This is a experimental project aiming to see the training result using different noise selection in NCE.
Create a data
folder first. Then run
python gen_data.py
This will preprocess the text8 dataset in suitable format
python word2vec_nce.py [path_to_result] [learning_rate]
or you see see the trained result in simple_LM
by
tensorboard --logdir=simple_LM
Trying to use pretrained models as noise