Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Speedup theano generators #267

Merged
merged 5 commits into from
Oct 15, 2015
Merged

Conversation

oliverlee
Copy link
Contributor

I've rebased this PR: #70
since it was made over a year ago.

@moorepants anything that would keep this from being merged?

@moorepants
Copy link
Member

Maybe we should only run the tests with theano on the newest released dep versions due to the conflict with NumPy 1.7.

@moorepants
Copy link
Member

This is the current offering for Theano from the anaconda repos:

moorepants@caramelmonkey:~$ conda search theano
Fetching package metadata: ......
theano                       0.5.0                np17py27_0  http://repo.continuum.io/pkgs/free/linux-64/ 
                             0.5.0                np17py26_0  http://repo.continuum.io/pkgs/free/linux-64/ 
                             0.5.0                np16py27_0  http://repo.continuum.io/pkgs/free/linux-64/ 
                             0.5.0                np16py26_0  http://repo.continuum.io/pkgs/free/linux-64/ 
                             0.5.0                np15py27_0  http://repo.continuum.io/pkgs/free/linux-64/ 
                             0.5.0                np15py26_0  http://repo.continuum.io/pkgs/free/linux-64/ 
                             0.5.0                np17py27_1  http://repo.continuum.io/pkgs/free/linux-64/ 
                             0.5.0                np17py26_1  http://repo.continuum.io/pkgs/free/linux-64/ 
                             0.5.0                np16py27_1  http://repo.continuum.io/pkgs/free/linux-64/ 
                             0.5.0                np16py26_1  http://repo.continuum.io/pkgs/free/linux-64/ 
                             0.6.0                np19py27_0  http://repo.continuum.io/pkgs/free/linux-64/ 
                             0.6.0                np19py26_0  http://repo.continuum.io/pkgs/free/linux-64/ 
                             0.6.0                np18py27_0  http://repo.continuum.io/pkgs/free/linux-64/ 
                             0.6.0                np18py26_0  http://repo.continuum.io/pkgs/free/linux-64/ 
                             0.6.0                np17py27_0  http://repo.continuum.io/pkgs/free/linux-64/ 
                             0.6.0                np17py26_0  http://repo.continuum.io/pkgs/free/linux-64/ 
                             0.7.0                np19py35_0  http://repo.continuum.io/pkgs/free/linux-64/ 
                             0.7.0                np19py34_0  http://repo.continuum.io/pkgs/free/linux-64/ 
                             0.7.0                np19py33_0  http://repo.continuum.io/pkgs/free/linux-64/ 
                          .  0.7.0                np19py27_0  http://repo.continuum.io/pkgs/free/linux-64/ 
                             0.7.0                np19py26_0  http://repo.continuum.io/pkgs/free/linux-64/ 
                             0.7.0               np110py35_0  http://repo.continuum.io/pkgs/free/linux-64/ 
                             0.7.0               np110py34_0  http://repo.continuum.io/pkgs/free/linux-64/ 
                             0.7.0               np110py27_0  http://repo.continuum.io/pkgs/free/linux-64/ 

NumPy 1.9 is the lowest Theano 0.7.0 supports. We could build a theano package for NumPy 1.7 and host it in the anaconda.org pydy channel, only test theano on the newest dep versions, or drop support for NumPy < 1.9.

@@ -22,7 +22,7 @@ install:
- source activate test-env
- if [[ $DEP_VERSIONS == "oldest" ]]; then
if [[ $TRAVIS_PYTHON_VERSION == 2.7 ]]; then
conda install numpy==1.7.0 scipy==0.11.0 sympy==0.7.4.1 cython==0.17 theano==0.6.0;
conda install numpy==1.7.0 scipy==0.11.0 sympy==0.7.4.1 cython==0.17 theano==0.7.0;
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I just added a NumPy 1.7 compatible Theano for Linux 64 here:

https://anaconda.org/pydy/theano

You can change this to conda install -c pydy numpy==1.7.0 scipy==0.11.0 sympy==0.7.4.1 cython==0.17 theano==0.7.0; to pull packages from the main anaconda repos and our pydy channel. I think that will fix this.

@moorepants
Copy link
Member

Please update with master.

@oliverlee oliverlee force-pushed the speedup-theano-generators branch from 31d84b7 to 88903d0 Compare October 13, 2015 11:42
nouiz and others added 4 commits October 13, 2015 13:42
Reduction of theano_function overhead requires Theano 0.7.0.
Bump modules to supported versions by Theano 0.7.0 in Anaconda.
@oliverlee oliverlee force-pushed the speedup-theano-generators branch from 14469cf to abfc6e7 Compare October 15, 2015 07:29
@oliverlee
Copy link
Contributor Author

NumPy 1.7 and SciPy 0.11 were released Feb 2013 and Sept 2012 respectively. NumPy 1.9 and Scipy 0.14 were released in Sept and May 2014. I suggest we drop support for earlier versions.

@moorepants
Copy link
Member

Moved version discussion here: #271

Let's try to merge this as is for now.

oliverlee added a commit that referenced this pull request Oct 15, 2015
@oliverlee oliverlee merged commit eaf45c6 into pydy:master Oct 15, 2015
@oliverlee oliverlee deleted the speedup-theano-generators branch October 17, 2015 11:17
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants