search

intel-analytics / ipex-llm

Accelerate local LLM inference and finetuning (LLaMA, Mistral, ChatGLM, Qwen, Mixtral, Gemma, Phi, MiniCPM, Qwen-VL, MiniCPM-V, etc.) on Intel XPU (e.g., local PC with iGPU and NPU, discrete GPU such as Arc, Flex and Max); seamlessly integrate with llama.cpp, Ollama, HuggingFace, LangChain, LlamaIndex, vLLM, GraphRAG, DeepSpeed, Axolotl, etc
Apache License 2.0
6.65k stars 1.26k forks source link

Segmentation fault (core dumped) while inferencing with MTL iGPU #11001

Closed eugeooi closed 5 months ago

eugeooi commented 5 months ago

After following the instructions: Install IPEX-LLM on Linux with Intel GPU, and executed the code Llama3, problem occurs: LIBXSMM_VERSION: main_stable-1.17-3651 (25693763) LIBXSMM_TARGET: adl [Intel(R) Core(TM) Ultra 7 165HL] Registry and code: 13 MB Command: python generate.py Uptime: 20.336332 s Aborted (core dumped)

Environments: Intel(R) Core(TM) Ultra 7 165HL Ubuntu 22.04.4 64GB DDR5x5600MHz ipex-llm[xpu]: 2.1.0b20240511 intel-extension-for-pytorch: 2.1.10+xpu oneAPI version 2024.0

Oscilloscope98 commented 5 months ago

Hi @eugeooi,

Here is a similar issue: https://github.com/intel-analytics/ipex-llm/issues/10936, maybe you could refer to here for more information :)

Please let us know if there are further problems

eugeooi commented 5 months ago

I have uninstalled NPU driver using this command: dpkg --purge --force-remove-reinstreq intel-driver-compiler-npu intel-fw-npu intel-level-zero-npu level-zero

But still failed to run generate.py with another errors shown: OSError: libmkl_intel_lp64.so.2: cannot open shared object file: No such file or directory

Oscilloscope98 commented 5 months ago

Hi @eugeooi,

It seems that some level zero packages were also accidentally uninstalled.

To uninstall NPU related packages, you could just uninstalled the ones listed by dpkg -l which includes -npu.

Please let us know for any further problems :)

eugeooi commented 5 months ago

Able to run now. Thanks!

sunshine19870316 commented 3 months ago

How did you solve this issue?