def get_feed_dict(self, seqs, labels=None, lr=None, dropout=None):
"""
:param seqs:
:param labels:
:param lr:
:param dropout:
:return: feed_dict
"""
word_ids, seq_len_list = pad_sequences(seqs, pad_mark=0)
feed_dict = {
self.word_ids: word_ids,
self.sequence_lengths: seq_len_list
}
if labels is not None:
labels_, _ = pad_sequences(labels, pad_mark=0)
feed_dict[self.labels] = labels_
if lr is not None:
feed_dict[self.lr_pl] = lr
if dropout is not None:
feed_dict[self.dropout_pl] = dropout
return feed_dict, seq_len_list
def predict_one_batch(seqs, stub, dropout=1.0):
request = predict_pb2.PredictRequest()
# request.model_spec.name = 'mnist' # for test
request.model_spec.name = 'bi_lstm_crf'
request.model_spec.signature_name = \
tf.saved_model.signature_constants.DEFAULT_SERVING_SIGNATURE_DEF_KEY
word_ids, seq_len_list = pad_sequences(seqs, pad_mark=0)
request.inputs['word_ids'].CopyFrom(
tf.contrib.util.make_tensor_proto(word_ids, shape=[1,
seq_len_list[0]]))
request.inputs['sequence_lengths'].CopyFrom(
tf.contrib.util.make_tensor_proto(seq_len_list, shape=[1]))
request.inputs['dropout'].CopyFrom(
tf.contrib.util.make_tensor_proto(dropout, shape=[]))
# result_future = stub.Predict.future(request, 2.0)
# e = result_future.exception()
# if e:
# print e
# return
# result = result_future.result()
result = stub.Predict(request, 2.0)
logits_proto = result.outputs['logits']
# logits_shape = [logits_proto.tensor_shape.dim[i].size
# for i in range(len(logits_proto.tensor_shape.dim))]
# logits = numpy.array(logits_proto.float_val).reshape(logits_shape)
logits = tf.contrib.util.make_ndarray(logits_proto)
transition_params_proto = result.outputs['transition_params']
# transition_params_shape = [transition_params_proto.tensor_shape.dim[i].size
# for i in range(len(transition_params_proto.tensor_shape.dim))]
# transition_params = numpy.array(transition_params_proto.float_val).reshape(transition_params_shape)
transition_params = tf.contrib.util.make_ndarray(transition_params_proto)
label_list = []
for logit, seq_len in zip(logits, seq_len_list):
viterbi, viterbi_score = viterbi_decode(logit[:seq_len],
transition_params)
label_list.append(viterbi)
return label_list, seq_len_list
def get_feed_dict(self, seqs, labels):
seqs, seqs_len, max_len = pad_sequences(seqs, pad_mark='。')
labels, _, _ = pad_sequences(labels, pad_mark='O')
return seqs, seqs_len, labels, max_len
def get_feed_dict_seq(self, seqs):
seqs, seqs_len, max_len = pad_sequences(seqs, pad_mark='。')
return seqs, seqs_len, max_len
