node-word2vec
node-word2vec copied to clipboard
Had to remove arr.pop() on line 373 in model.js to get this library working
This line here was removing the last item from each vector which made their lengths different which caused a whole bunch of chaos down the line (multiplying by undefined).
I have no idea why it's there, but I did notice that it works for the example vector.txt
in this project. Maybe something to do with \r and \n?
Also I added this:
if(isNaN(words) || isNaN(size)) {
throw new Error("First line of input text file should be <number of words> <length of vector>. See example data 'vectors.txt' in repo");
}
After this line since that caused me a lot of trouble (I don't think it's mentioned anywhere in the readme).
Thanks for the awesome library!
Thanks for this report, I will look into it. If I recall correctly, I indeed put the .pop()
call there to remove newline and carriage returns from the array.
Regarding your isNaN
checks, any call to word2vev
should write a file where the first line consists of <number of words> <length of vector>
. I guess you have been using the library differently?
Oh, right, yes I was using a pretrained file that I found on the internet. I just added the <number of words> <length of vector>
manually.
Also, sorry to pester, but did you see this issue on your node-wordnet-magic project?
Yeah, sorry for not coming back to that earlier. I will have a look today into what's wrong with the WordNet module.
Thanks :+1: