error acessing weights of gmm about pykaldi

[fgvbrt]

Forgot to mention, that i tried to do it in Jupiter notebook and according to error something wrong with printing, I created PR to fix it.

[dogancan]

Hmm, this is a weird problem. Importing _str module directly in matrix/__init__.py shouldn’t be necessary since it is being imported indirectly when set_printoptions function is imported. This is a quirk of the python import system but it only works inside __init__.py files. From the error message I see that the __init__.py file was somehow renamed to init.py when installed and I think that is why the direct import is needed. What I don’t understand is how and why that __init__.py file was renamed to init.py. @vrmpx Do you know if this is something that happens on some platforms? If this is the case then we might have a bunch of these problems in the codebase.

[vrmpx]

Hi @fgvbrt,

I see that you are using Anaconda and Python 3.5.
In order to replicate this problem, would you mind telling me:

1. what OS are you using?
2. what anaconda version are you using?

Thanks

[fgvbrt]

Hi @vrmpx, please find details below:

1. Ubuntu 16.04.4
2. conda 4.3.29, anaconda Command line client (version 1.5.1), for installation I used conda environments with Python 3.5.5

[dogancan]

Hmm, scratch that renaming theory. It was github's formatting which turned those double underscores in the error message to bold text.

[dogancan]

@fgvbrt I have been looking into why this happened but I couldn't find anything online and we cannot replicate it on our end. The reason I haven't merged #31 yet is because I want to understand the reason why this is an issue on your end only.

Could you help us debug this by running these test scripts on your system and sending us the output? Please run them as scripts, i.e. do not type them into an ipython session, so we can make sure no other packages are interfering with their operation. Also make sure you run them against vanilla PyKaldi, i.e. without the change in #31.

Here is the output I am getting

>>> python test1.py

 1  2
 3  4
[kaldi.matrix.Matrix of size 2x2]

>>> python test2.py

 0
 0
 0
 0
[kaldi.matrix.SubVector of size 4]

[fgvbrt]

Hi @dogancan I am getting the same output. And with latest master I also can't reproduce the error in first message. Thanks for help.