8000 Update TF install documentation to CUDA 10.1 and cuDNN 7.6. · tensorflow/docs@1fad670 · GitHub
[go: up one dir, main page]
Skip to content

Commit 1fad670

Browse files
lambertacopybara-github
lamberta
authored and
copybara-github
committed
Update TF install documentation to CUDA 10.1 and cuDNN 7.6.
PiperOrigin-RevId: 288745772
1 parent 8d69d5a commit 1fad670

File tree

2 files changed

+23
-21
lines changed

2 files changed

+23
-21
lines changed

‎site/en/install/gpu.md

Lines changed: 22 additions & 21 deletions
Original file line numberDiff line numberDiff line change
@@ -50,11 +50,12 @@ The following GPU-enabled devices are supported:
5050

5151
The following NVIDIA® software must be installed on your system:
5252

53-
* [NVIDIA® GPU drivers](https://www.nvidia.com/drivers){:.external} —CUDA 10.0 requires 410.x or higher.
53+
* [NVIDIA® GPU drivers](https://www.nvidia.com/drivers){:.external} —CUDA 10.1
54+
requires 418.x or higher.
5455
* [CUDA® Toolkit](https://developer.nvidia.com/cuda-toolkit-archive){:.external}
55-
—TensorFlow supports CUDA 10.0 (TensorFlow >= 1.13.0)
56+
—TensorFlow supports CUDA 10.1 (TensorFlow >= 2.1.0)
5657
* [CUPTI](http://docs.nvidia.com/cuda/cupti/){:.external} ships with the CUDA Toolkit.
57-
* [cuDNN SDK](https://developer.nvidia.com/cudnn){:.external} (>= 7.4.1)
58+
* [cuDNN SDK](https://developer.nvidia.com/cudnn){:.external} (>= 7.6)
5859
* *(Optional)* [TensorRT 6.0](https://docs.nvidia.com/deeplearning/sdk/tensorrt-install-guide/index.html){:.external}
5960
to improve latency and throughput for inference on some models.
6061

@@ -87,12 +88,12 @@ Caution: [Secure Boot](https://wiki.ubuntu.com/UEFI/SecureBoot){:.external}
8788
complicates installation of the NVIDIA driver and is beyond the scope of these instructions.
8889

8990

90-
#### Ubuntu 18.04 (CUDA 10)
91+
#### Ubuntu 18.04 (CUDA 10.1)
9192

9293
<pre class="prettyprint lang-bsh">
9394
# Add NVIDIA package repositories
94-
<code class="devsite-terminal">wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/cuda-repo-ubuntu1804_10.0.130-1_amd64.deb</code>
95-
<code class="devsite-terminal">sudo dpkg -i cuda-repo-ubuntu1804_10.0.130-1_amd64.deb</code>
95+
<code class="devsite-terminal">wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/cuda-repo-ubuntu1804_10.1.243-1_amd64.deb</code>
96+
<code class="devsite-terminal">sudo dpkg -i cuda-repo-ubuntu1804_10.1.243-1_amd64.deb</code>
9697
<code class="devsite-terminal">sudo apt-key adv --fetch-keys https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/7fa2af80.pub</code>
9798
<code class="devsite-terminal">sudo apt-get update</code>
9899
<code class="devsite-terminal">wget http://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1804/x86_64/nvidia-machine-learning-repo-ubuntu1804_1.0.0-1_amd64.deb</code>
@@ -105,14 +106,14 @@ complicates installation of the NVIDIA driver and is beyond the scope of these i
105106

106107
# Install development and runtime libraries (~4GB)
107108
<code class="devsite-terminal">sudo apt-get install --no-install-recommends \
108-
cuda-10-0 \
109-
libcudnn7=7.6.2.24-1+cuda10.0 \
110-
libcudnn7-dev=7.6.2.24-1+cuda10.0
109+
cuda-10-1 \
110+
libcudnn7=7.6.4.38-1+cuda10.1 \
111+
libcudnn7-dev=7.6.4.38-1+cuda10.1
111112
</code>
112113

113114
# Install TensorRT. Requires that libcudnn7 is installed above.
114-
<code class="devsite-terminal">sudo apt-get install -y --no-install-recommends libnvinfer6=6.0.1-1+cuda10.0 \
115-
libnvinfer-dev=6.0.1-1+cuda10.0
115+
<code class="devsite-terminal">sudo apt-get install -y --no-install-recommends libnvinfer6=6.0.1-1+cuda10.1 \
116+
libnvinfer-dev=6.0.1-1+cuda10.1
116117
</code>
117118
</pre>
118119

@@ -123,8 +124,8 @@ complicates installation of the NVIDIA driver and is beyond the scope of these i
123124
# Add NVIDIA package repositories
124125
# Add HTTPS support for apt-key
125126
<code class="devsite-terminal">sudo apt-get install gnupg-curl</code>
126-
<code class="devsite-terminal">wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1604/x86_64/cuda-repo-ubuntu1604_10.0.130-1_amd64.deb</code>
127-
<code class="devsite-terminal">sudo dpkg -i cuda-repo-ubuntu1604_10.0.130-1_amd64.deb</code>
127+
<code class="devsite-terminal">wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1604/x86_64/cuda-repo-ubuntu1604_10.1.243-1_amd64.deb</code>
128+
<code class="devsite-terminal">sudo dpkg -i cuda-repo-ubuntu1604_10.1.243-1_amd64.deb</code>
128129
<code class="devsite-terminal">sudo apt-key adv --fetch-keys https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1604/x86_64/7fa2af80.pub</code>
129130
<code class="devsite-terminal">sudo apt-get update</code>
130131
<code class="devsite-terminal">wget http://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1604/x86_64/nvidia-machine-learning-repo-ubuntu1604_1.0.0-1_amd64.deb</code>
@@ -140,13 +141,13 @@ complicates installation of the NVIDIA driver and is beyond the scope of these i
140141
# Install development and runtime libraries (~4GB)
141142
<code class="devsite-terminal">sudo apt-get install --no-install-recommends \
142143
cuda-10-0 \
143-
libcudnn7=7.6.2.24-1+cuda10.0 \
144-
libcudnn7-dev=7.6.2.24-1+cuda10.0
144+
libcudnn7=7.6.4.38-1+cuda10.1 \
145+
libcudnn7-dev=7.6.4.38-1+cuda10.1
145146
</code>
146147

147148
# Install TensorRT. Requires that libcudnn7 is installed above.
148-
<code class="devsite-terminal">sudo apt-get install -y --no-install-recommends libnvinfer5=5.1.5-1+cuda10.0 \
149-
libnvinfer-dev=5.1.5-1+cuda10.0
149+
<code class="devsite-terminal">sudo apt-get install -y --no-install-recommends libnvinfer5=6.0.1-1+cuda10.1 \
150+
libnvinfer-dev=6.0.1-1+cuda10.1
150151
</code>
151152
</pre>
152153

@@ -191,12 +192,12 @@ different version, see the [Windows build from source](./source_windows.md) guid
191192

192193
Add the CUDA, CUPTI, and cuDNN installation directories to the `%PATH%`
193194
environmental variable. For example, if the CUDA Toolkit is installed to
194-
`C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0` and cuDNN to
195+
`C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1` and cuDNN to
195196
`C:\tools\cuda`, update your `%PATH%` to match:
196197

197198
<pre class="devsite-click-to-copy">
198-
<code class="devsite-terminal tfo-terminal-windows">SET PATH=C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0\bin;%PATH%</code>
199-
<code class="devsite-terminal tfo-terminal-windows">SET PATH=C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0\extras\CUPTI\libx64;%PATH%</code>
200-
<code class="devsite-terminal tfo-terminal-windows">SET PATH=C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0\include;%PATH%</code>
199+
<code class="devsite-terminal tfo-terminal-windows">SET PATH=C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1\bin;%PATH%</code>
200+
<code class="devsite-terminal tfo-terminal-windows">SET PATH=C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1\extras\CUPTI\libx64;%PATH%</code>
201+
<code class="devsite-terminal tfo-terminal-windows">SET PATH=C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1\include;%PATH%</code>
201202
<code class="devsite-terminal tfo-terminal-windows">SET PATH=C:\tools\cuda\bin;%PATH%</code>
202203
</pre>

‎site/en/install/source.md

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -451,6 +451,7 @@ Success: TensorFlow is now installed.
451451

452452
<table>
453453
<tr><th>Version</th><th>Python version</th><th>Compiler</th><th>Build tools</th><th>cuDNN</th><th>CUDA</th></tr>
454+
<tr><td>tensorflow-2.1.0</td><td>2.7, 3.3-3.7</td><td>GCC 7.3.1</td><td>Bazel 0.26.1</td><td>7.6</td><td>10.1</td></tr>
454455
<tr><td>tensorflow-2.0.0</td><td>2.7, 3.3-3.7</td><td>GCC 7.3.1</td><td>Bazel 0.26.1</td><td>7.4</td><td>10.0</td></tr>
455456
<tr><td>tensorflow_gpu-1.14.0</td><td>2.7, 3.3-3.7</td><td>GCC 4.8</td><td>Bazel 0.24.1</td><td>7.4</td><td>10.0</td></tr>
456457
<tr><td>tensorflow_gpu-1.13.1</td><td>2.7, 3.3-3.7</td><td>GCC 4.8</td><td>Bazel 0.19.2</td><td>7.4</td><td>10.0</td></tr>

0 commit comments

Comments
 (0)
0