dl-sentiword.py word2vec variant training is extremely slow !

attardi / deepnl

Deep Learning for Natural Language Processing

GNU General Public License v3.0

457 stars 116 forks source link

dl-sentiword.py word2vec variant training is extremely slow ! #55

Open waniss opened 6 years ago

waniss commented 6 years ago

Hi, dl-sentiwords.py training.tsv --vocab SSWE_words.txt --vectorsSSWE_vectors.txt --variant word2vec --w 3 -e 50 --eps 10 --min-occur 10 --threads 30 My training set is a file of 1.5 million tweet, and the execution for 17h hours just moved by 1 or 2 epochs ? Any ideas to make things little faster please ?

Regards.

fuzihaofzh commented 6 years ago

There is the same problem with my data. Anyone have idea?

rudra0713 commented 5 years ago

I am also facing the same issue. But more importantly, with each epoch error value is increasing and accuracy always remains 0.00 . Can anyone explain why the error value is not decreasing?

samoliver360 commented 5 years ago

WhatsApp Image 2019-08-27 at 11 07 07 Total 64G RAM is used after 3 days and the code just reach 13 epochs!!!! See the graphic.....

samoliver360 commented 5 years ago

Anyone has the solution???????