FCN ObjectSegmentation with PyTorch backend #2041
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
iory
approved these changes
Mar 28, 2017
|
what's happens if we run `pip install pytorch`
…--
◉ Kei Okada
2017-03-27 2:52 GMT+09:00 Kentaro Wada <notifications@github.com>:
For start-jsk/jsk_apc#2042
<start-jsk/jsk_apc#2042>
@k-okada <https://github.com/k-okada>
Actually, installing pytorch by rosdep is difficult, because of its
dependency on cuda.
So I decided to put install script: jsk_perception/scripts/
install_pytorch.sh
------------------------------
You can view, comment on, or merge this pull request online at:
#2041
Commit Summary
- FCNObjectSegmentation with PyTorch backend
File Changes
- *M* jsk_perception/node_scripts/fcn_object_segmentation.py
<https://github.com/jsk-ros-pkg/jsk_recognition/pull/2041/files#diff-0>
(115)
- *A* jsk_perception/scripts/install_pytorch.sh
<https://github.com/jsk-ros-pkg/jsk_recognition/pull/2041/files#diff-1>
(15)
Patch Links:
- https://github.com/jsk-ros-pkg/jsk_recognition/pull/2041.patch
- https://github.com/jsk-ros-pkg/jsk_recognition/pull/2041.diff
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#2041>, or mute the
thread
<https://github.com/notifications/unsubscribe-auth/AAeG3I_vuLfEgG0UW_XIkS9ouaerlMt4ks5rpqXQgaJpZM4Mpg5X>
.
|
|
% pip install torch -I
Collecting torch
Using cached torch-0.1.2.post1.tar.gz
Complete output from command python setup.py egg_info:
Traceback (most recent call last):
File "<string>", line 1, in <module>
File "/tmp/pip-build-L8pNGM/torch/setup.py", line 11, in <module>
raise RuntimeError(README)
RuntimeError: PyTorch does not currently provide packages for PyPI (see
status at pytorch/pytorch#566).
Please follow the instructions at http://pytorch.org/ to install with
miniconda instead.
…----------------------------------------
Command "python setup.py egg_info" failed with error code 1 in
/tmp/pip-build-L8pNGM/torch/
On Tue, Mar 28, 2017 at 11:55 AM, Kei Okada ***@***.***> wrote:
what's happens if we run `pip install pytorch`
--
◉ Kei Okada
2017-03-27 2:52 GMT+09:00 Kentaro Wada ***@***.***>:
> For start-jsk/jsk_apc#2042
> <start-jsk/jsk_apc#2042>
>
> @k-okada <https://github.com/k-okada>
> Actually, installing pytorch by rosdep is difficult, because of its
> dependency on cuda.
> So I decided to put install script: jsk_perception/scripts/
> install_pytorch.sh
> ------------------------------
> You can view, comment on, or merge this pull request online at:
>
> #2041
> Commit Summary
>
> - FCNObjectSegmentation with PyTorch backend
>
> File Changes
>
> - *M* jsk_perception/node_scripts/fcn_object_segmentation.py
> <https://github.com/jsk-ros-pkg/jsk_recognition/pull/2041/files#diff-0>
> (115)
> - *A* jsk_perception/scripts/install_pytorch.sh
> <https://github.com/jsk-ros-pkg/jsk_recognition/pull/2041/files#diff-1>
> (15)
>
> Patch Links:
>
> - https://github.com/jsk-ros-pkg/jsk_recognition/pull/2041.patch
> - https://github.com/jsk-ros-pkg/jsk_recognition/pull/2041.diff
>
> —
> You are receiving this because you were mentioned.
> Reply to this email directly, view it on GitHub
> <#2041>, or mute the
> thread
> <https://github.com/notifications/unsubscribe-auth/AAeG3I_vuLfEgG0UW_
XIkS9ouaerlMt4ks5rpqXQgaJpZM4Mpg5X>
> .
>
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
<#2041 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AEHFk3LTwNfcMwkcaxKM_XVRVxqzLlhUks5rqOclgaJpZM4Mpg5X>
.
--
和田 健太郎 / Kentaro Wada
http://wkentaro.com
|
|
… --
◉ Kei Okada
2017-03-28 20:08 GMT+09:00 Kentaro Wada <notifications@github.com>:
% pip install torch -I
Collecting torch
Using cached torch-0.1.2.post1.tar.gz
Complete output from command python setup.py egg_info:
Traceback (most recent call last):
File "<string>", line 1, in <module>
File "/tmp/pip-build-L8pNGM/torch/setup.py", line 11, in <module>
raise RuntimeError(README)
RuntimeError: PyTorch does not currently provide packages for PyPI (see
status at pytorch/pytorch#566).
Please follow the instructions at http://pytorch.org/ to install with
miniconda instead.
----------------------------------------
Command "python setup.py egg_info" failed with error code 1 in
/tmp/pip-build-L8pNGM/torch/
On Tue, Mar 28, 2017 at 11:55 AM, Kei Okada ***@***.***>
wrote:
> what's happens if we run `pip install pytorch`
>
> --
> ◉ Kei Okada
>
>
> 2017-03-27 2:52 GMT+09:00 Kentaro Wada ***@***.***>:
>
> > For start-jsk/jsk_apc#2042
> > <start-jsk/jsk_apc#2042>
> >
> > @k-okada <https://github.com/k-okada>
> > Actually, installing pytorch by rosdep is difficult, because of its
> > dependency on cuda.
> > So I decided to put install script: jsk_perception/scripts/
> > install_pytorch.sh
> > ------------------------------
> > You can view, comment on, or merge this pull request online at:
> >
> > #2041
> > Commit Summary
> >
> > - FCNObjectSegmentation with PyTorch backend
> >
> > File Changes
> >
> > - *M* jsk_perception/node_scripts/fcn_object_segmentation.py
> > <https://github.com/jsk-ros-pkg/jsk_recognition/pull/2041/files#diff-0
>
> > (115)
> > - *A* jsk_perception/scripts/install_pytorch.sh
> > <https://github.com/jsk-ros-pkg/jsk_recognition/pull/2041/files#diff-1
>
> > (15)
> >
> > Patch Links:
> >
> > - https://github.com/jsk-ros-pkg/jsk_recognition/pull/2041.patch
> > - https://github.com/jsk-ros-pkg/jsk_recognition/pull/2041.diff
> >
> > —
> > You are receiving this because you were mentioned.
> > Reply to this email directly, view it on GitHub
> > <#2041>, or mute
the
> > thread
> > <https://github.com/notifications/unsubscribe-auth/AAeG3I_vuLfEgG0UW_
> XIkS9ouaerlMt4ks5rpqXQgaJpZM4Mpg5X>
> > .
> >
>
> —
> You are receiving this because you authored the thread.
> Reply to this email directly, view it on GitHub
> <#2041#
issuecomment-289734131>,
> or mute the thread
> <https://github.com/notifications/unsubscribe-auth/AEHFk3LTwNfcMwkcaxKM_
XVRVxqzLlhUks5rqOclgaJpZM4Mpg5X>
> .
>
--
和田 健太郎 / Kentaro Wada
http://wkentaro.com
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#2041 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AAeG3HjZHTGDrJ0G1DlUBhf-1-y3aU4Kks5rqOoygaJpZM4Mpg5X>
.
|
|
What do you mean?
…On Tue, Mar 28, 2017 at 1:52 PM, Kei Okada ***@***.***> wrote:
pytorch/pytorch#566 (comment) ?
--
◉ Kei Okada
2017-03-28 20:08 GMT+09:00 Kentaro Wada ***@***.***>:
> % pip install torch -I
> Collecting torch
> Using cached torch-0.1.2.post1.tar.gz
> Complete output from command python setup.py egg_info:
> Traceback (most recent call last):
> File "<string>", line 1, in <module>
> File "/tmp/pip-build-L8pNGM/torch/setup.py", line 11, in <module>
> raise RuntimeError(README)
> RuntimeError: PyTorch does not currently provide packages for PyPI (see
> status at pytorch/pytorch#566).
>
> Please follow the instructions at http://pytorch.org/ to install with
> miniconda instead.
>
>
> ----------------------------------------
> Command "python setup.py egg_info" failed with error code 1 in
> /tmp/pip-build-L8pNGM/torch/
>
>
> On Tue, Mar 28, 2017 at 11:55 AM, Kei Okada ***@***.***>
> wrote:
>
> > what's happens if we run `pip install pytorch`
> >
> > --
> > ◉ Kei Okada
> >
> >
> > 2017-03-27 2:52 GMT+09:00 Kentaro Wada ***@***.***>:
> >
> > > For start-jsk/jsk_apc#2042
> > > <start-jsk/jsk_apc#2042>
> > >
> > > @k-okada <https://github.com/k-okada>
> > > Actually, installing pytorch by rosdep is difficult, because of its
> > > dependency on cuda.
> > > So I decided to put install script: jsk_perception/scripts/
> > > install_pytorch.sh
> > > ------------------------------
> > > You can view, comment on, or merge this pull request online at:
> > >
> > > #2041
> > > Commit Summary
> > >
> > > - FCNObjectSegmentation with PyTorch backend
> > >
> > > File Changes
> > >
> > > - *M* jsk_perception/node_scripts/fcn_object_segmentation.py
> > > <#2041/
files#diff-0
> >
> > > (115)
> > > - *A* jsk_perception/scripts/install_pytorch.sh
> > > <#2041/
files#diff-1
> >
> > > (15)
> > >
> > > Patch Links:
> > >
> > > - https://github.com/jsk-ros-pkg/jsk_recognition/pull/2041.patch
> > > - https://github.com/jsk-ros-pkg/jsk_recognition/pull/2041.diff
> > >
> > > —
> > > You are receiving this because you were mentioned.
> > > Reply to this email directly, view it on GitHub
> > > <#2041>, or mute
> the
> > > thread
> > > <https://github.com/notifications/unsubscribe-
auth/AAeG3I_vuLfEgG0UW_
> > XIkS9ouaerlMt4ks5rpqXQgaJpZM4Mpg5X>
> > > .
> > >
> >
> > —
> > You are receiving this because you authored the thread.
> > Reply to this email directly, view it on GitHub
> > <#2041#
> issuecomment-289734131>,
> > or mute the thread
> > <https://github.com/notifications/unsubscribe-
auth/AEHFk3LTwNfcMwkcaxKM_
> XVRVxqzLlhUks5rqOclgaJpZM4Mpg5X>
> > .
> >
>
>
>
> --
> 和田 健太郎 / Kentaro Wada
> http://wkentaro.com
>
> —
> You are receiving this because you were mentioned.
> Reply to this email directly, view it on GitHub
> <#2041#
issuecomment-289736749>,
> or mute the thread
> <https://github.com/notifications/unsubscribe-
auth/AAeG3HjZHTGDrJ0G1DlUBhf-1-y3aU4Kks5rqOoygaJpZM4Mpg5X>
> .
>
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
<#2041 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AEHFk_Jf4hUA44JEqu64q3FNsKcFVcU9ks5rqQKjgaJpZM4Mpg5X>
.
--
和田 健太郎 / Kentaro Wada
http://wkentaro.com
|
|
try
```
pytoach-pip:
ubuntu:
pip:
http://download.pytorch.org/whl/torch-0.1.10.post1-cp35-cp35m-macosx_10_6_x86_64.whl
```
```
s$ rosdep install --from-paths .
executing command [sudo -H pip install -U
http://download.pytorch.org/whl/torch-0.1.10.post1-cp35-cp35m-macosx_10_6_x86_64.whl
]
torch-0.1.10.post1-cp35-cp35m-macosx_10_6_x86_64.whl is not a supported
wheel on this platform.
/usr/local/lib/python2.7/dist-packages/pip/_vendor/requests/packages/urllib3/util/ssl_.py:318:
SNIMissingWarning: An HTTPS request has been made, but the SNI (Subject
Name Indication) extension to TLS is not available on this platform. This
may cause the server to present an incorrect TLS certificate, which can
cause validation failures. You can upgrade to a newer version of Python to
solve this. For more information, see
https://urllib3.readthedocs.io/en/latest/security.html#snimissingwarning.
SNIMissingWarning
/usr/local/lib/python2.7/dist-packages/pip/_vendor/requests/packages/urllib3/util/ssl_.py:122:
InsecurePlatformWarning: A true SSLContext object is not available. This
prevents urllib3 from configuring SSL appropriately and may cause certain
SSL connections to fail. You can upgrade to a newer version of Python to
solve this. For more information, see
https://urllib3.readthedocs.io/en/latest/security.html#insecureplatformwarning
.
InsecurePlatformWarning
ERROR: the following rosdeps failed to install
pip: command [sudo -H pip install -U
http://download.pytorch.org/whl/torch-0.1.10.post1-cp35-cp35m-macosx_10_6_x86_64.whl]
failed
```
…--
◉ Kei Okada
2017-03-29 6:28 GMT+09:00 Kentaro Wada <notifications@github.com>:
What do you mean?
On Tue, Mar 28, 2017 at 1:52 PM, Kei Okada ***@***.***>
wrote:
> pytorch/pytorch#566 (comment) ?
>
> --
> ◉ Kei Okada
>
>
>
> 2017-03-28 20:08 GMT+09:00 Kentaro Wada ***@***.***>:
>
> > % pip install torch -I
> > Collecting torch
> > Using cached torch-0.1.2.post1.tar.gz
> > Complete output from command python setup.py egg_info:
> > Traceback (most recent call last):
> > File "<string>", line 1, in <module>
> > File "/tmp/pip-build-L8pNGM/torch/setup.py", line 11, in <module>
> > raise RuntimeError(README)
> > RuntimeError: PyTorch does not currently provide packages for PyPI (see
> > status at pytorch/pytorch#566).
> >
> > Please follow the instructions at http://pytorch.org/ to install with
> > miniconda instead.
> >
> >
> > ----------------------------------------
> > Command "python setup.py egg_info" failed with error code 1 in
> > /tmp/pip-build-L8pNGM/torch/
> >
> >
> > On Tue, Mar 28, 2017 at 11:55 AM, Kei Okada ***@***.***>
> > wrote:
> >
> > > what's happens if we run `pip install pytorch`
> > >
> > > --
> > > ◉ Kei Okada
> > >
> > >
> > > 2017-03-27 2:52 GMT+09:00 Kentaro Wada ***@***.***>:
> > >
> > > > For start-jsk/jsk_apc#2042
> > > > <start-jsk/jsk_apc#2042>
> > > >
> > > > @k-okada <https://github.com/k-okada>
> > > > Actually, installing pytorch by rosdep is difficult, because of its
> > > > dependency on cuda.
> > > > So I decided to put install script: jsk_perception/scripts/
> > > > install_pytorch.sh
> > > > ------------------------------
> > > > You can view, comment on, or merge this pull request online at:
> > > >
> > > > #2041
> > > > Commit Summary
> > > >
> > > > - FCNObjectSegmentation with PyTorch backend
> > > >
> > > > File Changes
> > > >
> > > > - *M* jsk_perception/node_scripts/fcn_object_segmentation.py
> > > > <#2041/
> files#diff-0
> > >
> > > > (115)
> > > > - *A* jsk_perception/scripts/install_pytorch.sh
> > > > <#2041/
> files#diff-1
> > >
> > > > (15)
> > > >
> > > > Patch Links:
> > > >
> > > > - https://github.com/jsk-ros-pkg/jsk_recognition/pull/2041.patch
> > > > - https://github.com/jsk-ros-pkg/jsk_recognition/pull/2041.diff
> > > >
> > > > —
> > > > You are receiving this because you were mentioned.
> > > > Reply to this email directly, view it on GitHub
> > > > <#2041>, or
mute
> > the
> > > > thread
> > > > <https://github.com/notifications/unsubscribe-
> auth/AAeG3I_vuLfEgG0UW_
> > > XIkS9ouaerlMt4ks5rpqXQgaJpZM4Mpg5X>
> > > > .
> > > >
> > >
> > > —
> > > You are receiving this because you authored the thread.
> > > Reply to this email directly, view it on GitHub
> > > <#2041#
> > issuecomment-289734131>,
> > > or mute the thread
> > > <https://github.com/notifications/unsubscribe-
> auth/AEHFk3LTwNfcMwkcaxKM_
> > XVRVxqzLlhUks5rqOclgaJpZM4Mpg5X>
> > > .
> > >
> >
> >
> >
> > --
> > 和田 健太郎 / Kentaro Wada
> > http://wkentaro.com
> >
> > —
> > You are receiving this because you were mentioned.
> > Reply to this email directly, view it on GitHub
> > <#2041#
> issuecomment-289736749>,
> > or mute the thread
> > <https://github.com/notifications/unsubscribe-
> auth/AAeG3HjZHTGDrJ0G1DlUBhf-1-y3aU4Kks5rqOoygaJpZM4Mpg5X>
> > .
> >
>
> —
> You are receiving this because you authored the thread.
> Reply to this email directly, view it on GitHub
> <#2041#
issuecomment-289759801>,
> or mute the thread
> <https://github.com/notifications/unsubscribe-auth/AEHFk_
Jf4hUA44JEqu64q3FNsKcFVcU9ks5rqQKjgaJpZM4Mpg5X>
> .
>
--
和田 健太郎 / Kentaro Wada
http://wkentaro.com
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#2041 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AAeG3HKJsJ-ZucXzUlu-6ZYlvIassvjbks5rqXtggaJpZM4Mpg5X>
.
|
|
The URL is for mac, and not work for ubuntu right?
|
|
yes , i thought there is a similar url for ubuntu
…--
◉ Kei Okada
2017-03-29 15:32 GMT+09:00 Kentaro Wada <notifications@github.com>:
The URL is for mac, and not work for ubuntu right?
2017年3月29日(水) 3:38 Kei Okada ***@***.***>:
try
```
pytoach-pip:
ubuntu:
pip:
http://download.pytorch.org/whl/torch-0.1.10.post1-cp35-
cp35m-macosx_10_6_x86_64.whl
```
```
s$ rosdep install --from-paths .
executing command [sudo -H pip install -U
http://download.pytorch.org/whl/torch-0.1.10.post1-cp35-
cp35m-macosx_10_6_x86_64.whl
]
torch-0.1.10.post1-cp35-cp35m-macosx_10_6_x86_64.whl is not a supported
wheel on this platform.
/usr/local/lib/python2.7/dist-packages/pip/_vendor/requests/
packages/urllib3/util/ssl_.py:318:
SNIMissingWarning: An HTTPS request has been made, but the SNI (Subject
Name Indication) extension to TLS is not available on this platform. This
may cause the server to present an incorrect TLS certificate, which can
cause validation failures. You can upgrade to a newer version of Python to
solve this. For more information, see
https://urllib3.readthedocs.io/en/latest/security.html#snimissingwarning.
SNIMissingWarning
/usr/local/lib/python2.7/dist-packages/pip/_vendor/requests/
packages/urllib3/util/ssl_.py:122:
InsecurePlatformWarning: A true SSLContext object is not available. This
prevents urllib3 from configuring SSL appropriately and may cause certain
SSL connections to fail. You can upgrade to a newer version of Python to
solve this. For more information, see
https://urllib3.readthedocs.io/en/latest/security.html#
insecureplatformwarning
.
InsecurePlatformWarning
ERROR: the following rosdeps failed to install
pip: command [sudo -H pip install -U
http://download.pytorch.org/whl/torch-0.1.10.post1-cp35-
cp35m-macosx_10_6_x86_64.whl
]
failed
```
--
◉ Kei Okada
2017-03-29 6:28 GMT+09:00 Kentaro Wada ***@***.***>:
> What do you mean?
>
>
> On Tue, Mar 28, 2017 at 1:52 PM, Kei Okada ***@***.***>
> wrote:
>
> > pytorch/pytorch#566 (comment) ?
> >
> > --
> > ◉ Kei Okada
> >
> >
> >
> > 2017-03-28 20:08 GMT+09:00 Kentaro Wada ***@***.***>:
> >
> > > % pip install torch -I
> > > Collecting torch
> > > Using cached torch-0.1.2.post1.tar.gz
> > > Complete output from command python setup.py egg_info:
> > > Traceback (most recent call last):
> > > File "<string>", line 1, in <module>
> > > File "/tmp/pip-build-L8pNGM/torch/setup.py", line 11, in <module>
> > > raise RuntimeError(README)
> > > RuntimeError: PyTorch does not currently provide packages for PyPI
(see
> > > status at pytorch/pytorch#566).
> > >
> > > Please follow the instructions at http://pytorch.org/ to install
with
> > > miniconda instead.
> > >
> > >
> > > ----------------------------------------
> > > Command "python setup.py egg_info" failed with error code 1 in
> > > /tmp/pip-build-L8pNGM/torch/
> > >
> > >
> > > On Tue, Mar 28, 2017 at 11:55 AM, Kei Okada <
***@***.***>
> > > wrote:
> > >
> > > > what's happens if we run `pip install pytorch`
> > > >
> > > > --
> > > > ◉ Kei Okada
> > > >
> > > >
> > > > 2017-03-27 2:52 GMT+09:00 Kentaro Wada ***@***.***>:
> > > >
> > > > > For start-jsk/jsk_apc#2042
> > > > > <start-jsk/jsk_apc#2042>
> > > > >
> > > > > @k-okada <https://github.com/k-okada>
> > > > > Actually, installing pytorch by rosdep is difficult, because of
its
> > > > > dependency on cuda.
> > > > > So I decided to put install script: jsk_perception/scripts/
> > > > > install_pytorch.sh
> > > > > ------------------------------
> > > > > You can view, comment on, or merge this pull request online at:
> > > > >
> > > > > #2041
> > > > > Commit Summary
> > > > >
> > > > > - FCNObjectSegmentation with PyTorch backend
> > > > >
> > > > > File Changes
> > > > >
> > > > > - *M* jsk_perception/node_scripts/fcn_object_segmentation.py
> > > > > <#2041/
> > files#diff-0
> > > >
> > > > > (115)
> > > > > - *A* jsk_perception/scripts/install_pytorch.sh
> > > > > <#2041/
> > files#diff-1
> > > >
> > > > > (15)
> > > > >
> > > > > Patch Links:
> > > > >
> > > > > - https://github.com/jsk-ros-pkg/jsk_recognition/pull/2041.patch
> > > > > - https://github.com/jsk-ros-pkg/jsk_recognition/pull/2041.diff
> > > > >
> > > > > —
> > > > > You are receiving this because you were mentioned.
> > > > > Reply to this email directly, view it on GitHub
> > > > > <#2041>, or
> mute
> > > the
> > > > > thread
> > > > > <https://github.com/notifications/unsubscribe-
> > auth/AAeG3I_vuLfEgG0UW_
> > > > XIkS9ouaerlMt4ks5rpqXQgaJpZM4Mpg5X>
> > > > > .
> > > > >
> > > >
> > > > —
> > > > You are receiving this because you authored the thread.
> > > > Reply to this email directly, view it on GitHub
> > > > <#2041#
> > > issuecomment-289734131>,
> > > > or mute the thread
> > > > <https://github.com/notifications/unsubscribe-
> > auth/AEHFk3LTwNfcMwkcaxKM_
> > > XVRVxqzLlhUks5rqOclgaJpZM4Mpg5X>
> > > > .
> > > >
> > >
> > >
> > >
> > > --
> > > 和田 健太郎 / Kentaro Wada
> > > http://wkentaro.com
> > >
> > > —
> > > You are receiving this because you were mentioned.
> > > Reply to this email directly, view it on GitHub
> > > <#2041#
> > issuecomment-289736749>,
> > > or mute the thread
> > > <https://github.com/notifications/unsubscribe-
> > auth/AAeG3HjZHTGDrJ0G1DlUBhf-1-y3aU4Kks5rqOoygaJpZM4Mpg5X>
> > > .
> > >
> >
> > —
> > You are receiving this because you authored the thread.
> > Reply to this email directly, view it on GitHub
> > <#2041#
> issuecomment-289759801>,
> > or mute the thread
> > <https://github.com/notifications/unsubscribe-auth/AEHFk_
> Jf4hUA44JEqu64q3FNsKcFVcU9ks5rqQKjgaJpZM4Mpg5X>
> > .
> >
>
>
>
> --
> 和田 健太郎 / Kentaro Wada
> http://wkentaro.com
>
> —
> You are receiving this because you were mentioned.
> Reply to this email directly, view it on GitHub
> <
#2041#
issuecomment-289910328
>,
> or mute the thread
> <
https://github.com/notifications/unsubscribe-auth/AAeG3HKJsJ-ZucXzUlu-
6ZYlvIassvjbks5rqXtggaJpZM4Mpg5X
>
> .
>
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
<#2041#
issuecomment-289964607>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/
AEHFkxPZ2ALsbJ3aePOZh82kslSddHMtks5rqcQdgaJpZM4Mpg5X>
.
--
和田 健太郎 / Kentaro Wada
http://wkentaro.com
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#2041 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AAeG3L9yUbq-3id8HBAKvssJgaoglN-Uks5rqfsRgaJpZM4Mpg5X>
.
|
|
Yeah you can see the installation script in this PR.
|
|
I need if condition to detect cuda and its version: https://github.com/wkentaro/jsk_recognition/blob/23efa116615f0fb7cf79a17a430c7656c3ad3034/jsk_perception/scripts/install_pytorch.sh |
|
i think we can assume everyone have latest? cuda or
assume everyone already install cuda by
ros/rosdistro#13782
…--
◉ Kei Okada
2017-03-29 17:40 GMT+09:00 Kentaro Wada <notifications@github.com>:
I need if condition to detect cuda and its version:
https://github.com/wkentaro/jsk_recognition/blob/
23efa11/jsk_perception/
scripts/install_pytorch.sh
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#2041 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AAeG3OtjAKUygkm7bQQ1iRDNcFCBdTY3ks5rqhjxgaJpZM4Mpg5X>
.
|
|
Which version of cuda is installed with apt? 7.5 or 8.0? |
|
http://packages.ubuntu.com/trusty/nvidia-cuda-toolkit
5 on T, 7.5 on X, 8 on YZ
…--
◉ Kei Okada
2017-03-29 22:51 GMT+09:00 Kentaro Wada <notifications@github.com>:
Which version of cuda is installed with apt? 7.5 or 8.0?
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#2041 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AAeG3Dd1UhGfPh7nna33LmRhOKse6pdaks5rqmHOgaJpZM4Mpg5X>
.
|
|
Then, we should set nocuda as the default? |
|
5 or 8
…--
◉ Kei Okada
2017-03-29 23:11 GMT+09:00 Kentaro Wada <notifications@github.com>:
Then, we should no-cuda as the default?
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#2041 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AAeG3OhAmDS8N5A4WPlYJ8Q3f1idFGehks5rqmZ2gaJpZM4Mpg5X>
.
|
|
But you said 8 was not for T, didn't you? |
|
So, we should set |
|
Do we have pytorch for 5?
If we set 8 for all ununtu distro, we assume everyone has installed 8.
If we set 5 for T, .... then we assume cuda is installed rosdep key.
Choose one you think you are comfortable when you asked questions
2017年3月29日(水) 23:27 Kentaro Wada <notifications@github.com>:
So, we should set
http://download.pytorch.org/whl/cu80/torch-0.1.10.post2-cp27-none-linux_x86_64.whl
for download target of python-torch rosdep key?
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#2041 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AAeG3HrHoWbfdaPa-oMMKVOumx2rfivJks5rqmo5gaJpZM4Mpg5X>
.
--
--
◉ Kei Okada
|
|
PyTorch supports only cuda 7.5 and 8.0, not 5. |
|
So possibly, current best choice using rosdep is |
|
But I think it is unlikely that the PR was merged if I send to rosdistro, because it is environment specific cuda-80. |
|
Enable to install pytorch with rosdep and how do we use them within
jsk_perception is different topic,
On rosdep, I think your approach would be nice, there are huge possibility
that the PR with http path is not accepted, but I think it is nice to show
our use cases.
On jsk_perception, it our decision to assume user environment, and my
though on this is ask users to `apt-get install nvidia-cuda-toolkit` or ask
them to "install cuda8", even if someone using cuda7.5.
For example, we're strongly assuming that everyone using 14.04 / Indigo for
jsk-ros-pkg, start-jsk, and if someone using 16.06 Kinetic, we'll assuming
those people
should fix their problem by themselves, very quickly.
…--
◉ Kei Okada
2017-03-30 1:42 GMT+09:00 Kentaro Wada <notifications@github.com>:
But I think it is unlikely that the PR was merged if I send to rosdistro,
because it is environment specific cuda-80.
Another idea was adding python-torch-cuda80-pip and
python-torch-cuda75-pip but it is useless because we can't use if
condition in package.xml.
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#2041 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AAeG3LDvEVv6Jk9PvuMLXwkFYC4rt4Scks5rqooKgaJpZM4Mpg5X>
.
|
IMO, I don't think so, and most of programs work on hydro, indigo, jade, and (kinetic). And I assume Ubuntu 12.04,14.04 (macOS is optional for me). But it's ok. |
|
what happens if we have cuda-9, cuda-10, cuda-11, do I have to continue
updating this script?
…--
◉ Kei Okada
2017-03-30 15:28 GMT+09:00 Kentaro Wada <notifications@github.com>:
For example, we're strongly assuming that everyone using 14.04 / Indigo for
jsk-ros-pkg, start-jsk, and if someone using 16.06 Kinetic, we'll assuming
those people
should fix their problem by themselves, very quickly.
IMO, I don't think so, and most of programs work on hydro, indigo, jade,
and (kinetic). And I assume Ubuntu 12.04,14.04 (macOS is optional for me).
But it's ok.
My objective is the use of PyTorch in jsk_perception, and maybe I can
abandon the automatic installation by rosdep or a script.
I'm not sure why you would like to use a single url not python script?
if we write
if (EXISTS /usr/local/cuda-8.0)
add_custom_command(COMMAND pip install http://torch-cuda-80 DEPENDS /usr/local/lib/python2.7/dist-packages/torch)
else if (EXISTS /usr/local/cuda-7.5)
add_custom_command(COMMAND pip install http://torch-cuda-75 DEPENDS /usr/local/lib/python2.7/dist-packages/torch)
else
add_custom_command(COMMAND pip install http://torch-cuda-75 DEPENDS /usr/local/lib/python2.7/dist-packages/torch)
endif
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#2041 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AAeG3OTLIryHpt_86sr1xXdJv6gpFyDdks5rq0uBgaJpZM4Mpg5X>
.
|
|
Yeah.
And you can automate that kind of thing if it is needed in the script.
for example
def install():
for i in xrange(7, 100):
for j in [0, 5]:
if os.path.exists('/usr/local/cuda-%d.%d' % (i, j)):
subprocess.call('pip install http://torch-cuda-%d.%d' % (i,
j), shell=True)
return
subprocess.call('pip install http://torch-no-cuda', shell=True)
…On Thu, Mar 30, 2017 at 11:48 AM, Kei Okada ***@***.***> wrote:
what happens if we have cuda-9, cuda-10, cuda-11, do I have to continue
updating this script?
--
◉ Kei Okada
2017-03-30 15:28 GMT+09:00 Kentaro Wada ***@***.***>:
> For example, we're strongly assuming that everyone using 14.04 / Indigo
for
> jsk-ros-pkg, start-jsk, and if someone using 16.06 Kinetic, we'll
assuming
> those people
> should fix their problem by themselves, very quickly.
>
> IMO, I don't think so, and most of programs work on hydro, indigo, jade,
> and (kinetic). And I assume Ubuntu 12.04,14.04 (macOS is optional for
me).
>
> But it's ok.
> My objective is the use of PyTorch in jsk_perception, and maybe I can
> abandon the automatic installation by rosdep or a script.
> I'm not sure why you would like to use a single url not python script?
> if we write
>
> if (EXISTS /usr/local/cuda-8.0)
> add_custom_command(COMMAND pip install http://torch-cuda-80 DEPENDS
/usr/local/lib/python2.7/dist-packages/torch)
> else if (EXISTS /usr/local/cuda-7.5)
> add_custom_command(COMMAND pip install http://torch-cuda-75 DEPENDS
/usr/local/lib/python2.7/dist-packages/torch)
> else
> add_custom_command(COMMAND pip install http://torch-cuda-75 DEPENDS
/usr/local/lib/python2.7/dist-packages/torch)
> endif
>
> —
> You are receiving this because you were mentioned.
> Reply to this email directly, view it on GitHub
> <#2041#
issuecomment-290314770>,
> or mute the thread
> <https://github.com/notifications/unsubscribe-auth/AAeG3OTLIryHpt_
86sr1xXdJv6gpFyDdks5rq0uBgaJpZM4Mpg5X>
> .
>
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
<#2041 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AEHFkw1zjzsikxx1QY--KZ7uFpVPOailks5rq4hqgaJpZM4Mpg5X>
.
--
和田 健太郎 / Kentaro Wada
http://wkentaro.com
|
|
My perspective on current situation is;
5 % users can install cuda and pytoach by them self and if there is an
problem they can fix by themselves, so we do not have to care about them,
for example. people like @wkentaro / @furushchev
Rest of people try to install by common procedure of ros systems , and
fall into trouble like
#1874, so for those
people, instead of take care about them for each different environment,
we'll encourage/assume them to use thinkpad, ubuntu, 14.04, .... and cuda
X. because we do not have enough resource to let them to install any cuda
version and support for each environment.
So what I'm talking here is we should advice them to install cuda8 (or 7.5)
by themselves or cuda 5 with nvidia-cuda-toolkit.
It is similar to lecture environment, soft2/soft3 shuuld run on any kind of
unix systmes, but I assume every one use T440p and 14.04. If someone have
trouble with that environment, it's my fault and my responsibility. I'll do
my best to fix the problem, and if they can not finish their homework with
that environment, because of system trouble. I'll do not immediately mark
D- for that.
But if a student ask me if he have trouble on OSX/vim environment and
could not submit homework because of that. I feel I do not have responsible
to that situation. I like these challenging students and I'll try to fix
his problem, but that is just for fun. If we can not fix that problem, I'd
advice him to use normal environment, T440 and 14.04 and emacs to work on
his homework within due.
…--
◉ Kei Okada
2017-03-30 20:01 GMT+09:00 Kentaro Wada <notifications@github.com>:
Yeah.
And you can automate that kind of thing if it is needed in the script.
for example
def install():
for i in xrange(7, 100):
for j in [0, 5]:
if os.path.exists('/usr/local/cuda-%d.%d' % (i, j)):
subprocess.call('pip install http://torch-cuda-%d.%d' % (i,
j), shell=True)
return
subprocess.call('pip install http://torch-no-cuda', shell=True)
On Thu, Mar 30, 2017 at 11:48 AM, Kei Okada ***@***.***>
wrote:
> what happens if we have cuda-9, cuda-10, cuda-11, do I have to continue
> updating this script?
>
> --
> ◉ Kei Okada
>
>
> 2017-03-30 15:28 GMT+09:00 Kentaro Wada ***@***.***>:
>
> > For example, we're strongly assuming that everyone using 14.04 / Indigo
> for
> > jsk-ros-pkg, start-jsk, and if someone using 16.06 Kinetic, we'll
> assuming
> > those people
> > should fix their problem by themselves, very quickly.
> >
> > IMO, I don't think so, and most of programs work on hydro, indigo,
jade,
> > and (kinetic). And I assume Ubuntu 12.04,14.04 (macOS is optional for
> me).
> >
> > But it's ok.
> > My objective is the use of PyTorch in jsk_perception, and maybe I can
> > abandon the automatic installation by rosdep or a script.
> > I'm not sure why you would like to use a single url not python script?
> > if we write
> >
> > if (EXISTS /usr/local/cuda-8.0)
> > add_custom_command(COMMAND pip install http://torch-cuda-80 DEPENDS
> /usr/local/lib/python2.7/dist-packages/torch)
> > else if (EXISTS /usr/local/cuda-7.5)
> > add_custom_command(COMMAND pip install http://torch-cuda-75 DEPENDS
> /usr/local/lib/python2.7/dist-packages/torch)
> > else
> > add_custom_command(COMMAND pip install http://torch-cuda-75 DEPENDS
> /usr/local/lib/python2.7/dist-packages/torch)
> > endif
> >
> > —
> > You are receiving this because you were mentioned.
> > Reply to this email directly, view it on GitHub
> > <#2041#
> issuecomment-290314770>,
> > or mute the thread
> > <https://github.com/notifications/unsubscribe-auth/AAeG3OTLIryHpt_
> 86sr1xXdJv6gpFyDdks5rq0uBgaJpZM4Mpg5X>
> > .
> >
>
> —
> You are receiving this because you authored the thread.
> Reply to this email directly, view it on GitHub
> <#2041#
issuecomment-290375139>,
> or mute the thread
> <https://github.com/notifications/unsubscribe-auth/AEHFkw1zjzsikxx1QY--
KZ7uFpVPOailks5rq4hqgaJpZM4Mpg5X>
> .
>
--
和田 健太郎 / Kentaro Wada
http://wkentaro.com
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#2041 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AAeG3P7juT8sdydVjHKz4I1labJSdZkgks5rq4t7gaJpZM4Mpg5X>
.
|
|
So always installing pytorch-cuda80 can be assumed. |
|
Depends on ros/rosdistro#14398 |
|
Installing pytorch by rosdep is currently not possible: ros/rosdistro#14398 (review) |
- new file: install_pytorch.py - deleted: install_pytorch.sh
|
|
I suggest an installation process above, and what do you think about that? |
After making some improvements on Chainer code, I found that Chainer one is faster than Pytorch implementation. Please take this into account. |
|
Yeah, Chainer's Relu was just a mess.. |
|
If you are concerned with performance, a rule of thumb is to profile the code by nvvp 🔢 |
|
@k-okada Well, what do you think about how to install pytorch? Can I just raise error message at runtime, to prompt user to install it manually? |
|
Please check |
|
note that I'm very serious on this question, we'd like to know how many DGX
machine needed for this purpose.
…--
◉ Kei Okada
2017-04-12 10:12 GMT+09:00 Kei Okada <notifications@github.com>:
Merged #2041 <#2041>.
—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
<#2041 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AAeG3HW-MAU2aO_bm3TqqUPt6CDNba_nks5rvCTlgaJpZM4Mpg5X>
.
|
|
Ok, I'll analyze it. |
1 task
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
For start-jsk/jsk_apc#2042
Why PyTorch?
Its much faster than Chainer.By using Chainer, we get 1Hz output with VGA image input, but we get 5Hz output by using PyTorch.For FCN with improvement of Relu implementation, now it is as fast as PyTorch: (See https://github.com/wkentaro/pytorch-fcn#speed)
I am currently mainly using PyTorch for its functionality: functions, layers and multi-gpu training.
Why not rosdep to install PyTorch?
@k-okada
Actually, installing pytorch by rosdep is difficult, because of its dependency on cuda.
So I decided to put install script:
jsk_perception/scripts/install_pytorch.sh