def test_step_report(logger,session,PadZeroBegin,max_length,test_path,
dropout_keep_prob,step,out_dir,char_alphabet,label_alphabet,word_alphabet,
word_column, label_column,char_embedd_dim,max_char_per_word):
# read test data
graph = tf.get_default_graph()
logger.info("Reading data from test set...")
word_sentences_test, _, word_index_sentences_test, label_index_sentences_test = dp.read_conll_sequence_labeling(
test_path, word_alphabet, label_alphabet, word_column, label_column)
logger.info("Padding test text and lables ...")
word_index_sentences_test_pad,test_seq_length = utils.padSequence(word_index_sentences_test,max_length, beginZero=PadZeroBegin)
label_index_sentences_test_pad,_= utils.padSequence(label_index_sentences_test,max_length, beginZero=PadZeroBegin)
logger.info("Creating character set FROM test set ...")
char_index_test,_= dp.generate_character_data(word_sentences_test,
char_alphabet=char_alphabet, setType="Test")
logger.info("Padding Test set ...")
char_index_test_pad = dp.construct_padded_char(char_index_test, char_alphabet, max_sent_length=max_length,max_char_per_word=max_char_per_word)
print(type(char_index_test_pad))
print(type(word_index_sentences_test_pad))
feed_dict=create_feed_Dict_Eval(graph,PadZeroBegin=PadZeroBegin,max_length=max_length,
x_batch=word_index_sentences_test_pad,
act_seq_lengths= test_seq_length, dropout_keep_prob=dropout_keep_prob,
char_batch=char_index_test_pad)
#tf.Print(feed_dict,feed_dict)
logit_op = graph.get_tensor_by_name('output/logits:0')
transition_params_op = graph.get_tensor_by_name('transitions:0')
logits,transition_params = session.run([logit_op, transition_params_op],feed_dict)
viterbi_decode(logits=logits,transition_params=transition_params,
seq_length=test_seq_length,x_batch=word_index_sentences_test_pad,word_alphabet=word_alphabet,label_alphabet=label_alphabet,
prefix_filename="test",beginZero=PadZeroBegin)
return
def test_step(logger,session,BiLSTM,PadZeroBegin,max_length,test_path,
dropout_keep_prob,step,out_dir,char_alphabet,label_alphabet,word_alphabet,
word_column, label_column,char_embedd_dim,max_char_per_word):
# read test data
logger.info("Reading data from test set...")
word_sentences_test, _, word_index_sentences_test, label_index_sentences_test = dp.read_conll_sequence_labeling(
test_path, word_alphabet, label_alphabet, word_column, label_column)
logger.info("Padding test text and lables ...")
word_index_sentences_test_pad,test_seq_length = utils.padSequence(word_index_sentences_test,max_length, beginZero=PadZeroBegin)
label_index_sentences_test_pad,_= utils.padSequence(label_index_sentences_test,max_length, beginZero=PadZeroBegin)
logger.info("Creating character set FROM test set ...")
char_index_test,_= dp.generate_character_data(word_sentences_test,
char_alphabet=char_alphabet, setType="Test")
logger.info("Padding Test set ...")
char_index_test_pad = dp.construct_padded_char(char_index_test, char_alphabet, max_sent_length=max_length,max_char_per_word=max_char_per_word)
# test summaries
#test_summary_op = tf.summary.merge([loss_summary])
#test_summary_dir = os.path.join(out_dir, "summaries", "test")
#test_summary_writer = tf.summary.FileWriter(test_summary_dir, sess.graph)
feed_dict=create_feed_Dict_Test(BiLSTM,PadZeroBegin=PadZeroBegin,max_length=max_length,
x_batch=word_index_sentences_test_pad, y_batch=label_index_sentences_test_pad,
act_seq_lengths= test_seq_length, dropout_keep_prob=dropout_keep_prob,
char_batch=char_index_test_pad)
'''#tf.Print(feed_dict,feed_dict)
logits, transition_params = session.run([BiLSTM.logits, BiLSTM.transition_params],feed_dict)
#logits is a list of numpy.ndarray
#transition_params : ndarray'''
logits, transition_params,embedded_char,embedded_words,char_pool_flat,input_x_test = session.run([BiLSTM.logits, BiLSTM.transition_params,
BiLSTM.W_char,BiLSTM.W_word,BiLSTM.char_pool_flat,BiLSTM.input_x],feed_dict)
accuracy,accuracy_low_classes = predictAccuracyAndWrite(logits,transition_params,test_seq_length,
label_index_sentences_test_pad,step,word_index_sentences_test_pad,word_alphabet,label_alphabet,prefix_filename="test",beginZero=PadZeroBegin)
#test_summary_writer.add_summary(summaries, step)
print("step {}, accuracy on test set {:g}, accuracy for classes except Others: {:g}".format(step,accuracy,accuracy_low_classes))
checkpoint_dir_test = os.path.abspath(os.path.join(out_dir, "checkpoints_test"))
if not os.path.exists(checkpoint_dir_test):
os.makedirs(checkpoint_dir_test)
fname_data = "input_x_test_"+str(step)+".pkl"
fname_conv_out = "char_pool_flat_"+str(step)+".pkl"
fname_seqLength = "act_seq_len_"+str(step)+".pkl"
fname_embedded_char = "embedded_char_"+str(step)+".pkl"
fname_embedded_words = "embedded_words_"+str(step)+".pkl"
pickle.dump(input_x_test,open(os.path.join(checkpoint_dir_test, fname_data),'wb'))
pickle.dump(char_pool_flat,open(os.path.join(checkpoint_dir_test, fname_conv_out),'wb'))
pickle.dump(test_seq_length,open(os.path.join(checkpoint_dir_test, fname_seqLength),'wb'))
pickle.dump(embedded_char,open(os.path.join(checkpoint_dir_test, fname_embedded_char),'wb'))
pickle.dump(embedded_words,open(os.path.join(checkpoint_dir_test, fname_embedded_words),'wb'))
print("Saved test data checkpoint to {}\n".format(checkpoint_dir_test))
return accuracy,accuracy_low_classes
char_index_dev, max_char_per_word_dev = dp.generate_character_data(
word_sentences_dev, char_alphabet=char_alphabet, setType="Dev")
logger.info("character alphabet size: %d" % (char_alphabet.size() - 1))
max_char_per_word = min(dp.MAX_CHAR_PER_WORD, max_char_per_word_train,
max_char_per_word_dev)
logger.info("Maximum character length is %d" % max_char_per_word)
logger.info("Constructing embedding table ...")
#TODO : modify network to use this
char_embedd_table = dp.build_char_embedd_table(
char_alphabet, char_embedd_dim=FLAGS.char_embedd_dim)
logger.info("Padding Training set ...")
char_index_train_pad = dp.construct_padded_char(
char_index_train,
char_alphabet,
max_sent_length=max_length,
max_char_per_word=max_char_per_word)
logger.info("Padding Dev set ...")
char_index_dev_pad = dp.construct_padded_char(
char_index_dev,
char_alphabet,
max_sent_length=max_length,
max_char_per_word=max_char_per_word)
#logger.info("Generating data with fine tuning...")
embedd_dict, embedd_dim, caseless = utils.load_word_embedding_dict(
embedding, embedding_path, logger)
logger.info("Dimension of embedding is %d, Caseless: %d" %
(embedd_dim, caseless))
#Create an embedding table where if the word from training/train/dev set is in glove , then assign glove values else assign random values