llama-cpp-ipex-llm-2.2.0-ubuntu-xeon NOT Support 4XARC770

[macafeeee]

Describe the bug I'm using the llama-cpp-ipex-llm-2.2.0-ubuntu-xeon on Ubuntu 24.04.02. The llama-cpp can't run and give me the error.

The program was built for 1 devices

Build program log for 'Intel(R) Arc(TM) A770 Graphics':

Exception caught at file:/home/intel/qiyuan/llama-cpp-bigdl/ggml/src/ggml-sycl/common.cpp, line:99

How to reproduce Steps to reproduce the error:

1. Install the new Ubuntu system.
2. Install the INTEL DRV according to the ipex-llm manual.
3. Using the deepseek r1 70B model.

Screenshots

Environment information 1.Mainboard: X12DPG-QT6 2.CPU:Xeon-6338*2 3.RAM:1TB DDR-4 3200MHZ 4.GPU:INTEL ARC 770 LIMITED EDITION

Additional context Ironically, the llama-cpp-ipex-llm-2.2.0b20250313-ubuntu-xeon run smoothly.

[cyita]

Hi macafeeee,

We can’t reproduce this issue. Since llama-cpp-ipex-llm-2.2.0b20250313-ubuntu-xeon runs smoothly, we recommend using this version.

[macafeeee]

Hi macafeeee,

We can’t reproduce this issue. Since llama-cpp-ipex-llm-2.2.0b20250313-ubuntu-xeon runs smoothly, we recommend using this version.

Try llama-server

[rnwang04]

Hi @macafeeee , we can only reproduce this error in very rare cases. Most of the time, we don't encounter this error, so we are still not sure what the root cause of this error. But in our latest nightly release (pip install --pre --upgrade ipex-llm[cpp]), we have some optimizations for server. Maybe you can try our latest release to see if this problem has been solved or improved. BTW, for latest nightly release, there is no portable zip, you need to install it by conda and maybe install OneAPI 2025.0.4 by yourself, related guide is here: https://github.com/intel/ipex-llm/blob/main/docs/mddocs/Quickstart/llama_cpp_quickstart.md .