[Bug][ROCm]: Process killed during tensor-parallel inference

[WoosukKwon]

Your current environment

I'm using a docker container built from Dockerfile.rocm with MI250x GPUs.

PyTorch version: 2.1.1+git011de5c
Is debug build: False
CUDA used to build PyTorch: N/A
ROCM used to build PyTorch: 6.0.32830-d62f6a171

OS: Ubuntu 20.04.6 LTS (x86_64)
GCC version: (Ubuntu 9.4.0-1ubuntu1~20.04.2) 9.4.0
Clang version: 17.0.0 (https://github.com/RadeonOpenCompute/llvm-project roc-6.0.0 23483 7208e8d15fbf218deb74483ea8c549c67ca4985e)
CMake version: version 3.29.1
Libc version: glibc-2.31

Python version: 3.9.18 (main, Sep 11 2023, 13:41:44)  [GCC 11.2.0] (64-bit runtime)
Python platform: Linux-5.19.0-45-generic-x86_64-with-glibc2.31
Is CUDA available: True
CUDA runtime version: 10.1.243
CUDA_MODULE_LOADING set to: LAZY
GPU models and configuration: AMD Instinct MI250X/MI250NoGCNArchNameOnOldPyTorch
Nvidia driver version: Could not collect
cuDNN version: Could not collect
HIP runtime version: 6.0.32830
MIOpen runtime version: 3.0.0
Is XNNPACK available: True

CPU:
Architecture:                    x86_64
CPU op-mode(s):                  32-bit, 64-bit
Byte Order:                      Little Endian
Address sizes:                   48 bits physical, 48 bits virtual
CPU(s):                          128
On-line CPU(s) list:             0-127
Thread(s) per core:              1
Core(s) per socket:              64
Socket(s):                       2
NUMA node(s):                    2
Vendor ID:                       AuthenticAMD
CPU family:                      25
Model:                           1
Model name:                      AMD EPYC 7713 64-Core Processor
Stepping:                        1
Frequency boost:                 enabled
CPU MHz:                         1500.000
CPU max MHz:                     3720.7029
CPU min MHz:                     1500.0000
BogoMIPS:                        3992.21
Virtualization:                  AMD-V
L1d cache:                       4 MiB
L1i cache:                       4 MiB
L2 cache:                        64 MiB
L3 cache:                        512 MiB
NUMA node0 CPU(s):               0-63
NUMA node1 CPU(s):               64-127
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: Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:        Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:        Mitigation; Retpolines, IBPB conditional, IBRS_FW, STIBP disabled, RSB filling, PBRSB-eIBRS Not affected
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 mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc rep_good nopl nonstop_tsc cpuid extd_apicid aperfmperf rapl pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_llc mwaitx cpb cat_l3 cdp_l3 invpcid_single hw_pstate ssbd mba ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 erms invpcid cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr rdpru wbnoinvd amd_ppin brs arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold v_vmsave_vmload vgif v_spec_ctrl umip pku ospke vaes vpclmulqdq rdpid overflow_recov succor smca fsrm

Versions of relevant libraries:
[pip3] mypy==1.4.1
[pip3] mypy-extensions==1.0.0
[pip3] numpy==1.22.4
[pip3] torch==2.1.1+git011de5c
[pip3] torchvision==0.16.1+fdea156
[pip3] triton==2.1.0
[conda] No relevant packagesROCM Version: 6.0.32830-d62f6a171
Neuron SDK Version: N/A
vLLM Version: 0.4.0.post1
vLLM Build Flags:
CUDA Archs: Not Set; ROCm: Disabled; Neuron: Disabled
GPU Topology:
Could not collect

🐛 Describe the bug

Got this error when running python benchmarks/benchmark_throughput.py --model meta-llama/Llama-2-7b-hf -tp 2 --dataset ShareGPT_Vicuna_unfiltered/resolve/main/ShareGPT_V3_unfiltered_cleaned_split.json. I tried multiple times, but always got the same error. I didn't get this error when using a single GPU.

Processed prompts:  80%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████▌                                                  | 799/1000 [04:55<00:17, 11.58it/s]
(RayWorkerVllm pid=12470) /var/lib/jenkins/pytorch/aten/src/ATen/native/hip/Indexing.hip:1294: indexSelectLargeIndex: Device-side assertion `srcIndex < srcSelectDimSize' failed.
(RayWorkerVllm pid=12470) /var/lib/jenkins/pytorch/aten/src/ATen/native/hip/Indexing.hip:1294: indexSelectLargeIndex: Device-side assertion `srcIndex < srcSelectDimSize' failed.

(RayWorkerVllm pid=12470) :0:rocdevice.cpp            :2726: 1292301652939 us: [pid:12470 tid:0x7eeb82ffd700] Callback: Queue 0x7eebf4200000 aborting with error : HSA_STATUS_ERROR_EXCEPTION: An HSAIL operation resulted in a hardware exception. code: 0x1016
(RayWorkerVllm pid=12470) *** SIGABRT received at time=1712868846 on cpu 107 ***
(RayWorkerVllm pid=12470) PC: @     0x7f1d4cad800b  (unknown)  raise
(RayWorkerVllm pid=12470)     @     0x7f1d4cdf5420  (unknown)  (unknown)
(RayWorkerVllm pid=12470) [2024-04-11 20:54:06,580 E 12470 16787] logging.cc:361: *** SIGABRT received at time=1712868846 on cpu 107 ***
(RayWorkerVllm pid=12470) [2024-04-11 20:54:06,580 E 12470 16787] logging.cc:361: PC: @     0x7f1d4cad800b  (unknown)  raise
(RayWorkerVllm pid=12470) [2024-04-11 20:54:06,580 E 12470 16787] logging.cc:361:     @     0x7f1d4cdf5420  (unknown)  (unknown)
(RayWorkerVllm pid=12470) Fatal Python error: Aborted
(RayWorkerVllm pid=12470) 

[hongxiayang]

@WoosukKwon This is a known issue when using non-eager mode. Please use --enforce-eager to run the benchmarking for now for tp > 1.

[sunway513]

From @WoosukKwon , --enforce-eager doesn't use hipgraph, it's functional but impact the perf.

[hongxiayang]

Closing this issue as the graph mode has been supported after ROCm 6.1 upgrade in current main branch.