def __init__(self,
batch_size=32,
pooling_strategy="REDUCE_MEAN",
max_seq_len=40):
"""
init BertVector
:param batch_size: Depending on your memory default is 32
"""
self.max_seq_length = max_seq_len
self.layer_indexes = args.layer_indexes
self.gpu_memory_fraction = 1
if pooling_strategy == "NONE":
pooling_strategy = args.PoolingStrategy.NONE
elif pooling_strategy == "REDUCE_MAX":
pooling_strategy = args.PoolingStrategy.REDUCE_MAX
elif pooling_strategy == "REDUCE_MEAN":
pooling_strategy = args.PoolingStrategy.REDUCE_MEAN
elif pooling_strategy == "REDUCE_MEAN_MAX":
pooling_strategy = args.PoolingStrategy.REDUCE_MEAN_MAX
self.graph_path = optimize_graph(pooling_strategy=pooling_strategy,
max_seq_len=self.max_seq_length)
self.tokenizer = tokenization.FullTokenizer(vocab_file=args.vocab_file,
do_lower_case=True)
self.batch_size = batch_size
self.estimator = self.get_estimator()
self.input_queue = Queue(maxsize=1)
self.output_queue = Queue(maxsize=1)
self.predict_thread = Thread(target=self.predict_from_queue,
daemon=True)
self.predict_thread.start()
def __init__(self, batch_size=32):
"""
init BertVector
:param batch_size: Depending on your memory default is 32
"""
self.max_seq_length = args.max_seq_len
self.layer_indexes = args.layer_indexes
self.gpu_memory_fraction = 1
self.graph_path = optimize_graph()
self.tokenizer = tokenization.FullTokenizer(vocab_file=args.vocab_file,
do_lower_case=True)
self.batch_size = batch_size
self.estimator = self.get_estimator()
self.input_queue = Queue(maxsize=1)
self.output_queue = Queue(maxsize=1)
self.predict_thread = Thread(target=self.predict_from_queue,
daemon=True)
self.predict_thread.start()
self.sentence_len = 0