Public » Linux

OneAPI i software¶

1. Instalacja bibliotek PokażUkryj

   mkdir /usr/src/neo/; \
   cd /usr/src/neo; \
   wget https://github.com/intel/intel-graphics-compiler/releases/download/v2.34.4/intel-igc-core-2_2.34.4+21428_amd64.deb; \
   wget https://github.com/intel/intel-graphics-compiler/releases/download/v2.34.4/intel-igc-opencl-2_2.34.4+21428_amd64.deb; \
   wget https://github.com/intel/compute-runtime/releases/download/26.18.38308.1/intel-ocloc-dbgsym_26.18.38308.1-0_amd64.ddeb; \
   wget https://github.com/intel/compute-runtime/releases/download/26.18.38308.1/intel-ocloc_26.18.38308.1-0_amd64.deb; \
   wget https://github.com/intel/compute-runtime/releases/download/26.18.38308.1/intel-opencl-icd-dbgsym_26.18.38308.1-0_amd64.ddeb; \
   wget https://github.com/intel/compute-runtime/releases/download/26.18.38308.1/intel-opencl-icd_26.18.38308.1-0_amd64.deb; \
   wget https://github.com/intel/compute-runtime/releases/download/26.18.38308.1/libigdgmm12_22.10.0_amd64.deb; \
   wget https://github.com/intel/compute-runtime/releases/download/26.18.38308.1/libze-intel-gpu1-dbgsym_26.18.38308.1-0_amd64.ddeb; \
   wget https://github.com/intel/compute-runtime/releases/download/26.18.38308.1/libze-intel-gpu1_26.18.38308.1-0_amd64.deb; \
    \
   apt-get install ./*
   

Intel Deep Learning Essentials¶

1. Instalacja PokażUkryj

   apt-get install -y gpg; \
   wget -O- https://apt.repos.intel.com/intel-gpg-keys/GPG-PUB-KEY-INTEL-SW-PRODUCTS.PUB | gpg --dearmor | tee /usr/share/keyrings/oneapi-archive-keyring.gpg > /dev/null; \
   echo "deb [signed-by=/usr/share/keyrings/oneapi-archive-keyring.gpg] https://apt.repos.intel.com/oneapi all main" | tee /etc/apt/sources.list.d/oneAPI.list; \
   apt-get update; \
   apt-get install intel-deep-learning-essentials;
   

2. Inicjalizacja PokażUkryj

   source /opt/intel/oneapi/setvars.sh;
   

llama.cpp + SYCL¶

1. Instalacja zależności(cmake itd.) PokażUkryj

   apt-get install -y cmake;
   

2. Kompilacja llama.cpp PokażUkryj

   git clone https://github.com/ggml-org/llama.cpp `date +%Y%m%d`; \
   cd `date +%Y%m%d`; \
   \
   cmake -B build \
     -DGGML_SYCL=ON \
     -DCMAKE_C_COMPILER=icx \
     -DCMAKE_CXX_COMPILER=icpx; \
   cmake --build build --config Release -j$(nproc);
   

Pobranie modeli i uruchomienie¶

1. Pobranie modeli PokażUkryj

   apt-get install -y pipx; \
    \
   pipx install huggingface_hub; \
   pipx ensurepath; \
   source ~/.bashrc; \
    \
   export HF_XET_HIGH_PERFORMANCE=1;
   
   hf download unsloth/Qwen3.6-27B-GGUF \
     Qwen3.6-27B-Q4_K_M.gguf \
     --local-dir ./models/qwen36-27b
   
   hf download unsloth/Qwen3.6-35B-A3B-MTP-GGUF Qwen3.6-35B-A3B-UD-Q4_K_M.gguf \
    --cache-dir ./models/Qwen3.6-35B-A3B-MTP-GGUF/
   
   

2. Uruchomienie PokażUkryj

   screen -S llm-b70 source /opt/intel/oneapi/setvars.sh; ONEAPI_DEVICE_SELECTOR="level_zero:0"; ./build/bin/llama-server \ 
     -m ./models/qwen36-27b/Qwen3.6-27B-Q4_K_M.gguf \
     --device SYCL0 -ngl 999   --ctx-size 196608   --jinja   --reasoning off   --host x.x.x.x --port 8080
   

Materiały dodatkowe:¶

• https://bibek-poudel.medium.com/how-to-run-qwen3-6-27b-locally-on-intel-arc-pro-b70-what-actually-works-c96dec67c6f7
• https://github.com/intel/compute-runtime/releases
• https://github.com/ggml-org/llama.cpp/releases
• https://www.intel.com/content/www/us/en/developer/tools/oneapi/oneapi-toolkit-download.html?packages=dl-essentials&dl-essentials-os=linux&dl-lin=apt
• https://www.intel.com/content/www/us/en/docs/oneapi-toolkit/installation-guide-linux/latest/install-deep-learning-essentials-with-apt.html#INSTALL-DEEP-LEARNING-ESSENTIALS-WITH-APT