glove-py 0.1.8

glove Python bindings

glove-py is an implementation of the GloVe algorithm for learning word vectors from a corpus.

##Installation

using pip: pip install glove-py.

install with github code :

git clone [e-mail]:f20500909/glove_py.git

python3 setup.py install

from glove import *

model = Glove(epoch=2)
model.train("small_text")
model.to_txt()
words = model.most_similary("one", 10)
print(words)

using in shell

Skipgram model

Train & load skipgram model

from glove import *

model = Glove(params)
model.train(input_file)
model.to_txt()
words = model.most_similary("one", 10)
print(words)

List of available params and their default value:

help info......
  -input_file               input_file [small_text]
  -log_dir                  a dirctory to save temp file [log/]
  -cofile                   max length of word ngram [cooccur.bin]
  -vocab_file               output vocabulary [vocab.txt]
  -temp_file                temporary file of common occurrence chunk [temp.bin_]
  -embd_file                embedded file[wordvec.txt]
  -vocab_size               vocabulary size to train  [0]
  -max_size                 maximum size to keep in courpus [10000000]
  -min_count                minimum times  [1]
  -window                   the window size to get common occurrence [10]
  -embed_size               embedded size  [80]
  -epoch                    train epoch [1]
  -threads                  max length of char ngram [40]
  -memory_limit             max length of char ngram [1]
  -lr                       max length of char ngram [0.05]
  -keep_case                max length of char ngram [0]
  -symmetric                max length of char ngram [1]

Reference

• GloVe
• Jeffrey Pennington, Richard Socher, and Christopher D. Manning. 2014. GloVe: Global Vectors for Word Representation.
• GloVe-cpp