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
Rebuild conda packages #556
Comments
Yes. Will release 2.3.3 today, so hold tight. |
It would be great to have x64 windows version as well. :) |
I think 64bit windows has problems with fortran. I can do 32 bit though. |
If you are up for experimenting, I saw some hints on 64bit windows and fortran here: https://twitter.com/ContinuumIO/status/481908466047803392 |
@twiecki Branch 2.3 contains my fixes and updates for 2.3.3. It passes locally and on Travis, so have a peek and I will tag it if you think it looks okay. |
Looks good to me 👍. Let me know once you updated pypi. |
RuntimeError: MD5 mismatch: '99645bf558a41376be0e687ccf90a39f' != On Tue, Jul 1, 2014 at 7:40 PM, Chris Fonnesbeck notifications@github.com
Thomas Wiecki |
PyPI created the MD5 automatically, so not sure what's up. |
linux 64 updated https://binstar.org/pymc/pymc/2.3.3/files On Tue, Jul 1, 2014 at 7:49 PM, Chris Fonnesbeck notifications@github.com
Thomas Wiecki |
I have the same problem with the md5 checksum. The last time something was wrong with one of the various caches that conda uses, didn't fix it this time. Maybe the problem is, that the md5 changes as soon as we put the md5 in the build recipe. In that way they will never match up. I've pushed a good build to binstar: |
@tobeplugged Yeah, I also just manually changed the md5 string. Should probably just replace that. I don't think we need to update the pypi version for that. Since you're set up to do this and have pymc installed successfully, would you be willing to help me upload a hddm conda for osx? Build files etc are all in place and it works well for linux so should just be a conda build and binstar upload. Of course no worries if not. |
I've just build 3.3 and 3.4 packages for OSX as well: https://binstar.org/pymc/pymc/2.3.3/download/osx-64/pymc-2.3.3-np18py33_0.tar.bz2 Sure building hddm should be no problem. But I cannot find the recipe in the repository, or am I missing something? |
I just uploaded windows 32 bit for py2. Building hddm should be the same, just checkout master, and then from the root |
I wasn't looking in the right repo... voila:
build like a charm. I did not check, but if the package does not have any compiled code, than it should be possible to build cross platform as well. I'll remove the file as soon as its up on binstar. |
Actually, you could just upload it to the pymc repo as well. It has cython On Wed, Jul 2, 2014 at 1:48 PM, Tobias notifications@github.com wrote:
Thomas Wiecki |
Okay, I'll do that. I thought you might want to keep the channels apart and make a new channel for hddm |
I had that for a while but the installation is easier that way. I don't think it could cause any problems. |
Turns out that I cannot upload to pymc I always get an error that the file was not found. Maybe I do not have the permissions. See here: Edit: I've build and uploaded 3.3 and 3.4 versions of pymc for 64 bit Linux as well. |
OK, I uploaded the dropbox file. Unfortunately it also needs kabuki which On Wed, Jul 2, 2014 at 2:04 PM, Tobias notifications@github.com wrote:
Thomas Wiecki |
No problem:
I did not build against Python 3 because kabuki did not seem to be compatible. I have a script that builds a package against all possible combinations of Python (and numpy) so doing the builds for the Pythons is not a big deal. |
Many thanks, I really appreciate it. A user just tried to install but gets a pymc-related error (on osx):
Any ideas? Is a separate fortran library required even with anaconda? |
Okay, that is weird. I can load and run pymc with no problem on my machine after installing from binstar. |
Do you have that directory? On Wed, Jul 2, 2014 at 3:03 PM, Tobias notifications@github.com wrote:
Thomas Wiecki |
Yes I do have that directory. But the even with a fresh environment, I have no problems running PyMC. And the |
Right, but it seems to be linked against the other lib dynamically which On Wed, Jul 2, 2014 at 3:10 PM, Tobias notifications@github.com wrote:
Thomas Wiecki |
That is from homebrew. The library it is linked against is gfortran which I use to build it in the first place. |
I see. We might want to statically link in that case or provide the library with pymc together. |
@asmeurer Where you able to give this a try with recipe we have on this repo? |
Seems like gfortran is installed by the conda
|
I plan to split out the libs from my gcc package soon, at least for Linux. |
@asmeurer Did you have a chance to give this another try (compiling the conda recipe that's in this repo) or is there some progress on the fortran libraries in conda? Would be great to make this available! |
I think the thing that needs to be done is to take my |
I can confirm that I got it to install on Anaconda Yosemite with: conda install -c asmeurer pymc Thx @asmeurer!
|
Yes, that should work. The issue is that it will be a very large download because it currently depends on the entire gcc package. |
Thought I should chip in with my 2p: I'm running OSX Yosemite (10.10) and have an Anaconda Python 3.4.2 env that I'm intending to use for some work in pymc 2.3. My first attempt was to install and build In a ipython session I then
So before diving into some potential fun with gcc etc, I thought I'd uninstall pymc and try asmeurer's build I'll report back with any subtle issues, but so far, thanks for all your help! |
I think a package with a larger download is better than no package. |
Humm... I've read the whole thread (quite a learning!) and in the end I've tried @asmeurer package. It installed gcc and pymc but after the first import I get the now all to familiar dlopen error: I'm running on Yosemite with Python 3.4.2 (Anaconda 3.7.3).
As you can see it is trying to load the inexistent local |
Can you show me the output of |
Below is the output. By the way did you consider linking it against the Accelerate framework in OSX? It has a tuned implementation of the BLAS/Linpack (no need for GCC) but for libquadmath I'm clueless except for this: http://stackoverflow.com/questions/17590525/correct-way-to-statically-link-in-gfortran-libraries-on-osx. Thanks for the support! Output:
Notice that my |
And this is the latest build of my package (what is As for using Accelerate, I don't know how to do it, but I agree that it's better. We build the Continuum numpy/scipy against the Accelerate framework on OS X. |
I assumed it is the latest as I installed with as shown in the previous post by jonsedar. Here are the versions:
|
That looks right. I'm not sure what's going on. Does |
No, I get the same error but this time I also got this:
By the way, does the output of my otool look right? Cheers, |
I managed to link pymc statically against libgfortran on MacOSX. Basically by adding the following to the conda recipe:
PyMC is now officially supported by Anaconda on Linux and Mac OS X! Conda packages are available for Python 2. |
Great! I just installed |
@ilanschnell Thanks so much for looking into that! What about Python 3 and Windows? Those should be pretty easy to add I think. |
Python 3 packages are also already available for Linux and MacOSX. We haven't looked into Windows yet, but it's on our list. |
Cool! FWIW I think the recipe we have works pretty well: Not sure if it successfully links in gfortran though. |
@ilanschnell Any update on a windows version? I think then we can close this issue and just advertise conda as the preferred way to install pymc! |
I've just created Windows conda packages for pymc, such that we now support pymc on all platforms and for all Python versions! |
@ilanschnell @asmeurer Thank you guys so much! 👍 |
Can the recipe to build them all be found anywhere? |
@tobeplugged You were extremely helpful in building OSX conda packages for HDDM before. Do you think I could ask you to help with this once more? Specifically, kabuki and hddm would need to be pulled and then |
Hi, i'm getting this error:
I think I've tried all, but may be I did something wrong, can you help me? thanks |
The conda packages for pymc 2.3 are not compatible with current anaconda versions. We thus need to rebuild and reupload. I can do linux 64bit, @fonnesbeck can you do osx 64bit?
The text was updated successfully, but these errors were encountered: