I want to use GPflow's AdamOptimizer in my python code.
As mentioned in the previous post, gpflow is working with a certain version of the TensorFlow.
I followed the answer in post, I tried three different combinations inside the virtual environment.
Tensorflow GPflow
1.5 1.4
1.13.0 1.2.0
1.11.0 1.2.0
I still can't implement:
gpflow.train.AdamOptimizer
Error is: Cannot find the reference in the init.py
Do you have any suggestions for me?
P.S. I think this question is not the duplicate of post, because the answer of that post is not working for me.
The combinations you suggest don't make sense - tensorflow 1.5 is really old, and we did not bother staying backwards compatible given how fast tensorflow is moving. Likewise, gpflow 1.2 is quite old at this point and can't be expected to work with more recent tensorflow versions.
If you look at the release notes, you'll see that GPflow 1.4 is the first version that supports tensorflow 1.12 and 1.13. We fixed support for tensorflow 1.14 in the develop branch (install as suggested in Mark van der Wilk's comment), but this has not yet been released as a PyPI/pip package. (GPflow 1.3 should support tensorflow 1.11, but I've not tested that out.)
Tensorflow 1.11.0 with GPflow 1.2.0 version is compatible.
Related
trying to run openpose on windows.
tensorflow = 1.14 installed already.
getting the error no module found : tensorflow.contrib
is there any way to work out on this on windows?
or any other library to work for pose detection?
Please suggest a alternate way to do pose estimation if there is no solution to this problem for windows.
Thanks.
You got above error because tensorflow.contrib has been removed from the packages.
Since all the projects in tf.contrib were not officially supported by Tensorflow and it had designated owners for maintaining it.
From Tensorflow 2.x version all the contrib projects has three options: move to core, move to a separate repository or delete most of of them have been reviewed with the respective project owners.
If the library you are using is moved to the core or separate repository, then tensorflow automatic code migration from 1.x to 2.x will not work for tf.contrib` projects. You have to change code manually for above parts, which is suggested going forward.
You can refer this article for Pose Estimation using TensorFlow 2.0.
The newest AnyLogic version is not supported by my current AnyLogic license. The license only supports up to 8.3.3. However, I do wish to install AnyLogic on another computer and activate it with my license. However, surprisingly I could not find downloads of earlier AnyLogic versions. Does anyone know if and where I can find these earlier versions, and whether I perhaps can downgrade a newer AnyLogic version?
https://www.anylogic.com/files/anylogic-professional-8.3.3.exe
For any version, just put the version you want and you will likely be able to download it
if using mac:
https://www.anylogic.com/files/anylogic-professional-8.3.3.dmg
In addition to Felipe's answer, you can always ask
support#anylogic.com
if you need very old versions. I believe that AL7.x is not available online anymore but they happily send the installers if you need them.
I have a similar issue to others with version numbers in the App Store being unable to upload but don't see a solution for mine.
I created version 1.21, which is a published app. I now realize that the format is supposed to be 1.2.1
Then I tried to create version 1.3 but it says the 1.21 is higher.
How can I get back on the 1.x version numbering format?
Thanks in advance,
~eric
~eric
1.21 is obvisously higher than 1.3.
I think your only option would be version number 1.22 or higher.
I am trying to locate the last version of Caffe that was still compatible with cuDNN v2. The current master requires cuDNN v3. I've been searching through the github repository for Caffe trying to locate what I need but there doesn't seem to be a clear way of doing this. Is there a relatively painless way of finding this on github ?
Here is a list of Caffe release, and rc2 is the one you are looking for.
I'm using the academic version of enthought python for Windows. Most of ogr works, but not geometry methods like geometry.Buffer() or predicates like geometry1.crosses(geometry2). According to this:
http://gdal.org/python/osgeo.ogr.Geometry-class.html#Crosses
ogr needs to have been built with GEOS. Has anybody got these methods to work with Enthought? If not, can you suggest a Windows python binary where these are working?
thanks,
jim
Correct, there are many possible build configurations of GDAL/OGR, and geos is not currently supported by Enthought's build configuration. We will consider adding it in a future build. Meanwhile, FYI, GEOS operations are fully supported in the powerful and popular Shapely package which is included in the Enthought repository.