train_word2vec（CNN情感分析yoom例子五）

关键函数：

sentences = [[vocabulary_inv[w] for w in s] for s in sentence_matrix]
embedding_model = word2vec.Word2Vec(sentences, workers=num_workers, size=num_features, min_count = min_word_count, window = context, sample = downsampling)

embedding_weights = [np.array([embedding_model[w] if w in embedding_modelelse np.random.uniform(-0.25,0.25,embedding_model.vector_size)for w in vocabulary_inv])]

from gensim.models import word2vec
from os.path import join, exists, split
import os
import numpy as np
import data_helpers

def train_word2vec(sentence_matrix, vocabulary_inv,
                   num_features=300, min_word_count=1, context=10):
    """
    Trains, saves, loads Word2Vec model
    Returns initial weights for embedding layer.
   
    inputs:
    sentence_matrix # int matrix: num_sentences x max_sentence_len
    vocabulary_inv  # dict {str:int}
    num_features    # Word vector dimensionality                      
    min_word_count  # Minimum word count                        
    context         # Context window size 
    """
    model_dir = 'word2vec_models'
    model_name = "{:d}features_{:d}minwords_{:d}context".format(num_features, min_word_count, context)
    model_name = join(model_dir, model_name)
    print(model_name)
    if exists(model_name):
        #important
        embedding_model = word2vec.Word2Vec.load(model_name)
        print(split(model_name))
        print('Loading existing Word2Vec model \'%s\'' % split(model_name)[-1])
    else:
        # Set values for various parameters
        num_workers = 2       # Number of threads to run in parallel
        downsampling = 1e-3   # Downsample setting for frequent words
        
        # Initialize and train the model
        print("Training Word2Vec model...")
        sentences = [[vocabulary_inv[w] for w in s] for s in sentence_matrix]
        embedding_model = word2vec.Word2Vec(sentences, workers=num_workers, \
                            size=num_features, min_count = min_word_count, \
                            window = context, sample = downsampling)
        
        # If we don't plan to train the model any further, calling 
        # init_sims will make the model much more memory-efficient.
        embedding_model.init_sims(replace=True)
        
        # Saving the model for later use. You can load it later using Word2Vec.load()
        if not exists(model_dir):
            os.mkdir(model_dir)
        print('Saving Word2Vec model \'%s\'' % split(model_name)[-1])
        embedding_model.save(model_name)
    print("<<<<<<<model>>>>>>>")
    print(embedding_model)
    print("<<<<<weight of 'the'>>>>>>>")
    print(embedding_model['the'])
    #  add unknown words
    embedding_weights = [np.array([embedding_model[w] if w in embedding_model\
                                                        else np.random.uniform(-0.25,0.25,embedding_model.vector_size)\
                                                        for w in vocabulary_inv])]
    print("<<<<<<<weight>>>>>>>")
    print(embedding_weights )
    #form:[[[300 dimension],[300 dimension]], dtype=float32]
    print("<<<<<<<num of wieghts>>>>>>>")
    print(len(embedding_weights[0]))
    print('<<<<<<<translate word0>>>>>>>>>')
    print(embedding_weights[0][0])
    return embedding_weights

if __name__=='__main__':
    print("Loading data...")
    x, _, _, vocabulary_inv = data_helpers.load_data()
    w = train_word2vec(x, vocabulary_inv,num_features=50)

结果：

Loading data...
word2vec_models\50features_1minwords_10context
('word2vec_models', '50features_1minwords_10context')
Loading existing Word2Vec model '50features_1minwords_10context'
<<<<<<<model>>>>>>>
Word2Vec(vocab=18765, size=50, alpha=0.025)
<<<<<weight of 'the'>>>>>>>
[-0.02123935 -0.05490068 -0.23091695  0.22885838  0.27167228 -0.05158678
 -0.05539207 -0.22459298 -0.13045608  0.1070291  -0.02403063 -0.00130523
  0.19467545  0.0296479  -0.08463751  0.06144359 -0.12156744 -0.10263078
 -0.06417726 -0.01894584 -0.24981096  0.06570466  0.18615869  0.0254211
  0.16925883  0.16957371  0.09030077 -0.21692845  0.06049829 -0.16586471
 -0.14992879  0.10567563  0.06595352 -0.2763775   0.19153105  0.12031604
 -0.04390313  0.17974497  0.0111691   0.02915976 -0.2177939  -0.10611911
  0.03313744 -0.1350788   0.14111236 -0.2494767   0.1069486   0.12323371
  0.12049541  0.06635255]
<<<<<<<weight>>>>>>>
[array([[ 0.22287713,  0.16943391, -0.18881463, ...,  0.14732356,
         0.13720259,  0.01359756],
       [-0.02123935, -0.05490068, -0.23091695, ...,  0.12323371,
         0.12049541,  0.06635255],
       [ 0.00441776,  0.18812552, -0.04573391, ...,  0.21478625,
         0.13668334,  0.07716116],
       ..., 
       [-0.12307893,  0.00517228, -0.15578713, ...,  0.14851454,
         0.03096815,  0.02698081],
       [ 0.10233459,  0.05561316, -0.14217722, ...,  0.13685261,
         0.06430191,  0.07483072],
       [ 0.38770503,  0.25239426,  0.05189984, ...,  0.05148352,
         0.05523289, -0.00138285]], dtype=float32)]
<<<<<<<num of wieghts>>>>>>>
18765
<<<<<<<translate word0>>>>>>>>>
[ 0.22287713  0.16943391 -0.18881463  0.17884882  0.18564233 -0.0442104
 -0.0878861  -0.1689464  -0.18261056  0.06646598 -0.01281965 -0.13765122
 -0.13511361 -0.09671792  0.00905229  0.15683898 -0.05347729 -0.19133224
 -0.19077209  0.06136192 -0.23724565  0.02753574  0.12075555  0.04671283
  0.24587892  0.04068457  0.13252144 -0.17025913  0.02412261 -0.09669705
 -0.1384121   0.23630536  0.02841855 -0.09812463  0.09301346  0.14504847
  0.00938961  0.20613976  0.11047759 -0.09049261 -0.303936    0.05824288
  0.03945156 -0.07627468  0.07813629 -0.24140984  0.11729364  0.14732356
  0.13720259  0.01359756]
[Finished in 2.4s]