-
Notifications
You must be signed in to change notification settings - Fork 167
[Bug]: Qwen3-8B crash with v0.7.2 on ascend 910C #959
Copy link
Copy link
Open
Labels
bugSomething isn't workingSomething isn't working
Description
Your environment
- inference Ascend 910C with ARM
- xLLM version:0.7.2
- startup parameters:
--port=8010
--max_memory_utilization=0.90
--block_size=128
--max_tokens_per_batch=30000
--max_memory_utilization=0.8
--max_seqs_per_batch=1000
--tool_call_parser=qwen3
--master_node_addr=127.0.0.1:13100
--nnodes=2
🐛 Describe the bug
- core dump:
Thread 129 (Thread 0xfffef20cae80 (LWP 10518)):
#0 0x0000ffff9f76cd84 in ?? () from /usr/lib64/libc.so.6
#1 0x0000ffff9f772944 in pthread_mutex_lock () from /usr/lib64/libc.so.6
#2 0x0000ffff7ffbd744 in ?? () from /usr/local/Ascend/ascend-toolkit/latest/lib64/libascend_trace.so
#3 0x0000ffff7ffbca48 in ?? () from /usr/local/Ascend/ascend-toolkit/latest/lib64/libascend_trace.so
#4 <signal handler called>
#5 0x0000ffff9f771620 in ?? () from /usr/lib64/libc.so.6
#6 0x0000ffff9f729f6c in raise () from /usr/lib64/libc.so.6
#7 0x0000ffff9f71703c in abort () from /usr/lib64/libc.so.6
#8 0x00000000006386ec in google::DumpStackTraceAndExit() ()
#9 0x000000000062dde4 in google::LogMessage::Fail() ()
#10 0x0000000000630f3c in google::LogMessage::SendToLog() ()
#11 0x000000000062d890 in google::LogMessage::Flush() ()
#12 0x000000000063151c in google::LogMessageFatal::~LogMessageFatal() ()
#13 0x0000000000b1b704 in xllm::Batch::process_sample_output(xllm::RawForwardOutput const&, bool) ()
#14 0x0000000000a9369c in xllm::LLMEngine::step(std::vector<xllm::Batch, std::allocator<xllm::Batch> >&) ()
#15 0x0000000000a5131c in xllm::ContinuousScheduler::step(absl::lts_20230802::Duration const&) ()
#16 0x00000000005e8a1c in std::thread::_State_impl<std::thread::_Invoker<std::tuple<xllm::LLMMaster::run()::{lambda()#1}> > >::_M_run() ()
#17 0x0000ffffaa3acd24 in std::execute_native_thread_routine (__p=0x93b17a90) at ../../../../../libstdc++-v3/src/nonshared11/../c++11/thread.cc:83
--Type <RET> for more, q to quit, c to continue without paging--
#18 0x0000ffff9f76fbd4 in ?? () from /usr/lib64/libc.so.6
#19 0x0000ffff9f7d7a8c in ?? () from /usr/lib64/libc.so.6
- accidental event, preduct issue
Reactions are currently unavailable
Metadata
Metadata
Assignees
Labels
bugSomething isn't workingSomething isn't working