Skip to content

Commit

Permalink
update CONTRIBUTING.rst
Browse files Browse the repository at this point in the history
  • Loading branch information
@kanishk16
kanishk16 committed Jul 2, 2022
1 parent d157a6c commit c94f119
Show file tree
Hide file tree
Showing 2 changed files with 60 additions and 214 deletions.
272 changes: 59 additions & 213 deletions CONTRIBUTING.rst
View file
Original file line number Diff line number Diff line change
Expand Up @@ -7,266 +7,112 @@ Thank you for your interest in contributing to ivadomed! This project uses the f
* The `NeuroPoly Contributing Guidelines <https://intranet.neuro.polymtl.ca/software-development/contributing>`_ provide instructions for development workflows, such as reporting issues or submitting pull requests.
* The `ivadomed Developer Wiki <https://github.com/ivadomed/ivadomed/wiki>`_ acts as a knowledge base for documenting internal design decisions specific to the ivadomed codebase. It also contains step-by-step walkthroughs for common ivadomed maintainer tasks.

Step 2: Install ``ivadomed``
++++++++++++++++++++++++++++
Contributor or Developer ``ivadomed`` installation
++++++++++++++++++++++++++++++++++++++++++++++++++

.. tabs::

.. tab:: NVIDIA GPU Support

``ivadomed`` requires a minimum PyTorch version of either
1.8.1 or 1.8.2 which supports CUDA 10.2 and CUDA 11.1 builds
by default.

Ampere-based GPUs (with a Compute Capability of 8.x) only work
CUDA>=11.1. Although CUDA 11.1 is backward compatible with older
hardware, CUDA 10.2 is preferred if available.

CUDA 10.2 and CUDA 11.1 require an NVIDIA driver version >=440.33
and >=450 respectively as indicated `here <https://docs.nvidia.com/deploy/cuda-compatibility/index.html#minor-version-compatibility>`__.
Please make sure to upgrade to the minimum NVIDIA driver version
requirements for the respective CUDA builds. To verify the NVIDIA
driver version, just run the command `nvidia-smi` and you'll find
your current driver version on the top.

.. tabs::

.. tab:: Package Installation (Recommended)

To install ``ivadomed``

.. tabs::

.. tab:: Linux

with CUDA 10.2:

::
pip install ivadomed

and, with CUDA 11.1:

::

pip install ivadomed --extra-index-url https://download.pytorch.org/whl/cu111

.. tab:: Windows

with CUDA 10.2:

::
pip install ivadomed --extra-index-url https://download.pytorch.org/whl/cu102

and, with CUDA 11.1:

::

pip install ivadomed --extra-index-url https://download.pytorch.org/whl/cu111


.. tab:: Source Installation

Bleeding-edge developments are available on the master branch of the project
on Github. To install ``ivadomed`` from source

.. tabs::

.. tab:: Linux

with CUDA 10.2:

::
git clone https://github.com/ivadomed/ivadomed.git

cd ivadomed

pip install -e .

and, with CUDA 11.1:

::

git clone https://github.com/ivadomed/ivadomed.git

cd ivadomed

pip install -e . --extra-index-url https://download.pytorch.org/whl/cu111

.. tab:: Windows

with CUDA 10.2:

::

git clone https://github.com/ivadomed/ivadomed.git

cd ivadomed

pip install -e . --extra-index-url https://download.pytorch.org/whl/cu102

and, with CUDA 11.1:
PyTorch is an integral part of ``ivadomed`` which ships
CUDA 10.2 and CUDA 11.1 runtime by default.

::
Ampere-based GPUs (with a `Compute Capability <https://developer.nvidia.com/cuda-gpus>`_
of 8.x) only work with CUDA>=11.1. Although CUDA 11.1 is
backward compatible with older hardware, CUDA 10.2 is
preferred if available.

git clone https://github.com/ivadomed/ivadomed.git

cd ivadomed

pip install -e . --extra-index-url https://download.pytorch.org/whl/cu111


.. tab:: Contributor or Developer Installation

To contribute to the project, we recommend installing ``ivadomed``
from source along with additional dependencies related to building
documentation and testing
To accelerate ``ivadomed`` with CUDA 10.2 on a Linux system, you'd
need to have GPUs installed with an `NVIDIA driver version >=440.33
<https://docs.nvidia.com/deploy/cuda-compatibility/index.html#minor-version-compatibility>`_.
And, for CUDA 11.1 you'd need an upgraded NVIDIA driver version >=450.

To verify the NVIDIA driver version, just look in ``/sys`` by
executing the command ``cat /sys/module/nvidia/version`` and you'll find
your current driver version.
Before proceeding with the installation, we suggest you to set up a virtual environment
by following the instructions as specified in the :ref:`Step 1 of the installation procedure <just-a-label>`.
``ivadomed`` from source along with additional dependencies related to building documentation and testing:

.. tabs::
.. tabs::

.. tab:: Linux
.. tab:: Linux

with CUDA 10.2:
with CUDA 10.2:

::
::

git clone https://github.com/ivadomed/ivadomed.git
git clone https://github.com/ivadomed/ivadomed.git

cd ivadomed
cd ivadomed

pip install -e .[dev]
pip install -e .[dev]

with CUDA 11.1:
with CUDA 11.1:

::
::

git clone https://github.com/ivadomed/ivadomed.git
git clone https://github.com/ivadomed/ivadomed.git

cd ivadomed
cd ivadomed

pip install -e .[dev] --extra-index-url https://download.pytorch.org/whl/cu111
pip install -e .[dev] --extra-index-url https://download.pytorch.org/whl/cu111

.. tab:: Windows
.. tab:: Mac/Windows

with CUDA 10.2:
with CUDA 10.2:

::
::

git clone https://github.com/ivadomed/ivadomed.git
git clone https://github.com/ivadomed/ivadomed.git

cd ivadomed
cd ivadomed

pip install -e .[dev] --extra-index-url https://download.pytorch.org/whl/cu102
pip install -e .[dev] --extra-index-url https://download.pytorch.org/whl/cu102

with CUDA 11.1:
with CUDA 11.1:

::
::

git clone https://github.com/ivadomed/ivadomed.git
git clone https://github.com/ivadomed/ivadomed.git

cd ivadomed
cd ivadomed

pip install -e .[dev] --extra-index-url https://download.pytorch.org/whl/cu111
pip install -e .[dev] --extra-index-url https://download.pytorch.org/whl/cu111


.. tab:: CPU Support

.. tabs::

.. tab:: Package Installation (Recommended)

.. tabs::

.. tab:: Linux

::

pip install ivadomed --extra-index-url https://download.pytorch.org/whl/cpu

.. tab:: Windows

::

pip install ivadomed

.. tab:: Mac

::
.. tabs::

pip install ivadomed
.. tab:: Linux

.. tab:: Source Installation

Bleeding-edge developments are available on the project's master branch
on Github. To install ``ivadomed`` from source:

.. tabs::

.. tab:: Linux

::

git clone https://github.com/ivadomed/ivadomed.git

cd ivadomed

pip install -e . --extra-index-url https://download.pytorch.org/whl/cpu

.. tab:: Windows

::

git clone https://github.com/ivadomed/ivadomed.git

cd ivadomed
pip install -e .

.. tab:: Mac

::

git clone https://github.com/ivadomed/ivadomed.git

cd ivadomed

pip install -e .

.. tab:: Contributor or Developer Installation

To contribute to the project, we recommend installing ``ivadomed`` from source along with additional dependencies related to building documentation and testing:

.. tabs::

.. tab:: Linux

::
::

git clone https://github.com/ivadomed/ivadomed.git
git clone https://github.com/ivadomed/ivadomed.git

cd ivadomed
cd ivadomed

pip install -e .[dev] --extra-index-url https://download.pytorch.org/whl/cpu
pip install -e .[dev] --extra-index-url https://download.pytorch.org/whl/cpu

.. tab:: Windows
.. tab:: Mac/Windows

::
::

git clone https://github.com/ivadomed/ivadomed.git
git clone https://github.com/ivadomed/ivadomed.git

cd ivadomed
cd ivadomed

pip install -e .[dev]
pip install -e .[dev]

.. tab:: Mac
.. tab:: Mac

::
::

git clone https://github.com/ivadomed/ivadomed.git
git clone https://github.com/ivadomed/ivadomed.git

cd ivadomed
cd ivadomed

pip install -e .[dev]
pip install -e .[dev]
2 changes: 1 addition & 1 deletion docs/source/installation.rst
View file
Original file line number Diff line number Diff line change
Expand Up @@ -119,7 +119,7 @@ Step 2: Install ``ivadomed``

To accelerate ``ivadomed`` with CUDA 10.2 on a Linux system, you'd
need to have GPUs installed with an `NVIDIA driver version >=440.33
<https://docs.nvidia.com/deploy/cuda-compatibility/index.html#minor-version-compatibility>`__.
<https://docs.nvidia.com/deploy/cuda-compatibility/index.html#minor-version-compatibility>`_.
And, for CUDA 11.1 you'd need an upgraded NVIDIA driver version >=450.

To verify the NVIDIA driver version, just look in ``/sys`` by
Expand Down

0 comments on commit c94f119

Please sign in to comment.