Llama7b crash during runtime create

DX-3 report crash error when try to run llama 7B model. 3B model is work.

Below is log:
adb -s usbvid0e8dpid200ami0179d1041a00001 shell “cd /data/local/tmp/llm_sdk; LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$PWD:/vendor/lib64 ./main config_openllama7b.yaml -i sample_prompt.txt -m 100”
Using yaml config file: config_openllama7b.yaml
Reading prompt from file: sample_prompt.txt
Begin model init…
Using Neuron Runtime
Begin loading neuron runtime.
dlopen neuron_runtime
Load neuron OK
Found function: NeuronRuntime_create
Found function: NeuronRuntime_create_with_options
Found function: NeuronRuntime_loadNetworkFromFile
Found function: NeuronRuntime_setInput
Found function: NeuronRuntime_setOutput
Found function: NeuronRuntime_setOffsetedInput
Found function: NeuronRuntime_setQoSOption
Found function: NeuronRuntime_getInputSize
Found function: NeuronRuntime_getOutputSize
Found function: NeuronRuntime_getInputPaddedSize
Found function: NeuronRuntime_getOutputPaddedSize
Found function: NeuronRuntime_getInputPaddedDimensions
Found function: NeuronRuntime_getOutputPaddedDimensions
Found function: NeuronRuntime_getInputNumber
Found function: NeuronRuntime_getOutputNumber
Found function: NeuronRuntime_getProfiledQoSData
Found function: NeuronRuntime_inference
Found function: NeuronRuntime_release
Found function: NeuronRuntime_getVersion
Begin loading dmabuf library.
dlopen dmabufflib
Load dmabuf OK
Found function: FreeDmabufHeapBufferAllocator
Found function: DmabufHeapAlloc
Found function: CreateDmabufHeapBufferAllocator
Number of cache per dla: 32
Loading DLA 0
Loading DLA 1
Loading TFLite: /data/local/tmp/llm_sdk/openllama/embedding_open_llama_7b.tflite
initRuntimes
createRuntime0
initRuntimes(): Loaded single exclusive model (Total=1)
initModelIOInfo
gatherPreallocInputBuffers
preInitBufferProcess
initBuffer
INFO: Initialized TensorFlow Lite runtime.
verifyInit
initialize DONE
Initialized TFLite
startInitialize DLA
dlaExecutors DLA
setModelInput DLA
initialize DLA
initRuntimes
createRuntime1:/data/local/tmp/llm_sdk/openllama/open_llama_7b_sym4W_sym16A_Overall_hessian_128t1024c_0_extracted.dla
Segmentation fault