(VllmWorkerProcess pid=352)�[0;0m ERROR 11-25 16:02:05 multiproc_worker_utils.py:229] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
�[1;36m(VllmWorkerProcess pid=352)�[0;0m ERROR 11-25 16:02:05 multiproc_worker_utils.py:229] File "/usr/local/lib/python3.12/dist-packages/torch/nn/modules/module.py", line 1747, in _call_impl
�[1;36m(VllmWorkerProcess pid=352)�[0;0m ERROR 11-25 16:02:05 multiproc_worker_utils.py:229] return forward_call(*args, **kwargs)
...

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.
@yessenzhar yessenzhar added the bug Something isn't working label Nov 26, 2024
@DarkLight1337
Copy link
Member

cc @heheda12345

@debraj135
Copy link

Is multi-lora supported for this model or for this meta-llama/Llama-3.2-11B-Vision-Instruct?

@DarkLight1337
Copy link
Member

This model doesn't support LoRA. Please check the Supported Models page for more details for each model.

@vlasenkoalexey
Copy link

Happens under model load for me as well, we don't use LoRA for this deployment, therefore issue is not LoRA related.

@heheda12345
Copy link
Collaborator

Can you share the log with CUDA_LAUNCH_BLOCKING=1?

@tjohnson31415
Copy link
Contributor

tjohnson31415 commented Jan 10, 2025 •
edited
Loading

@wallashss and I have found a way to reproduce this crash by processing a request with more (groups of) <|image|> tokens than images. When running the OpenAI server, this can be triggered if the user message includes the <|image|> special token that is non-adjacent to the image message, eg.:

curl http://localhost:8000/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "meta-llama/Llama-3.2-11B-Vision",
    "messages": [
      {
        "role": "user",
        "content": [
          {"type": "text", "text": " <|image|> "},
          {"type": "image_url", "image_url": {"url": "https://upload.wikimedia.org/wikipedia/commons/d/da/2015_Kaczka_krzy%C5%BCowka_w_wodzie_%28samiec%29.jpg"}}
        ]
      }
    ]
  }'

(NB: this crashes it with both the 90B and 11B model)

@heheda12345
Copy link
Collaborator

@ywang96 @DarkLight1337 @Isotr0py Any suggestions on rejecting the requests with different number of image tokens and input images instead of making the server crash?

@tjohnson31415 tjohnson31415 mentioned this issue Jan 10, 2025
Merged
@tjohnson31415
Copy link
Contributor

tjohnson31415 commented Jan 17, 2025 •
edited
Loading

An update:
we deployed vLLM internally with my bugfix PR (#11939) included, but we are still seeing the server crash. With my patch, it now crashes with an assert ts != -1 AssertionError instead of a CUDA error, so there must be another way to cause a crash even with matching counts of image tokens and input images 🤔

@yessenzhar
Copy link
Author

We see that in production too, even with checks that there are not two <|image|> tags in the array. From what we can tell it's happening under heavy load.

@tjohnson31415 tjohnson31415 mentioned this issue Jan 23, 2025
Merged
@tjohnson31415
Copy link
Contributor

@wallashss and I got another bug fix merged in for another way to trigger the crash: #12347

We still need to deploy to see if we get crashes in prouction, but hopefully this is the last fix we need for this issue.

cc: @yessenzhar in case you are able to test it as well

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
bug Something isn't working
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
6 participants
@vlasenkoalexey @tjohnson31415 @yessenzhar @debraj135 @heheda12345 @DarkLight1337