Intel® Extension for TensorFlow*🔗
Intel® Extension for TensorFlow* extends TensorFlow* with up-to-date feature optimizations for an extra performance boost on Intel hardware.
Intel® Extension for TensorFlow* is based on the TensorFlow PluggableDevice interface to bring Intel XPU(GPU, CPU, etc.) devices into TensorFlow* with flexibility for on-demand performance on the following Intel GPUs:
Note: There are two dockerhub repositories (
intel/intel-extension-for-tensorflowandintel/intel-optimized-tensorflow) that are routinely updated with the latest images, however, some legacy images have not be published to both repositories.
XPU images🔗
The images below include support for both CPU and GPU optimizations:
| Tag(s) | TensorFlow | ITEX | Driver | Dockerfile |
|---|---|---|---|---|
2.15.0.1-xpu-pip-base, xpu |
v2.15.1 | v2.15.0.1 | 803.63 | v0.4.0-Beta |
2.15.0.0-xpu |
v2.15.0 | v2.15.0.0 | 803 | v0.4.0-Beta |
2.14.0.1-xpu |
v2.14.1 | v2.14.0.1 | 736 | v0.3.4 |
2.13.0.0-xpu |
v2.13.0 | v2.13.0.0 | 647 | v0.2.3 |
Run the XPU Container🔗
docker run -it --rm \
--device /dev/dri \
-v /dev/dri/by-path:/dev/dri/by-path \
--ipc=host \
intel/intel-extension-for-tensorflow:xpu
The images below additionally include Jupyter Notebook server:
| Tag(s) | TensorFlow | IPEX | Driver | Dockerfile |
|---|---|---|---|---|
2.15.0.1-xpu-pip-jupyter |
v2.15.1 | v2.15.0.1 | 803.63 | v0.4.0-Beta |
xpu-jupyter |
v2.14.1 | v2.14.0.1 | 736 | v0.3.4 |
Run the XPU Jupyter Container🔗
docker run -it --rm \
-p 8888:8888 \
--net=host \
--device /dev/dri \
-v /dev/dri/by-path:/dev/dri/by-path \
--ipc=host \
intel/intel-extension-for-tensorflow:2.15.0.1-xpu-pip-jupyter
After running the command above, copy the URL (something like http://127.0.0.1:$PORT/?token=***) into your browser to access the notebook server.
The images below are TensorFlow* Serving with GPU Optimizations:
| Tag(s) | TensorFlow | IPEX |
|---|---|---|
2.14.0.1-serving-gpu, serving-gpu |
v2.14.1 | v2.14.0.1 |
2.13.0.0-serving-gpu, |
v2.13.0 | v2.13.0.0 |
Run the Serving GPU Container🔗
docker run -it --rm \
-p 8500:8500 \
--device /dev/dri \
-v /dev/dri/by-path:/dev/dri/by-path \
-v $PWD/workspace:/workspace \
-w /workspace \
-e MODEL_NAME=<your-model-name> \
-e MODEL_DIR=<your-model-dir> \
intel/intel-extension-for-tensorflow:serving-gpu
For more details, follow the procedure in the Intel® Extension for TensorFlow* Serving instructions.
CPU only images🔗
The images below are built only with CPU optimizations (GPU acceleration support was deliberately excluded):
| Tag(s) | TensorFlow | ITEX | Dockerfile |
|---|---|---|---|
2.15.1-pip-base, latest |
v2.15.1 | v2.15.0.1 | v0.4.0-Beta |
2.15.0-pip-base |
v2.15.0 | v2.15.0.0 | v0.4.0-Beta |
2.14.0-pip-base |
v2.14.1 | v2.14.0.1 | v0.3.4 |
2.13-pip-base |
v2.13.0 | v2.13.0.0 | v0.2.3 |
The images below additionally include Jupyter Notebook server:
| Tag(s) | TensorFlow | ITEX | Dockerfile |
|---|---|---|---|
2.15.1-pip-jupyter |
v2.15.1 | v2.15.0.1 | v0.4.0-Beta |
2.15.0-pip-jupyter |
v2.15.0 | v2.15.0.0 | v0.4.0-Beta |
2.14.0-pip-jupyter |
v2.14.1 | v2.14.0.1 | v0.3.4 |
2.13-pip-jupyter |
v2.13.0 | v2.13.0.0 | v0.2.3 |
Run the CPU Jupyter Container🔗
docker run -it --rm \
-p 8888:8888 \
--net=host \
-v $PWD/workspace:/workspace \
-w /workspace \
intel/intel-extension-for-tensorflow:2.15.1-pip-jupyter
After running the command above, copy the URL (something like http://127.0.0.1:$PORT/?token=***) into your browser to access the notebook server.
The images below additionally include Horovod:
| Tag(s) | Tensorflow | ITEX | Horovod | Dockerfile |
|---|---|---|---|---|
2.15.1-pip-multinode |
v2.15.1 | v2.15.0.1 | v0.28.1 | v0.4.0-Beta |
2.15.0-pip-multinode |
v2.15.0 | v2.15.0.0 | v0.28.1 | v0.4.0-Beta |
2.14.0-pip-openmpi-multinode |
v2.14.1 | v2.14.0.1 | v0.28.1 | v0.3.4 |
2.13-pip-openmpi-mulitnode |
v2.13.0 | v2.13.0.0 | v0.28.0 | v0.2.3 |
Note
Passwordless SSH connection is also enabled in the image, but the container does not contain any SSH ID keys. The user needs to mount those keys at /root/.ssh/id_rsa and /etc/ssh/authorized_keys.
Tip
Before mounting any keys, modify the permissions of those files with chmod 600 authorized_keys; chmod 600 id_rsa to grant read access for the default user account.
Setup and Run ITEX Multi-Node Container🔗
Important
Maintainence, Bug Fixes, and Releases of Intel® Extension for TensorFlow* Multi-Node Container for Xeon Processors have ceased development. The last supported version is 2.15.1. For future releases, please use the Intel® Extension for TensorFlow* Multi-Node Container for XPU.
Some additional assembly is required to utilize this container with OpenSSH. To perform any kind of DDP (Distributed Data Parallel) execution, containers are assigned the roles of launcher and worker respectively:
SSH Server (Worker)
- Authorized Keys :
/etc/ssh/authorized_keys
SSH Client (Launcher)
- Private User Key :
/root/.ssh/id_rsa
To add these files correctly please follow the steps described below.
-
Setup ID Keys
You can use the commands provided below to generate the identity keys for OpenSSH.
ssh-keygen -q -N "" -t rsa -b 4096 -f ./id_rsa touch authorized_keys cat id_rsa.pub >> authorized_keys -
Configure the permissions and ownership for all of the files you have created so far
chmod 600 id_rsa config authorized_keys chown root:root id_rsa.pub id_rsa config authorized_keys -
Create a hostfile for horovod. (Optional)
Host host1 HostName <Hostname of host1> IdentitiesOnly yes IdentityFile ~/.root/id_rsa Port <SSH Port> Host host2 HostName <Hostname of host2> IdentitiesOnly yes IdentityFile ~/.root/id_rsa Port <SSH Port> ... -
Configure Horovod in your python script
import horovod.torch as hvd hvd.init() -
Now start the workers and execute DDP on the launcher
-
Worker run command:
docker run -it --rm \ --net=host \ -v $PWD/authorized_keys:/etc/ssh/authorized_keys \ -v $PWD/tests:/workspace/tests \ -w /workspace \ intel/intel-optimized-tensorflow:2.15.1-pip-multinode \ bash -c '/usr/sbin/sshd -D' -
Launcher run command:
docker run -it --rm \ --net=host \ -v $PWD/id_rsa:/root/.ssh/id_rsa \ -v $PWD/tests:/workspace/tests \ -v $PWD/hostfile:/root/ssh/config \ -w /workspace \ intel/intel-optimized-tensorflow:2.15.1-pip-multinode \ bash -c 'horovodrun --verbose -np 2 -H host1:1,host2:1 /workspace/tests/tf_base_test.py'
-
Note
Intel® MPI can be configured based on your machine settings. If the above commands do not work for you, see the documentation for how to configure based on your network.
The images below are TensorFlow* Serving with CPU Optimizations:
| Tag(s) | TensorFlow | ITEX |
|---|---|---|
2.14.0.1-serving-cpu, serving-cpu |
v2.14.1 | v2.14.0.1 |
2.13.0.0-serving-cpu |
v2.13.0 | v2.13.0.0 |
Run the Serving CPU Container🔗
docker run -it --rm \
-p 8500:8500 \
--device /dev/dri \
-v /dev/dri/by-path:/dev/dri/by-path \
-v $PWD/workspace:/workspace \
-w /workspace \
-e MODEL_NAME=<your-model-name> \
-e MODEL_DIR=<your-model-dir> \
intel/intel-extension-for-tensorflow:serving-cpu
For more details, follow the procedure in the Intel® Extension for TensorFlow* Serving instructions.
CPU only images with Intel® Distribution for Python*🔗
The images below are built only with CPU optimizations (GPU acceleration support was deliberately excluded) and include Intel® Distribution for Python*:
| Tag(s) | TensorFlow | ITEX | Dockerfile |
|---|---|---|---|
2.15.1-idp-base |
v2.15.1 | v2.15.0.1 | v0.4.0-Beta |
2.15.0-idp-base |
v2.15.0 | v2.15.0.0 | v0.4.0-Beta |
2.14.0-idp-base |
v2.14.1 | v2.14.0.1 | v0.3.4 |
2.13-idp-base |
v2.13.0 | v2.13.0.0 | v0.2.3 |
The images below additionally include Jupyter Notebook server:
| Tag(s) | TensorFlow | ITEX | Dockerfile |
|---|---|---|---|
2.15.1-idp-jupyter |
v2.15.1 | v2.15.0.1 | v0.4.0-Beta |
2.15.0-idp-jupyter |
v2.15.0 | v2.15.0.0 | v0.4.0-Beta |
2.14.0-idp-jupyter |
v2.14.1 | v2.14.0.1 | v0.3.4 |
2.13-idp-jupyter |
v2.13.0 | v2.13.0.0 | v0.2.3 |
The images below additionally include Horovod:
| Tag(s) | Tensorflow | ITEX | Horovod | Dockerfile |
|---|---|---|---|---|
2.15.1-idp-multinode |
v2.15.1 | v2.15.0.1 | v0.28.1 | v0.4.0-Beta |
2.15.0-idp-multinode |
v2.15.0 | v2.15.0.0 | v0.28.1 | v0.4.0-Beta |
2.14.0-idp-openmpi-multinode |
v2.14.1 | v2.14.0.1 | v0.28.1 | v0.3.4 |
2.13-idp-openmpi-mulitnode |
v2.13.0 | v2.13.0.0 | v0.28.0 | v0.2.3 |
XPU images with Intel® Distribution for Python*🔗
The images below are built only with CPU and GPU optimizations and include Intel® Distribution for Python*:
| Tag(s) | Pytorch | ITEX | Driver | Dockerfile |
|---|---|---|---|---|
2.15.0.1-xpu-idp-base |
v2.15.1 | v2.15.0.1 | 803 | v0.4.0-Beta |
2.15.0-xpu-idp-base |
v2.15.0 | v2.15.0.0 | 803 | v0.4.0-Beta |
The images below additionally include Jupyter Notebook server:
| Tag(s) | Pytorch | IPEX | Driver | Jupyter Port | Dockerfile |
|---|---|---|---|---|---|
2.15.0.1-xpu-idp-jupyter |
v2.15.1 | v2.15.0.1 | 803 | 8888 |
v0.4.0-Beta |
2.15.0-xpu-idp-jupyter |
[v2.1.0] | v2.15.0.0 | 803 | 8888 |
v0.4.0-Beta |
Build from Source🔗
To build the images from source, clone the AI Containers repository, follow the main README.md file to setup your environment, and run the following command:
cd pytorch
docker compose build tf-base
docker compose run tf-base
You can find the list of services below for each container in the group:
| Service Name | Description |
|---|---|
tf-base |
Base image with Intel® Extension for TensorFlow* |
jupyter |
Adds Jupyter Notebook server |
multinode |
Adds Intel® MPI, Horovod and INC |
xpu |
Adds Intel GPU Support |
xpu-jupyter |
Adds Jupyter notebook server to GPU image |
License🔗
View the License for the Intel® Extension for TensorFlow*.
The images below also contain other software which may be under other licenses (such as TensorFlow, Jupyter, Bash, etc. from the base).
It is the image user's responsibility to ensure that any use of The images below comply with any relevant licenses for all software contained within.
* Other names and brands may be claimed as the property of others.