vosk-server-gpu Segmentation fault (core dumped)

Hi! Can help me to identify what i'm doing wrong that after some transcriptions i got an
Segmentation fault (core dumped)

I sent audio chunks of 30 seconds to transcribe, one after the other... in some cases we split into multiple workers as you can see bellow, using 3 workers.

The path for the server is created dynamically to be uniq by chunk.

```
root@f66db52da9e0:/opt/vosk-server/websocket-gpu-batch# python3 ./asr_server_gpu.py
WARNING ([5.5.1089~1-a25f2]:SelectGpuId():cu-device.cc:243) Not in compute-exclusive mode. Suggestion: use 'nvidia-smi -c 3' to set compute exclusive mode
LOG ([5.5.1089~1-a25f2]:SelectGpuIdAuto():cu-device.cc:438) Selecting from 1 GPUs
LOG ([5.5.1089~1-a25f2]:SelectGpuIdAuto():cu-device.cc:453) cudaSetDevice(0): Tesla T4 free:14791M, used:118M, total:14910M, free/total:0.992023
LOG ([5.5.1089~1-a25f2]:SelectGpuIdAuto():cu-device.cc:501) Device: 0, mem_ratio: 0.992023
LOG ([5.5.1089~1-a25f2]:SelectGpuId():cu-device.cc:382) Trying to select device: 0
LOG ([5.5.1089~1-a25f2]:SelectGpuIdAuto():cu-device.cc:511) Success selecting device 0 free mem ratio: 0.992023
LOG ([5.5.1089~1-a25f2]:FinalizeActiveGpu():cu-device.cc:338) The active GPU is [0]: Tesla T4 free:14455M, used:454M, total:14910M, free/total:0.969489 version 7.5
LOG ([5.5.1089~1-a25f2]:RemoveOrphanNodes():nnet-nnet.cc:948) Removed 0 orphan nodes.
LOG ([5.5.1089~1-a25f2]:RemoveOrphanComponents():nnet-nnet.cc:847) Removing 0 orphan components.
LOG ([5.5.1089~1-a25f2]:BatchModel():batch_model.cc:52) Loading HCLG from model/graph/HCLG.fst
LOG ([5.5.1089~1-a25f2]:BatchModel():batch_model.cc:56) Loading words from model/graph/words.txt
LOG ([5.5.1089~1-a25f2]:BatchModel():batch_model.cc:64) Loading winfo model/graph/phones/word_boundary.int
LOG ([5.5.1089~1-a25f2]:ComputeDerivedVars():ivector-extractor.cc:183) Computing derived variables for iVector extractor
LOG ([5.5.1089~1-a25f2]:ComputeDerivedVars():ivector-extractor.cc:204) Done.
server listening on 0.0.0.0:2700
INFO:websockets.server:server listening on 0.0.0.0:2700
connection open
INFO:websockets.server:connection open
INFO:root:Connection from ('34.30.88.55', 49442)
INFO:root:Config {'words': 1, 'sample_rate': 16000}
connection closed
INFO:websockets.server:connection closed
connection open
INFO:websockets.server:connection open
INFO:root:Connection from ('34.30.88.55', 49456)
INFO:root:Config {'words': 1, 'sample_rate': 16000}
connection open
INFO:websockets.server:connection open
INFO:root:Connection from ('35.224.62.142', 60480)
INFO:root:Config {'words': 1, 'sample_rate': 16000}
connection open
INFO:websockets.server:connection open
INFO:root:Connection from ('34.134.42.203', 40700)
INFO:root:Config {'words': 1, 'sample_rate': 16000}
connection closed
INFO:websockets.server:connection closed
Segmentation fault (core dumped)
```

```
root@f66db52da9e0:/opt/vosk-server/websocket-gpu-batch# gdb python3 core

Type "apropos word" to search for commands related to "word"...
Reading symbols from python3...
Reading symbols from /usr/lib/debug/.build-id/14/8e086667839ef13939196984d6f717c331bd76.debug...

warning: Can't open file /dev/zero (deleted) during file-backed mapping note processing
[New LWP 2564]
[New LWP 2563]
[New LWP 2568]
[New LWP 2046]
[New LWP 2049]
[New LWP 2562]
[New LWP 2565]
[New LWP 2566]
[New LWP 2045]
[New LWP 2567]
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
Core was generated by `python3 ./asr_server_gpu.py'.
Program terminated with signal SIGSEGV, Segmentation fault.
#0 0x00007fc915ea7b36 in BatchRecognizer::PushLattice(fst::VectorFst, int> >, fst::VectorState, int> >, std::allocator, int> > > > >&, float) () from /usr/local/lib/python3.10/dist-packages/vosk-0.3.45-py3.10.egg/vosk/libvosk.so
[Current thread is 1 (Thread 0x7fc67e5d2640 (LWP 2564))]
(gdb)
```

```
(gdb) bt
#0 0x00007fc915ea7b36 in BatchRecognizer::PushLattice(fst::VectorFst, int> >, fst::VectorState, int> >, std::allocator, int> > > > >&, float) () from /usr/local/lib/python3.10/dist-packages/vosk-0.3.45-py3.10.egg/vosk/libvosk.so
#1 0x00007fc915eb9f81 in kaldi::cuda_decoder::BatchedThreadedNnet3CudaOnlinePipeline::FinalizeDecoding(int) () from /usr/local/lib/python3.10/dist-packages/vosk-0.3.45-py3.10.egg/vosk/libvosk.so
#2 0x00007fc915eae5a5 in kaldi::cuda_decoder::ThreadPoolLightWorker::Work() () from /usr/local/lib/python3.10/dist-packages/vosk-0.3.45-py3.10.egg/vosk/libvosk.so
#3 0x00007fc8d8cb22b3 in ?? () from /lib/x86_64-linux-gnu/libstdc++.so.6
#4 0x00007fc91ab2eb43 in start_thread (arg=) at ./nptl/pthread_create.c:442
#5 0x00007fc91abbfbb4 in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:100
```