Verified Hardware Platforms:
- Intel® Data Center GPU Max Series, Driver Version: 540
- Intel® Data Center GPU Flex Series 170, Driver Version: 540
- Experimental: Intel® Arc™ A-Series
For experimental support of the Intel® Arc™ A-Series GPUs, please refer to Intel® Arc™ A-Series GPU Software Installation for details.
- Ubuntu 22.04, RedHat 8.6 (64-bit)
- Intel® Data Center GPU Flex Series
- SUSE Linux Enterprise Server(SLES) 15 SP3/SP4, RedHat 8.6 (64-bit)
- Intel® Data Center GPU Max Series
- Intel® oneAPI Base Toolkit 2023.0
- TensorFlow 2.11.0 or 2.10.0
- Python 3.7-3.10
- pip 19.0 or later (requires manylinux2014 support)
Intel ® Extension for TensorFlow* | Stock Tensorflow | OS | oneAPI toolkit | Driver |
---|---|---|---|---|
v1.1.0 | 2.10 & 2.11 | Ubuntu 22.04 & RHEL 8.6 & SLES 15SP3 & SLES 15SP4 | 2023.0 | 540 |
v1.0.0 | 2.10 | Ubuntu 20.04 & RHEL 8.5 | 2022.3 | 419.4 |
Release | OS | Intel GPU | Install Intel GPU Driver |
---|---|---|---|
v1.1.0 | Ubuntu 22.04, RedHat 8.6 | Intel® Data Center GPU Flex Series | Refer to the Installation Guides for latest driver installation. If install the verified Intel® Data Center GPU Max Series/Intel® Data Center GPU Flex Series 540, please append the specific version after components, such as sudo apt-get install intel-opencl-icd=22.43.24595.35+i538~22.04 |
v1.1.0 | SLES 15 SP3/SP4, RedHat 8.6 | Intel® Data Center GPU Max Series | Refer to the Installation Guides for latest driver installation. If install the verified Intel® Data Center GPU Max Series/Intel® Data Center GPU Flex Series 540, please append the specific version after components. |
The Docker container includes the Intel® oneAPI Base Toolkit, and all other software stack except Intel GPU Drivers. User only needs to install the GPU driver in host machine bare metal environment, and then launch the docker container directly.
Run the following Dockerfile build procedure to build the pip based deployment container.
Pre-built docker images will be available at DockerHub around January 13.
Please run the following command to pull the Intel® Data Center GPU Flex Series Docker container image (gpu-flex
) to your local machine.
$ docker pull intel/intel-extension-for-tensorflow:gpu-flex
$ docker run -it -p 8888:8888 --device /dev/dri -v /dev/dri/by-path:/dev/dri/by-path intel/intel-extension-for-tensorflow:gpu-flex
Or pull the Intel® Data Center GPU Max Series Docker container image (gpu-max
) to your local machine by the following command.
$ docker pull intel/intel-extension-for-tensorflow:gpu-max
$ docker run -it -p 8888:8888 --device /dev/dri -v /dev/dri/by-path:/dev/dri/by-path intel/intel-extension-for-tensorflow:gpu-max
Then go to your browser on http://localhost:8888/
Need to install components of Intel® oneAPI Base Toolkit:
- Intel® oneAPI DPC++ Compiler
- Intel® oneAPI Threading Building Blocks (oneTBB)
- Intel® oneAPI Math Kernel Library (oneMKL)
- Intel® oneAPI Collective Communications Library (oneCCL), required by Intel® Optimization for Horovod* only
$ wget https://registrationcenter-download.intel.com/akdlm/irc_nas/19079/l_BaseKit_p_2023.0.0.25537_offline.sh
# 4 components are necessary: DPC++/C++ Compiler, DPC++ Libiary, oneTBB and oneMKL
# if you want to run distributed training with Intel® Optimization for Horovod*, oneCCL is needed too(Intel® oneAPI MPI Library will be installed automatically as its dependency)
$ sudo sh ./l_BaseKit_p_2023.0.0.25537_offline.sh
For any more details, please follow the procedure in https://www.intel.com/content/www/us/en/developer/tools/oneapi/base-toolkit.html.
# DPC++ Compiler/oneMKL/oneTBB
source /path to basekit/intel/oneapi/compiler/latest/env/vars.sh
source /path to basekit/intel/oneapi/mkl/latest/env/vars.sh
source /path to basekit/intel/oneapi/tbb/latest/env/vars.sh
# oneCCL (and Intel® oneAPI MPI Library as its dependency), required by Intel® Optimization for Horovod* only
source /path to basekit/intel/oneapi/mpi/latest/env/vars.sh
source /path to basekit/intel/oneapi/ccl/latest/env/vars.sh
A user may install more components than Intel® Extension for TensorFlow* needs, and if required, setvars.sh
can be customized to point to a specific directory by using a configuration file:
source /opt/intel/oneapi/setvars.sh --config="full/path/to/your/config.txt"
The Python development and virtual environment setup recommendation by TensorFlow to isolate package installation from the system.
The Intel® Extension for TensorFlow* requires stock TensorFlow, and the version should be == 2.11.0 or 2.10.0.
You can follow the instructions in stock tensorflow install to activate the virtual environment.
On Linux, it is often necessary to first update pip to a version that supports manylinux2014 wheels.
(tf)$ pip install --upgrade pip
To install in virtual environment, you can run
(tf)$ pip install tensorflow==2.11.0
If want to system install in $HOME, please append --user
to the commands.
$ pip install --user tensorflow==2.11.0
And the following system environment install for Intel® Extension for TensorFlow* will use the same practice.
To install a GPU-only version in virtual environment, which depends on Intel GPU drivers and oneAPI BaseKit, you can run
(tf)$ pip install intel-extension-for-tensorflow[gpu]==1.1.0
python -c "import intel_extension_for_tensorflow as itex; print(itex.__version__)"