Skip to content

Commit 463a418

Browse files
authored
Update installation guide etc. regarding compilers and NCCL (horovod#3107)
Signed-off-by: Max H. Gerlach <[email protected]>
1 parent 2a401a3 commit 463a418

File tree

5 files changed

+13
-16
lines changed

5 files changed

+13
-16
lines changed

README.rst

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -123,9 +123,9 @@ To install Horovod:
123123

124124
<p/>
125125

126-
2. If you've installed TensorFlow from `PyPI <https://pypi.org/project/tensorflow>`__, make sure that the ``g++-4.8.5`` or ``g++-4.9`` or above is installed.
126+
2. If you've installed TensorFlow from `PyPI <https://pypi.org/project/tensorflow>`__, make sure that ``g++-5`` or above is installed.
127127

128-
If you've installed PyTorch from `PyPI <https://pypi.org/project/torch>`__, make sure that the ``g++-4.9`` or above is installed.
128+
If you've installed PyTorch from `PyPI <https://pypi.org/project/torch>`__, make sure that ``g++-5`` or above is installed.
129129

130130
If you've installed either package from `Conda <https://conda.io>`_, make sure that the ``gxx_linux-64`` Conda package is installed.
131131

docs/gpus.rst

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -32,9 +32,9 @@ by installing an `nv_peer_memory <https://github.com/Mellanox/nv_peer_memory>`__
3232

3333
**Note**: Open MPI 3.1.3 has an issue that may cause hangs. The recommended fix is to downgrade to Open MPI 3.1.2 or upgrade to Open MPI 4.0.0.
3434

35-
4. If you installed TensorFlow from `PyPI <https://pypi.org/project/tensorflow>`__, make sure that the ``g++-4.8.5`` or ``g++-4.9`` or above is installed.
35+
4. If you installed TensorFlow from `PyPI <https://pypi.org/project/tensorflow>`__, make sure that ``g++-5`` or above is installed.
3636

37-
If you installed PyTorch from `PyPI <https://pypi.org/project/torch>`__, make sure that the ``g++-4.9`` or above is installed.
37+
If you installed PyTorch from `PyPI <https://pypi.org/project/torch>`__, make sure that ``g++-5`` or above is installed.
3838

3939
If you installed either package from `Conda <https://conda.io>`_, make sure that the ``gxx_linux-64`` Conda package is installed.
4040

docs/index.rst

Lines changed: 3 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -14,7 +14,7 @@ Choose your deep learning framework to learn how to get started with Horovod.
1414
<ol>
1515
<li><a href="https://www.open-mpi.org/faq/?category=building#easy-build">Install Open MPI 3.1.2 or 4.0.0</a>, or another MPI implementation. </li>
1616
<li>
17-
If you've installed TensorFlow from <a href="https://pypi.org/project/tensorflow">PyPI</a>, make sure that the <code>g++-4.8.5</code> or <code>g++-4.9</code> or above is installed.<br/>
17+
If you've installed TensorFlow from <a href="https://pypi.org/project/tensorflow">PyPI</a>, make sure that <code>g++-5</code> or above is installed.<br/>
1818
If you've installed TensorFlow from <a href="https://conda.io">Conda</a>, make sure that the <code>gxx_linux-64</code> Conda package is installed.
1919
</li>
2020
<li>Install the Horovod pip package: <code>pip install horovod</code></li>
@@ -30,7 +30,7 @@ Choose your deep learning framework to learn how to get started with Horovod.
3030
<ol>
3131
<li><a href="https://www.open-mpi.org/faq/?category=building#easy-build">Install Open MPI 3.1.2 or 4.0.0</a>, or another MPI implementation. </li>
3232
<li>
33-
If you've installed TensorFlow from <a href="https://pypi.org/project/tensorflow">PyPI</a>, make sure that the <code>g++-4.8.5</code> or <code>g++-4.9</code> or above is installed.<br/>
33+
If you've installed TensorFlow from <a href="https://pypi.org/project/tensorflow">PyPI</a>, make sure that <code>g++-5</code> or above is installed.<br/>
3434
If you've installed TensorFlow from <a href="https://conda.io">Conda</a>, make sure that the <code>gxx_linux-64</code> Conda package is installed.
3535
</li>
3636
<li>Install the Horovod pip package: <code>pip install horovod</code></li>
@@ -46,7 +46,7 @@ Choose your deep learning framework to learn how to get started with Horovod.
4646
<ol>
4747
<li><a href="https://www.open-mpi.org/faq/?category=building#easy-build">Install Open MPI 3.1.2 or 4.0.0</a>, or another MPI implementation. </li>
4848
<li>
49-
If you've installed PyTorch from <a href="https://pypi.org/project/torch">PyPI</a>, make sure that the <code>g++-4.9</code> or above is installed.<br/>
49+
If you've installed PyTorch from <a href="https://pypi.org/project/torch">PyPI</a>, make sure that <code>g++-5</code> or above is installed.<br/>
5050
If you've installed PyTorch from <a href="https://conda.io">Conda</a>, make sure that the <code>gxx_linux-64</code> Conda package is installed.
5151
</li>
5252
<li>Install the Horovod pip package: <code>pip install horovod</code></li>

docs/install.rst

Lines changed: 4 additions & 7 deletions
Original file line numberDiff line numberDiff line change
@@ -7,6 +7,7 @@ Requirements
77
------------
88

99
- Python >= 3.6
10+
- `g++-5` or above, or another compiler supporting C++14
1011
- CMake
1112
- TensorFlow, PyTorch, or MXNet
1213
- (Optional) MPI
@@ -51,9 +52,6 @@ To ensure that Horovod is built with TensorFlow support enabled:
5152
5253
To skip TensorFlow, set ``HOROVOD_WITHOUT_TENSORFLOW=1`` in your environment.
5354

54-
If you've installed TensorFlow from `PyPI <https://pypi.org/project/tensorflow>`__, make sure that
55-
the ``g++-4.8.5`` or ``g++-4.9`` or above is installed.
56-
5755
PyTorch
5856
~~~~~~~
5957

@@ -65,9 +63,6 @@ To ensure that Horovod is built with PyTorch support enabled:
6563
6664
To skip PyTorch, set ``HOROVOD_WITHOUT_PYTORCH=1`` in your environment.
6765

68-
If you've installed PyTorch from `PyPI <https://pypi.org/project/torch>`__, make sure that the ``g++-4.9`` or
69-
above is installed.
70-
7166
MXNet
7267
~~~~~
7368

@@ -165,12 +160,14 @@ will be used for CPU operations. You can override this by setting ``HOROVOD_CPU_
165160
NCCL
166161
~~~~
167162

168-
NCCL is supported for Allreduce, Allgather, and Broadcast operations. You can enable these by setting
163+
NCCL is supported for Allreduce, Allgather, Broadcast, and Alltoall operations. You can enable these by setting
169164
``HOROVOD_GPU_OPERATIONS=NCCL`` during installation.
170165

171166
NCCL operations are supported on both Nvidia (CUDA) and AMD (ROCm) GPUs. You can set ``HOROVOD_GPU`` in your
172167
environment to specify building with CUDA or ROCm. CUDA will be assumed if not specified.
173168

169+
Note that Alltoall requires NCCL version >= 2.7.0.
170+
174171
MPI
175172
~~~
176173

docs/summary.rst

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -115,9 +115,9 @@ To install Horovod:
115115

116116
<p/>
117117

118-
2. If you've installed TensorFlow from `PyPI <https://pypi.org/project/tensorflow>`__, make sure that the ``g++-4.8.5`` or ``g++-4.9`` or above is installed.
118+
2. If you've installed TensorFlow from `PyPI <https://pypi.org/project/tensorflow>`__, make sure that ``g++-5`` or above is installed.
119119

120-
If you've installed PyTorch from `PyPI <https://pypi.org/project/torch>`__, make sure that the ``g++-4.9`` or above is installed.
120+
If you've installed PyTorch from `PyPI <https://pypi.org/project/torch>`__, make sure that ``g++-5`` or above is installed.
121121

122122
If you've installed either package from `Conda <https://conda.io>`_, make sure that the ``gxx_linux-64`` Conda package is installed.
123123

0 commit comments

Comments
 (0)