def train(model, batch_size, max_nb_cha, nb_epoch, save_dir, train_data,
val_data):
print 'X_train shape:', X_train.shape
print X_train.shape[0], 'train samples'
if os.path.exists(save_dir) == False:
os.mkdir(save_dir)
start_time = time.time()
save_path = save_dir + 'weights.{epoch:02d}-{val_loss:.2f}.hdf5'
check_pointer = ModelCheckpoint(save_path, save_best_only=True)
# history = model.fit(train_data, batch_size=batch_size, nb_epoch=nb_epoch,
# validation_data=val_data,
# validation_split=0.01,
# callbacks=[check_pointer])
history = model.fit(
train_data['input'],
train_data['output1'],
batch_size=batch_size,
nb_epoch=nb_epoch,
validation_data=(val_data['input'], val_data['output1']),
# validation_split=0.1,
callbacks=[check_pointer])
plot_loss_figure(
history,
save_dir + str(datetime.now()).split('.')[0].split()[1] + '.jpg')
print 'Training time(h):', (time.time() - start_time) / 3600
def train(model, batch_size, nb_epoch, save_dir, train_data, val_data,
char_set):
X_train, y_train = train_data[0], train_data[1]
sample_weight = get_sample_weight(y_train, char_set)
print 'X_train shape:', X_train.shape
print X_train.shape[0], 'train samples'
if os.path.exists(save_dir) == False:
os.mkdir(save_dir)
start_time = time.time()
save_path = save_dir + 'weights.{epoch:02d}-{val_loss:.2f}.hdf5'
check_pointer = ModelCheckpoint(save_path, save_best_only=True)
history = model.fit(X_train,
y_train,
batch_size=batch_size,
nb_epoch=nb_epoch,
validation_data=val_data,
validation_split=0.1,
callbacks=[check_pointer],
sample_weight=sample_weight)
plot_loss_figure(
history,
save_dir + str(datetime.now()).split('.')[0].split()[1] + '.jpg')
print 'Training time(h):', (time.time() - start_time) / 3600
def train(model, batch_size, nb_epoch, save_dir, train_X, train_y, val_data):
print('train_X shape:', train_X.shape)
print(train_X.shape[0], 'train samples')
if os.path.exists(save_dir) == False:
os.mkdir(save_dir)
start_time = time.time()
save_path = save_dir + 'weights.{epoch:02d}-{val_loss:.3f}.h5'
check_pointer = ModelCheckpoint(save_path,
save_best_only=True)
history = model.fit(train_X, train_y, batch_size=batch_size, nb_epoch=nb_epoch,
validation_data=val_data,
validation_split=0.1,
callbacks=[check_pointer],
)
#model.save(save_path)
plot_loss_figure(history, save_dir + str(datetime.now()).split('.')[0].split()[1]+'.png')
print('Training time(h):', (time.time()-start_time) / 3600)
def train_on_generator(model, batch_size, max_nb_cha, nb_epoch, save_dir,
generator, val_data):
print 'Train using generator'
if os.path.exists(save_dir) == False:
os.mkdir(save_dir)
start_time = time.time()
save_path = save_dir + 'weights.{epoch:02d}-{val_loss:.2f}.hdf5'
check_pointer = ModelCheckpoint(save_path)
samples_per_epoch = 50000 # 每个epoch跑多少数据
history = model.fit_generator(generator,
samples_per_epoch=samples_per_epoch,
nb_epoch=nb_epoch,
nb_worker=4,
validation_data=val_data,
callbacks=[check_pointer])
plot_loss_figure(
history,
save_dir + str(datetime.now()).split('.')[0].split()[1] + '.jpg')
print 'Training time(h):', (time.time() - start_time) / 3600
def train(model, batch_size, nb_epoch, save_dir, train_data, val_data, char_set):
X_train, y_train = train_data[0], train_data[1]
sample_weight = get_sample_weight(y_train, char_set)
print 'X_train shape:', X_train.shape
print X_train.shape[0], 'train samples'
if os.path.exists(save_dir) == False:
os.mkdir(save_dir)
start_time = time.time()
save_path = save_dir + 'weights.{epoch:02d}-{val_loss:.2f}.hdf5'
check_pointer = ModelCheckpoint(save_path,
save_best_only=True)
history = model.fit(X_train, y_train, batch_size=batch_size, nb_epoch=nb_epoch,
validation_data=val_data,
validation_split=0.1,
callbacks=[check_pointer],
sample_weight=sample_weight
)
plot_loss_figure(history, save_dir + str(datetime.now()).split('.')[0].split()[1]+'.jpg')
print 'Training time(h):', (time.time()-start_time) / 3600
def train(model, batch_size, nb_epoch, save_dir, train_data, val_data,
char_set):
#pdb.set_trace()
X_train, y_train = train_data[0], train_data[1]
print('X_train shape:', X_train.shape)
print(X_train.shape[0], 'train samples')
if os.path.exists(save_dir) == False:
os.mkdir(save_dir)
start_time = time.time()
save_path = save_dir + 'weights.{epoch:02d}-{val_loss:.2f}.hdf5'
check_pointer = ModelCheckpoint(save_path, save_best_only=True)
hist = model.fit(X_train,
y_train,
batch_size=batch_size,
nb_epoch=nb_epoch,
validation_data=val_data,
validation_split=0.1,
callbacks=[check_pointer],
sample_weight=None) #返回记录字典，包括每一次迭代的训练误差率和验证误差率
plot_loss_figure(
hist, save_dir + str(datetime.now()).split('.')[0].split()[1] + '.png')
print('Training time(h):', (time.time() - start_time) / 3600)
