You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi. First of all thank you very much for your help. You have saved my life at least several times :) My question is that I have experinced some problems while training word2vec with large data corpus. The data i'd like to use for training process is almost 4 Gb. I wonder whether if it's possible or not. I tried to train word2vec with 2 Gb data and it didn't work too.Shall i increase the heap-size or something like that ?
The text was updated successfully, but these errors were encountered:
I find out where is this issue and #1
I use some 100mb internet data and it was surprise that there is exception, but ther i understand that when I do StreamReader.ReadLine() I read a whole file which is storing with only spaces and thats cause an exception. And actually I dont even sure what to do to save same performance, because there is threads and seek, but you cant just seek through single line so
I solve this by preprocessing train file and separating some amount of words in single line because solid line cause issues even when opening with notepad++ when opening processed files occurs instantly
Hi. First of all thank you very much for your help. You have saved my life at least several times :) My question is that I have experinced some problems while training word2vec with large data corpus. The data i'd like to use for training process is almost 4 Gb. I wonder whether if it's possible or not. I tried to train word2vec with 2 Gb data and it didn't work too.Shall i increase the heap-size or something like that ?
The text was updated successfully, but these errors were encountered: