In order to explore FastText, I prepared a train.txt containing 90 examples tagged with a single label, and a test.txt containing 4 examples. The training is performed successfully, but i got no results after testing the model.
Is it due to the number of examples in the files train and test?
After installing those files provided in the link https://github.com/xiamx/fastText/releases
Please explain what to do next?
I extracted those files but not able to open.. Can anyone tell me the procedure what i need to do after extracting those files
How will i get the fasttext?
Please reply to my as soon as poss
Facebook published some pretrained vectors that can be used. I have tried running it on this windows version but the command takes a very long time, it's been at least 15 minutes, will it even work?
I also noticed that the size of the binaries produced is 0.16 times of the self-built binaries. How is that so?
My text file exist 1537584 lines, but after the prediction process, there are only 1537490 results.
I've already check that my file didn't include any empty line.
After I published this issues to facebookresearch/fasttext, they said it might be a old bug.
(facebookresearch#344)
Do you have latest build at windows binary file? It will really help me !! Because my environment is windows server 2016, and your binary file is my only available source to run it. I suffered several issue at R/Python fasttext wrapper, only your one can be executed successfully.
Hi, I am using your fasttext bin version. Everything works fine, esp. applying nn on a model I produced by myself. Trying to use predict or predict-prob on this or any other models produces the following error code:
Assertion failed: A.m_ == m_, file src\vector.cc, line 93
I'm trying to train FastText for performing Information Extraction (which is considered as a text classification problem) on a big corpus where the positive examples (speakers) are not organized one per line, like the paragrapgh below.
Can FastText perform the classification based on this kind of input?
"The Student Alumni Relations Council (SARC) invites you to a Career Talk featuring __label__speaker Andrew Gault HS'80 HNZ'94 and three university speakers, __label__speaker Mary Francis McLaughlin (volunteer and community service opportunities), __label__speaker Jessie Ramey (research opportunities) and __label__speaker Judi Mancuso (part-time work-study summer and internship opportunities) on Tuesday January 31 at 7:00 P M in the Carnegie Conference Room 1st Floor Warner Hall."
f = open('train.txt', 'w')
f.write('__label__1 i love you\n')
f.write('__label__1 he loves me\n')
f.write('__label__1 she likes baseball\n')
f.write('__label__0 i hate you\n')
f.write('__label__0 sorry for that\n')
f.write('__label__0 this is awful')
f.close()
f = open('test.txt', 'w')
f.write('sorry hate you')
f.close()
`
Running fasttext supervised -input train.txt -output model -dim 2 yields a model.bin file with 0 bytes