Skip to content

Commit dba9cc3

Browse files
MarkDaoustcopybara-github
MarkDaoust
authored and
copybara-github
committed
Mention the 3P builds on the install page.
The blog post is published: https://blog.tensorflow.org/2022/09/announcing-tensorflow-official-build-collaborators.html PiperOrigin-RevId: 472783812
1 parent 767618c commit dba9cc3

File tree

1 file changed

+104
-52
lines changed

1 file changed

+104
-52
lines changed

‎site/en/install/pip.md

Lines changed: 104 additions & 52 deletions
Original file line numberDiff line numberDiff line change
@@ -13,74 +13,100 @@ step-by-step instructions.
1313

1414
* {Linux}
1515

16-
```bash
17-
conda install -c conda-forge cudatoolkit=11.2 cudnn=8.1.0
18-
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$CONDA_PREFIX/lib/
19-
python3 -m pip install tensorflow
20-
# Verify install:
21-
python3 -c "import tensorflow as tf; print(tf.config.list_physical_devices('GPU'))"
22-
```
16+
Note: Starting with TensorFlow `2.10`, Linux CPU-builds for Aarch64/ARM64
17+
processors are are built, maintained, tested and released by a third party:
18+
[AWS](https://aws.amazon.com/).
19+
Installing the [`tensorflow`](https://pypi.org/project/tensorflow/)
20+
package on an ARM machine installs AWS's
21+
[`tensorflow-cpu-aws`](https://pypi.org/project/tensorflow-cpu-aws/) package.
22+
They are provided as-is. Tensorflow will use reasonable efforts to maintain
23+
the availability and integrity of this pip package. There may be delays if
24+
the third party fails to release the pip package. See
25+
[this blog post](https://blog.tensorflow.org/2022/09/announcing-tensorflow-official-build-collaborators.html)
26+
for more information about this collaboration.
27+
28+
```bash
29+
conda install -c conda-forge cudatoolkit=11.2 cudnn=8.1.0
30+
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$CONDA_PREFIX/lib/
31+
python3 -m pip install tensorflow
32+
# Verify install:
33+
python3 -c "import tensorflow as tf; print(tf.config.list_physical_devices('GPU'))"
34+
```
2335

2436
* {MacOS}
2537

26-
```bash
27-
# There is currently no official GPU support for MacOS.
28-
python3 -m pip install tensorflow
29-
# Verify install:
30-
python3 -c "import tensorflow as tf; print(tf.reduce_sum(tf.random.normal([1000, 1000])))"
31-
```
38+
```bash
39+
# There is currently no official GPU support for MacOS.
40+
python3 -m pip install tensorflow
41+
# Verify install:
42+
python3 -c "import tensorflow as tf; print(tf.reduce_sum(tf.random.normal([1000, 1000])))"
43+
```
3244

3345
* {Windows Native}
3446

35-
Caution: The current TensorFlow version, `2.10`, is the **last** TensorFlow
36-
release that will support GPU on native-Windows.
37-
Starting with TensorFlow `2.11`, you will need to install
38-
[TensorFlow in WSL2](https://tensorflow.org/install/pip#windows-wsl2),
39-
or install `tensorflow_cpu` and, optionally, try the
40-
[TensorFlow-DirectML-Plugin](https://github.com/microsoft/tensorflow-directml-plugin#tensorflow-directml-plugin-)
47+
Caution: The current TensorFlow version, `2.10`, is the **last** TensorFlow
48+
release that will support GPU on native-Windows.
49+
Starting with TensorFlow `2.11`, you will need to install
50+
[TensorFlow in WSL2](https://tensorflow.org/install/pip#windows-wsl2),
51+
or install `tensorflow_cpu` and, optionally, try the
52+
[TensorFlow-DirectML-Plugin](https://github.com/microsoft/tensorflow-directml-plugin#tensorflow-directml-plugin-)
4153

42-
```bash
43-
conda install -c conda-forge cudatoolkit=11.2 cudnn=8.1.0
44-
python3 -m pip install tensorflow
45-
# Verify install:
46-
python3 -c "import tensorflow as tf; print(tf.config.list_physical_devices('GPU'))"
47-
```
54+
```bash
55+
conda install -c conda-forge cudatoolkit=11.2 cudnn=8.1.0
56+
python3 -m pip install tensorflow
57+
# Verify install:
58+
python3 -c "import tensorflow as tf; print(tf.config.list_physical_devices('GPU'))"
59+
```
4860

4961
* {Windows WSL2}
5062

51-
Note: TensorFlow with GPU access is supported for WSL2 on Windows 10 19044 or
52-
higher. This corresponds to Windows 10 version 21H2, the November 2021
53-
update. You can get the latest update from here:
54-
[Download Windows 10](https://www.microsoft.com/software-download/windows10){:.external}.
55-
For instructions, see
56-
[Install WSL2](https://docs.microsoft.com/windows/wsl/install){:.external}
57-
and
58-
[NVIDIA’s setup docs](https://docs.nvidia.com/cuda/wsl-user-guide/index.html){:.external}
59-
for CUDA in WSL.
60-
61-
```bash
62-
conda install -c conda-forge cudatoolkit=11.2 cudnn=8.1.0
63-
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$CONDA_PREFIX/lib/
64-
python3 -m pip install tensorflow
65-
# Verify install:
66-
python3 -c "import tensorflow as tf; print(tf.config.list_physical_devices('GPU'))"
67-
```
63+
Note: TensorFlow with GPU access is supported for WSL2 on Windows 10 19044 or
64+
higher. This corresponds to Windows 10 version 21H2, the November 2021
65+
update. You can get the latest update from here:
66+
[Download Windows 10](https://www.microsoft.com/software-download/windows10){:.external}.
67+
For instructions, see
68+
[Install WSL2](https://docs.microsoft.com/windows/wsl/install){:.external}
69+
and
70+
[NVIDIA’s setup docs](https://docs.nvidia.com/cuda/wsl-user-guide/index.html){:.external}
71+
for CUDA in WSL.
72+
73+
```bash
74+
conda install -c conda-forge cudatoolkit=11.2 cudnn=8.1.0
75+
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$CONDA_PREFIX/lib/
76+
python3 -m pip install tensorflow
77+
# Verify install:
78+
python3 -c "import tensorflow as tf; print(tf.config.list_physical_devices('GPU'))"
79+
```
6880

6981
* {CPU}
7082

71-
```bash
72-
python3 -m pip install tensorflow
73-
# Verify install:
74-
python3 -c "import tensorflow as tf; print(tf.reduce_sum(tf.random.normal([1000, 1000])))"
75-
```
83+
Note: Starting with TensorFlow `2.10`, Windows CPU-builds for x86/x64
84+
processors are built, maintained, tested and released by a third party:
85+
[Intel](https://www.intel.com/).
86+
Installing the windows-native
87+
[`tensorflow-cpu`](https://pypi.org/project/tensorflow-cpu/)
88+
package installs Intel's
89+
[`tensorflow-intel`](https://pypi.org/project/tensorflow-intel/)
90+
package. These packages are provided as-is. Tensorflow will use reasonable
91+
efforts to maintain the availability and integrity of this pip package.
92+
There may be delays if the third party fails to release the pip package. See
93+
[this blog post](https://blog.tensorflow.org/2022/09/announcing-tensorflow-official-build-collaborators.html)
94+
for more information about this
95+
collaboration.
96+
97+
```bash
98+
python3 -m pip install tensorflow
99+
# Verify install:
100+
python3 -c "import tensorflow as tf; print(tf.reduce_sum(tf.random.normal([1000, 1000])))"
101+
```
76102
77103
* {Nightly}
78104
79-
```bash
80-
python3 -m pip install tf-nightly
81-
# Verify install:
82-
python3 -c "import tensorflow as tf; print(tf.reduce_sum(tf.random.normal([1000, 1000])))"
83-
```
105+
```bash
106+
python3 -m pip install tf-nightly
107+
# Verify install:
108+
python3 -c "import tensorflow as tf; print(tf.reduce_sum(tf.random.normal([1000, 1000])))"
109+
```
84110
85111
## Hardware requirements
86112
@@ -145,6 +171,18 @@ The following NVIDIA® software are only required for GPU support.
145171
TensorFlow only officially support Ubuntu. However, the following
146172
instructions may also work for other Linux distros.
147173
174+
Note: Starting with TensorFlow `2.10`, Linux CPU-builds for Aarch64/ARM64
175+
processors are are built, maintained, tested and released by a third party:
176+
[AWS](https://aws.amazon.com/).
177+
Installing the [`tensorflow`](https://pypi.org/project/tensorflow/)
178+
package on an ARM machine installs AWS's
179+
[`tensorflow-cpu-aws`](https://pypi.org/project/tensorflow-cpu-aws/) package.
180+
They are provided as-is. Tensorflow will use reasonable efforts to maintain
181+
the availability and integrity of this pip package. There may be delays if
182+
the third party fails to release the pip package. See
183+
[this blog post](https://blog.tensorflow.org/2022/09/announcing-tensorflow-official-build-collaborators.html)
184+
for more information about this collaboration.
185+
148186
### 2. Install Miniconda
149187

150188
[Miniconda](https://docs.conda.io/en/latest/miniconda.html){:.external} is the
@@ -357,6 +395,20 @@ The following NVIDIA® software are only required for GPU support.
357395

358396
* Windows 7 or higher (64-bit)
359397

398+
Note: Starting with TensorFlow `2.10`, Windows CPU-builds for x86/x64
399+
processors are built, maintained, tested and released by a third party:
400+
[Intel](https://www.intel.com/).
401+
Installing the windows-native
402+
[`tensorflow-cpu`](https://pypi.org/project/tensorflow-cpu/)
403+
package installs Intel's
404+
[`tensorflow-intel`](https://pypi.org/project/tensorflow-intel/)
405+
package. These packages are provided as-is. Tensorflow will use reasonable
406+
efforts to maintain the availability and integrity of this pip package.
407+
There may be delays if the third party fails to release the pip package. See
408+
[this blog post](https://blog.tensorflow.org/2022/09/announcing-tensorflow-official-build-collaborators.html)
409+
for more information about this
410+
collaboration.
411+
360412
### 2. Install Microsoft Visual C++ Redistributable
361413
362414
Install the *Microsoft Visual C++ Redistributable for Visual Studio 2015,

0 commit comments

Comments
 (0)