Error when loading reddit_vectors-1.1.0

[GregSilverman]

Adding to existing Jupyter notebook that is run from a docker container, I do the following:

import sense2vec
s2v = sense2vec.load('reddit_vectors-1.1.0')

which throws the error:

/usr/local/lib/python3.5/dist-packages/sense2vec/util.py in get_package_by_name(name, via)
     18                                name or about.default_model, data_path=via)
     19     except PackageNotFoundException as e:
---> 20         raise RuntimeError("Model not installed. Please run 'python -m "
     21                            "sense2vec.download' to install latest compatible "
     22                            "model.")

Model not installed. Please run 'python -m sense2vec.download' to install latest compatible model.

I then create a local image of the container using a TensorFlow Dockerfile, and in the Dockerfile I issue RUN pip install sense2vec when building my Docker container, but if I do a RUN ipython -m sense2vec.download I get an error that print("Model already installed. Please run '%s --force to reinstall." % sys.argv[0], file=sys.stderr)
AttributeError: module 'sense2vec.about' has no attribute '__title__'.

The same happens if I SSH into the Docker container, that is, if I do an ipython -m sense2vec.download from the command line I get the error that print("Model already installed. Please run '%s --force to reinstall." % sys.argv[0], file=sys.stderr)

The iPython version is 3.5. Docker container's OS version is 64-bit: SMP Wed Mar 14 15:12:16 UTC 2018 x86_64 x86_64 x86_64 GNU/Linux

I just noticed the requirement of CPython. Since Jupyter uses iPython, I suppose there is no way to get this to work? even if so, why the discrepancy is error from outside the container and inside it?

Please advise.

[ines]

It looks like you might be running v0.x instead of v1.0.0a0 (i.e. the latest alpha version). Sorry, this isn't very clear in the README – I might add an explicit note. To install the alpha, you have to tell pip the version number explicitly:

pip install sense2vec==1.0.0a0

I just noticed the requirement of CPython. Since Jupyter uses iPython, I suppose there is no way to get this to work?

That should be no problem – under the hood, you're most likely running CPython, which then powers your iPython environment.

[GregSilverman]

Got it. Now, I am getting an error when installing the alpha version:

Step 5/8 : RUN pip install sense2vec==1.0.0a0
---> Running in a72e0965423d
Collecting sense2vec==1.0.0a0
https://files.pythonhosted.org/packages/28/4a/a1d9a28545adc839789c1442e7314cb0c70b
Downloading https://files.pythonhosted.org/packages/28/4a/a1d9a28545adc839789c1442e7314cb0c70b8657a885f9e5b287fade7814/sense2vec-1.0.0a0.tar.gz (311kB)
   Complete output from command python setup.py egg_info:
   Traceback (most recent call last):
     File "<string>", line 1, in <module>
     File "/tmp/pip-build-vln916it/sense2vec/setup.py", line 169, in <module>
       setup_package()
     File "/tmp/pip-build-vln916it/sense2vec/setup.py", line 107, in setup_package
       readme = f.read()
     File "/usr/lib/python3.5/encodings/ascii.py", line 26, in decode
       return codecs.ascii_decode(input, self.errors)[0]
   UnicodeDecodeError: 'ascii' codec can't decode byte 0xe2 in position 5413: ordinal not in range(128)

Appreciate the help!

[GregSilverman]

I've tried everything here, to no avail: pip-install-numpy-throws-an-error-ascii-codec-cant-decode-byte-0xe2

I was able to successfully install it on my MacBook just now, so I'm wondering if there is a potential memory or disk space issue happening. I increased memory on the Docker VM to 4 Gig and there are 60 Gigs free, but if that it not enough to install the models, that could cause an issue.

[GregSilverman]

Unable to get this working in my Docker container (using the Dockerfile here: https://github.com/adeshpande3/LSTM-Sentiment-Analysis). Will try using it locally with spaCy at some point.

[GregSilverman]

As soon as I posted this, I tried downgrading the version of Python in the container, and lo' and behold, it worked! Not sure why, since on my laptop I am running Anaconda version 3.6.5. Anyway, all is well!

As a side-note: I did try upgrading the version of python in the container to 3.6.3, but even then it still threw the error. In any case, since I need to run this with syntaxnet, I will just use 2.7, for the time being. I did need to add from __future__ import unicode_literals to the top of my code to properly handle unicode encodings.