search

meta-toolkit / meta

A Modern C++ Data Sciences Toolkit
https://meta-toolkit.org
MIT License
696 stars 236 forks source link

PerceptronTagger hangs #188

Closed magic-lantern closed 7 years ago

magic-lantern commented 7 years ago

From today's demo at KDD 2017 - wether running

tagger = metapy.sequence.PerceptronTagger("perceptron-tagger/")

either inside Jupyter or via command-line python, the above line of code appears to hang - no error is output, but never completed. Others said that killing the kernel and restarting worked - but didn't work for me.

As an FYI, here's what Python I'm running on macOS Sierra 10.12.6:

python --version Python 3.6.1 :: Anaconda custom (x86_64)

python --version Python 2.7.13 :: Continuum Analytics, Inc.

skystrife commented 7 years ago

Thanks for the report. I was just able to reproduce this; it looks like it is a deadlock contention for the python GIL to print loading progress output, but it only happens if metapy.log_to_stderr() was called first.

This has been fixed in meta-toolkit/metapy@23bd3c0 and should be available once the build jobs for v0.2.6 finish and I can upload the new library version to PyPI.

magic-lantern commented 7 years ago

Thanks for the quick look and resolution.

On Wed, Aug 16, 2017 at 2:57 PM, Chase Geigle notifications@github.com wrote:

Thanks for the report. I was just able to reproduce this; it looks like it is a deadlock contention for the python GIL to print loading progress output if metapy.log_to_stderr() was called first.

This has been fixed in meta-toolkit/metapy@23bd3c0 https://github.com/meta-toolkit/metapy/commit/23bd3c0 and should be available once the build jobs for v0.2.6 finish and I can upload the new library version to PyPI.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/meta-toolkit/meta/issues/188#issuecomment-322850684, or mute the thread https://github.com/notifications/unsubscribe-auth/AI8R-FTubhUIlfbv37sAIXOnjiIXmOm6ks5sYy2kgaJpZM4O5DVw .

skystrife commented 7 years ago

The new version has just finished deployment, so you should be able to grab the latest version with the fix with:

pip install --upgrade metapy
magic-lantern commented 7 years ago

Just wanted to let you know that I've confirmed this is resolved as well.