def load(cls, filename, lm, config):
logger.info('Loading rule table from "%s"...' % filename)
timer = Timer()
timer.tic()
table = cls.__load_rules(filename, lm, config)
logger.info('Rule table loaded in %f seconds.' % timer.toc())
return table
def load(cls, filename, lm, config):
logger.info('Loading rule table from "%s"...' % filename)
timer = Timer()
timer.tic()
table = cls.__load_rules(filename, lm, config)
logger.info('Rule table loaded in %f seconds.' % timer.toc())
return table
print >> stderr, 'lambda_reg: %20.18f' % lambda_reg
print >> stderr, 'Max iterations: %d' % maxiter
if _seed:
print >> stderr, 'Random seed: %s' % _seed
print >> stderr, ''
print >> stderr, 'load Word2Vec Model...'
word_vectors = WordVectors.load_vectors(word_vector_file)
embsize = word_vectors.embsize()
print >> stderr, 'preparing data...'
instances, _, total_internal_node = prepare_data(word_vectors, instances_file)
print >> stderr, 'init. RAE parameters...'
timer = Timer()
timer.tic()
if _seed != None:
_seed = int(_seed)
else:
_seed = None
print >> stderr, 'seed: %s' % str(_seed)
theta0 = init_theta(embsize, _seed=_seed)
theta0_init_time = timer.toc()
print >> stderr, 'shape of theta0 %s' % theta0.shape
timer.tic()
if save_theta0:
print >> stderr, 'saving theta0...'
pos = model.rfind('.')
if pos < 0:
print >> stderr, 'lambda_reg: %20.18f' % lambda_reg
print >> stderr, 'Max iterations: %d' % maxiter
if _seed:
print >> stderr, 'Random seed: %s' % _seed
print >> stderr, ''
print >> stderr, 'load word vectors...'
word_vectors = WordVectors.load_vectors(word_vector_file)
embsize = word_vectors.embsize()
print >> stderr, 'preparing data...'
instances, _, total_internal_node = prepare_data(
word_vectors, instances_file)
print >> stderr, 'init. RAE parameters...'
timer = Timer()
timer.tic()
if _seed != None:
_seed = int(_seed)
else:
_seed = None
print >> stderr, 'seed: %s' % str(_seed)
theta0 = init_theta(embsize, _seed=_seed)
theta0_init_time = timer.toc()
print >> stderr, 'shape of theta0 %s' % theta0.shape
timer.tic()
if save_theta0:
print >> stderr, 'saving theta0...'
pos = model.rfind('.')
if pos < 0:
rule_table,
lm,
recombination_checker=recombination_checker,
extra_feature_funcs=extra_feature_funcs,
checking_hypo=checking,
expend_loser=expend_loser)
logger.info('Start decoding...')
def translate(data):
_, sentence = data
translations = decoder.translate(sentence, k, drop_oov, with_rule_tree)
return translations
total_timer = Timer()
total_timer.tic()
timer = Timer()
sentences = [
re.sub(r'\s+', ' ', sentence).strip().split(' ') for sentence in source
]
data = [(sid, sentences[sid]) for sid in range(len(sentences))]
if threads > 1:
pool = Pool(threads, init_worker)
all_translations = []
try:
all_translations = pool.map(translate, data)
pool.close()
pool.join()
except KeyboardInterrupt:
logger.critical('Caught KeyboardInterrupt, terminating workers')
extra_feature_funcs = build_extra_feature_funcs(config)
recombination_checker = CombinedRecombinationChecker(extra_feature_funcs)
decoder = CKYDecoder(config, rule_table, lm,
recombination_checker=recombination_checker,
extra_feature_funcs=extra_feature_funcs,
checking_hypo=checking, expend_loser=expend_loser)
logger.info('Start decoding...')
def translate(data):
_, sentence = data
translations = decoder.translate(sentence, k, drop_oov, with_rule_tree)
return translations
total_timer = Timer()
total_timer.tic()
timer = Timer()
sentences = [re.sub(r'\s+', ' ', sentence).strip().split(' ') for
sentence in source]
data = [(sid, sentences[sid]) for sid in range(len(sentences))]
if threads > 1:
pool = Pool(threads, init_worker)
all_translations = []
try:
all_translations = pool.map(translate, data)
pool.close()
pool.join()
except KeyboardInterrupt:
logger.critical('Caught KeyboardInterrupt, terminating workers')
pool.terminate()
print >> stderr, 'Random seed: %s' % _seed
print >> stderr, ''
print >> stderr, 'load word vectors...'
# 载入词向量的输入放入word_vectors中
word_vectors = WordVectors.load_vectors( word_vector_file )
Xidx = [word_vectors.get_word_index( '$X_1' ), word_vectors.get_word_index( '$X_2' )]
#embsize为词向量的维度
embsize = word_vectors.embsize()
print >> stderr, 'preparing data...'
#载入训练短语数据，将短语转化为instance的数组放入instances中
instances, _, total_internal_node = prepare_data( word_vectors, instances_file )
print >> stderr, 'init. RAE parameters...'
timer = Timer()
timer.tic()
if _seed != None:
_seed = int(_seed)
else:
_seed = None
print >> stderr, 'seed: %s' % str(_seed)
# 初始化参数
theta0 = init_theta( embsize, word_vectors, _seed = _seed )
theta0_init_time = timer.toc()
print >> stderr, 'shape of ' + tp + ' theta0 %s' % theta0.shape
timer.tic()
if save_theta0:
print >> stderr, 'saving ' + tp + ' theta0...'
pos = model.rfind('.')