Error installing implicit on Ubuntu about implicit HOT 9 CLOSED

What version of scipy are you using? I think these functions were added in v0.16

from implicit.

scipy==0.19.1.

The same requirements set worked on Mac OS X 10.12.14, if that helps, this is an error that I encountered when pushing out to the CI server.

from implicit.

Interesting. I'm using ubuntu w/ scipy v 0.19.0 (via the python3 anaconda distribution) without any issues myself.

Can you verify that the 'cython_lapack.pxd' file is included in the scipy thats getting picked up?

python -c "import scipy.linalg; import os; dirname = os.path.dirname(scipy.linalg.__file__); print(dirname); print(scipy.__version__); print('\n'.join(os.listdir(dirname)))"

from implicit.

Hmm you know, I think this might actually be an issue with the CI service provider. I've tried ssh-ing into their CI boxes and doing a manual install of all packages, and can't replicate the error. Have also tried deploying to my own Ubuntu box, and I can't replicate the error. It only seems to happen when the CI service runs the dependency installation as part of their installation chain, so I'm going to assume that something odd is going on on their end.

from implicit.

Ok, managed to replicate the error. For some reason, if I install the components manually, i.e.

pip install Cython==0.27.2
pip install scipy==0.19.1
pip install implicit==0.2.6
and then
pip install -r requirements.txt for the rest of my requirements

that succeeds on Ubuntu 14.04.

However, moving scipy and implicit into the requirements.txt file results in a failure. As far as I can tell, there shouldn't be a reason that the two are different, since the requirements.txt file output shows that the right package versions are installed or cached:

Collecting scipy==0.19.1 (from -r requirements.txt (line 69))
  Using cached scipy-0.19.1-cp35-cp35m-manylinux1_x86_64.whl
Collecting implicit==0.2.6 (from -r requirements.txt (line 70))
  Using cached implicit-0.2.6.tar.gz
    Complete output from command python setup.py egg_info:
    
    Error compiling Cython file:
    ------------------------------------------------------------
    ...

and yet I still get that odd error that suggests that it's using the wrong version of scipy. No idea why this might be happening since all of the versions are pegged, but installing outside of requirements.txt works fine for now.

from implicit.

interesting - I think it must be picking up the wrong version somehow =( glad you managed to work around this problem!

from implicit.

HI, I am using implicit to implement ALS . While fitting model, I am getting error "AttributeError: 'DataFrame' object has no attribute 'tocsr'"
Using scipy version 0.16
implicit version - 0.2.7

from implicit.

@erseema That looks like you're passing a pandas dataframe to the fit function, when you should be passing a sparse matrix (like a scipy.sparse.csr_matrix or similar) in.

Check out the examples for how to convert: https://github.com/benfred/implicit/blob/master/examples/lastfm.py#L71

from implicit.

Hi, can you provide an example of how you suggest to convert a pandas dataframe to sparse matrix?

I did not understand the link you provided.

I tried this one:
https://hippocampus-garden.com/pandas_sparse/#:~:text=To%20convert%20a%20DataFrame%20to,a%20coordinate%20(COO)%20matrix.

which worked, but I don't know if this is the best way..

from implicit.