os.environ["CUDA_VISIBLE_DEVICES"] = '2'
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
file = '../datasets/sentiment_train.csv'
xtrain, xtest, y_train, y_test, real_xtest, vocab_size = get_train_test_df(
file)
maxlen = len(max((xtrain + xtest), key=len))
maxlen = 100
embedding_dim = 128
model = Sequential()
model.add(
layers.Embedding(input_dim=vocab_size,
output_dim=embedding_dim,
input_length=maxlen))
model.add(layers.GlobalAveragePooling1D())
model.add(layers.Dense(10))
model.add(layers.Dense(1, activation="sigmoid"))
# train_model(xtrain, xtest, y_train, y_test, maxlen, model, epochs=15)
train_model(xtrain,
xtest,
y_train,
y_test,
maxlen,
model,
epochs=30,
real_xtest=real_xtest)
from keras.models import Sequential
from keras import layers, regularizers
import os
from sentiment.cross_base import train_model, get_train_test_df
os.environ["CUDA_VISIBLE_DEVICES"] = '2'
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
file = '../datasets/sentiment_train.csv'
xtrain, xtest, y_train, y_test, real_xtest, vocab_size = get_train_test_df(
file)
maxlen = len(max((xtrain + xtest), key=len))
maxlen = 128
embedding_dim = 128
model = Sequential()
model.add(
layers.Embedding(input_dim=vocab_size,
output_dim=embedding_dim,
input_length=maxlen))
# model.add(layers.Bidirectional(layers.LSTM(32, return_sequences=True)))
model.add(
layers.Bidirectional(
layers.LSTM(16, kernel_regularizer=regularizers.l2(0.001))))
model.add(layers.Dense(10, activation='relu'))
model.add(layers.Dense(1, activation='sigmoid'))
train_model(xtrain, xtest, y_train, y_test, maxlen, model, epochs=10)
from keras.models import Sequential
from keras import layers
import os
from sentiment.cross_base import train_model, get_train_test_df
os.environ["CUDA_VISIBLE_DEVICES"] = '2'
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
file = '../datasets/sentiment_train.csv'
xtrain, xtest, y_train, y_test, vocab_size = get_train_test_df(file)
maxlen = len(max((xtrain + xtest), key=len))
maxlen = 100
embedding_dim = 128
model = Sequential()
model.add(layers.Embedding(input_dim=vocab_size,
output_dim=embedding_dim,
input_length=maxlen))
model.add(layers.SimpleRNN(36, input_shape=(1, maxlen), activation='sigmoid'))
model.add(layers.Dense(10))
model.add(layers.Dense(1, activation="sigmoid"))
model.compile(optimizer="adam", loss="binary_crossentropy",
metrics=['accuracy'])
train_model(xtrain, xtest, y_train, y_test, maxlen, model)