DONT USE THIS, use gensim instead, but emb_reduce may caching and gensim may not
fire, tqdm
retrive word emb from large word emb
python ./reduce_emb/reduce_emb.py --fword wordlist --femb word_emb_file --fout reduced_emb_file
relatively small txt
each word perline
very large txt
optional wordnum:num emb_dim:num
for the first line, see ./tests/Tencent_AILab_ChineseEmbedding_sample for example. glove doesnt have this line
word:num dim0 dim1 [<dim2, >]
for rest
read until empty line
txt where to output, format will like golve without the first line
python-snippet/upload_to_pip.md at master · xsthunder/python-snippet