undefined symbol: cblas_dasum

[boscotsang]

After deployment the import error raised during runtime. The error report as follow:

ImportError: /app/.heroku/python/lib/python2.7/site-packages/sklearn/metrics/pairwise_fast.so: undefined symbol: cblas_dasum

[branliu0]

Can you provide more information about what versions of packages you're using?

There are known issues with installing sklearn on this buildpack with certain versions. I personally haven't had the time to look into sklearn in detail yet, since I don't use it very much.

[boscotsang]

My requirements.txt is as follow

numpy==1.9.2 scipy==0.15.1 scikit-learn==0.15.2 simplejson>=3.8.0 Flask>=0.10 Jinja2>=2.8 Werkzeug>=0.9 gunicorn>=0.17.2

[mitchelljkotler]

I am seeing this issue as well

[branliu0]

Is it also for sklearn specifically? I'm afraid I haven't been able to investigate sklearn issues with this buildpack.

[mitchelljkotler]

Yes, for sklearn. Not having any luck getting sklearn going on heroku by any means. On Oct 24, 2015 5:30 PM, "Brandon Liu" [email protected] wrote:

Is it also for sklearn specifically? I'm afraid I haven't been able to investigate sklearn issues with this buildpack.

— Reply to this email directly or view it on GitHub https://github.com/thenovices/heroku-buildpack-scipy/issues/22#issuecomment-150852672 .

[mitchelljkotler]

Using this buildpack instead resolved the issue for me: https://github.com/andrewychoi/heroku-buildpack-scipy

[WhiskySoda]

Still having this issue, tried the build pack... unfortunately, not much luck. sklearn==0.15.2

[72L]

I'm also having this issue. Switching to https://github.com/andrewychoi/heroku-buildpack-scipy did not work for me scikit-learn==0.16.1, scipy==0.15.1, numpy==1.9.2

[anitameh]

I'm also having the same issue, even after switching to andrewychoi's buildpack.

ImportError: /app/.heroku/python/lib/python2.7/site-packages/sklearn/metrics/pairwise_fast.so: undefined symbol: cblas_dasum

anybody manage to get this working?

[jdelman]

Also having this issue.

@mitchelljkotler and @thenovices - could you tell us which versions of sklearn, numpy, and scipy that you're using?

[branliu0]

Sorry guys -- I don't use sklearn at all, but I can try to look into this. I'm pretty sure we would have to make a new precompiled binary for sklearn as well. I don't have that much time right now to work on that -- but hopefully will get to it at some point

[jdelman]

Running using Choi's buildpack with the following versions:

numpy==1.9.1
scipy==0.14.0
scikit-learn==0.15.2

[mitchelljkotler]

Also with Choi's buildpack: numpy==1.9.1 scipy==0.14.0 scikit-learn==0.16.1

On Wed, Nov 11, 2015 at 1:36 PM, Josh Delman [email protected] wrote:

Running using Choi's buildpack https://github.com/andrewychoi/heroku-buildpack-scipy with the following versions:

numpy==1.9.1 scipy==0.14.0 scikit-learn==0.15.2

— Reply to this email directly or view it on GitHub https://github.com/thenovices/heroku-buildpack-scipy/issues/22#issuecomment-155870576 .

[anitameh]

sweet got this working with choi's buildpack too! thanks!!

[seyonv]

I'm having the same problem. (Using Choi's build pack and versions specified by mitchelljkotler

Getting the following error "ImportError: liblapack.so.3: cannot open shared object file: No such file or directory"

[troyshu]

@seyonv @mitchelljkotler's version numbers just worked for me. Make sure numpy, scipy, and scikit-learn are the very first things in your requirements.txt; also, I'm on Python 2.7.

[alexcombessie]

Same issue here. I solved it with Choi's build pack and the following versions on my requirements.txt:

numpy==1.9.1
scipy==0.14.0
scikit-learn==0.16.1

Be careful to reset all your buildpacks beforehand to do a clean install. You can do heroku buildpacks:clear

Just to be sure, I also pip uninstalled everything with pip freeze | xargs pip uninstall -y.

[FreshDonut]

Following all the tips. I feel like I'm getting closer. I do require pandas to read from a csv file though. I seem to be getting the following error on app load ImportError: C extension: liblapack.so.3: cannot open shared object file: No such file or directory not built. If you want to import pandas from the source directory, you may need to run 'python setup.py build_ext --inplace' to build the C extensions first. Is this bc it can't find the file from S3 or something to do with pandas? Any thoughts guys?