# 处理数据
X, y, classes = load_hotel_comment()
X_train, X_test, y_train, y_test = train_test_split(X,
y,
train_size=0.8,
random_state=7432)
num_classes = len(classes)
# 转化成字id
tokenizer = SimpleTokenizer()
tokenizer.fit(X_train)
X_train = tokenizer.transform(X_train)
X_test = tokenizer.transform(X_test)
maxlen = find_best_maxlen(X_train, mode="max")
maxlen = 128
X_train = sequence.pad_sequences(X_train,
maxlen=maxlen,
dtype="int32",
padding="post",
truncating="post",
value=0.0)
X_test = sequence.pad_sequences(X_test,
maxlen=maxlen,
dtype="int32",
padding="post",
truncating="post",
value=0.0)
self).train_step(data) # 执行普通的train_step
embeddings.assign_sub(delta) # 删除Embedding矩阵上的扰动
return results
X, y, classes = load_hotel_comment()
X_train, X_test, y_train, y_test = train_test_split(X,
y,
train_size=0.6,
random_state=7384672)
num_classes = len(classes)
tokenizer = SimpleTokenizer()
tokenizer.fit(X_train)
maxlen = find_best_maxlen(X_train)
# maxlen = 256
def create_dataset(X, y, maxlen=maxlen):
X = tokenizer.transform(X)
X = sequence.pad_sequences(X,
maxlen=maxlen,
dtype="int32",
padding="post",
truncating="post",
value=0.0)
y = tf.keras.utils.to_categorical(y)
return X, y
value=0)
return X
def create_dataset(Xa, Xp, Xn, maxlen):
Xa = tokenizer.transform(Xa)
Xp = tokenizer.transform(Xp)
Xn = tokenizer.transform(Xn)
Xa = pad(Xa, maxlen)
Xp = pad(Xp, maxlen)
Xn = pad(Xn, maxlen)
return Xa, Xp, Xn
maxlen = find_best_maxlen(Xa)
maxlen = 48
hdims = 128
epochs = 1
num_words = len(tokenizer)
embedding_dims = 128
x1_input = Input(shape=(maxlen, ))
x2_input = Input(shape=(maxlen, ))
x3_input = Input(shape=(maxlen, ))
# 计算全局mask
x1_mask = Lambda(lambda x: tf.not_equal(x, 0))(x1_input)
x2_mask = Lambda(lambda x: tf.not_equal(x, 0))(x2_input)
x3_mask = Lambda(lambda x: tf.not_equal(x, 0))(x3_input)
embedding = Embedding(num_words,