def on_epoch_end(self, epoch, logs={}):
y_prob = self.model.predict(self.X_test)
np.save(self.saveDir + 'epoch{:0>2}.npy'.format(epoch), y_prob)
y_predict = MyCorpus.topK(y_prob, topK=1)
predictions = MyCorpus.oneHotDecode(y_predict)
ground_truth = MyCorpus.oneHotDecode(self.y_test)
metrics = evaluate(predictions, ground_truth)
logging.info('acc, ma_f1, mi_f1: {}'.format(metrics[0], metrics[3],
metrics[6]))
with open(self.saveDir + 'metric.txt', 'a') as metricFile:
metricFile.write('\t'.join(map(str, metrics)) + '\n')
self.history.append(metrics)
def getRawLabels():
global trainFilePath, testFilePath, corpus
_, y_train = corpus.loadFile(filePath=trainFilePath)
predictions = MyCorpus.oneHotDecode(y_train)
with open('raw.pkl', 'wb') as outputFile:
pickle.dump(predictions, outputFile, pickle.HIGHEST_PROTOCOL)
with open(dataSetName + 'Label2idx.pkl', 'wb') as outputFile:
pickle.dump(corpus.label2idx, outputFile, pickle.HIGHEST_PROTOCOL)
def stat1():
from util.corpus import MyCorpus
from eval.evaluation import evaluate
dataSetName = 'BBN'
trainFilePath = 'data/{}/train.json'.format(dataSetName)
testFilePath = 'data/{}/test.json'.format(dataSetName)
corpus = MyCorpus(filePathList=[trainFilePath, testFilePath])
X_train, y_train = corpus.loadFile(filePath=trainFilePath)
X_test, y_test = corpus.loadFile(filePath=testFilePath)
y = np.load('10_24_17_32/epoch17.npy')
ground = MyCorpus.oneHotDecode(y_test)
scores = []
outputFile = open('a.txt', 'w', encoding='utf-8')
for threshold in np.arange(0.3, 0.6, 0.01):
y1 = MyCorpus.threshold(y, threshold)
y1 = MyCorpus.oneHotDecode(y1)
score = evaluate(y1, ground)
scores.append((threshold, score))
outputFile.write('%.3f\t%.3f\t%.3f\t%.3f\n' %
(threshold, score[0], score[3], score[6]))
print(scores)
print('Max: ' +
str(max(scores, key=lambda kv: sum([kv[1][0], kv[1][3], kv[1][6]]))))
outputFile.close()
def on_epoch_end(self, epoch, logs={}):
y_prob = self.model.predict(self.X_test)
np.save(self.saveDir + 'epoch{:0>2}.npy'.format(epoch), y_prob)
scores = []
for threshold in np.arange(0.3, 0.6, 0.1):
y_predict = MyCorpus.threshold(y_prob, threshold=threshold)
predictions = MyCorpus.oneHotDecode(y_predict)
ground_truth = MyCorpus.oneHotDecode(self.y_test)
metrics = evaluate(predictions, ground_truth)
scores.append(metrics)
maxScore = max(scores, key=lambda kv: sum([kv[0], kv[3], kv[6]]))
logging.info('acc, ma_f1, mi_f1: {}'.format(maxScore[0], maxScore[3],
maxScore[6]))
with open(self.saveDir + 'metric.txt', 'a') as metricFile:
metricFile.write('\t'.join(map(str, maxScore)) + '\n')
self.history.append(maxScore)
def trainLSTMSingle():
global trainFilePath, testFilePath, corpus
X_train, y_train = corpus.loadFileSingleSent(filePath=trainFilePath)
X_test, y_test = corpus.loadFileSingleSent(filePath=testFilePath)
model = lstmSingle(corpus)
metricHistory = MetricHistory(X_test, y_test)
model.fit(X_train,
y_train,
epochs=10,
batch_size=128,
validation_split=0.1,
shuffle=True,
callbacks=[metricHistory])
y_prob = model.predict(X_train)
y_predict = MyCorpus.hybrid(y_prob, threshold=0.5)
predictions = MyCorpus.oneHotDecode(y_predict)
with open(metricHistory.saveDir + 'LSTMSingle.pkl', 'wb') as outputFile:
pickle.dump(predictions, outputFile, pickle.HIGHEST_PROTOCOL)
from util.corpus import MyCorpus
from util.callback import MetricHistory
# :: Logging level ::
logging.basicConfig(
level=logging.INFO,
format=
'%(asctime)s %(filename)s[line:%(lineno)d] %(levelname)s %(message)s',
datefmt='%a, %d %b %Y %H:%M:%S',
)
dataSetName = 'Wiki'
trainFilePath = 'data/{}/train.json'.format(dataSetName)
testFilePath = 'data/{}/test.json'.format(dataSetName)
corpus = MyCorpus(filePathList=[trainFilePath, testFilePath])
X_train, y_train = corpus.loadFile(filePath=trainFilePath)
X_test, y_test = corpus.loadFile(filePath=testFilePath)
modelName = dataSetName + '_bilstm.h5'
if len(sys.argv) > 1 and sys.argv[1] == 'eval':
model = load_model(modelName)
else:
if os.path.exists(modelName):
model = load_model(modelName)
else:
model = lstm(corpus)
metricHistory = MetricHistory(X_test, y_test)
model.fit(X_train,
y_train,
from util.corpus import MyCorpus
from util.callback import MetricHistory, MetricHistorySoftmax
# :: Logging level ::
logging.basicConfig(
level=logging.INFO,
format=
'%(asctime)s %(filename)s[line:%(lineno)d] %(levelname)s %(message)s',
datefmt='%a, %d %b %Y %H:%M:%S',
)
dataSetName = 'baike'
trainFilePath = 'data/{}/train.json'.format(dataSetName)
testFilePath = 'data/{}/test.json'.format(dataSetName)
corpus = MyCorpus(filePathList=[trainFilePath, testFilePath])
def getRawLabels():
global trainFilePath, testFilePath, corpus
_, y_train = corpus.loadFile(filePath=trainFilePath)
predictions = MyCorpus.oneHotDecode(y_train)
with open('raw.pkl', 'wb') as outputFile:
pickle.dump(predictions, outputFile, pickle.HIGHEST_PROTOCOL)
with open(dataSetName + 'Label2idx.pkl', 'wb') as outputFile:
pickle.dump(corpus.label2idx, outputFile, pickle.HIGHEST_PROTOCOL)
def trainLSTM():
global trainFilePath, testFilePath, corpus