def event_handler(event):
if isinstance(event, paddle.event.EndIteration):
if event.batch_id % 100 == 0:
print "\nPass %d, Batch %d, Cost %f, %s" % (
event.pass_id, event.batch_id, event.cost, event.metrics)
else:
sys.stdout.write('.')
sys.stdout.flush()
if isinstance(event, paddle.event.EndPass):
# save parameters
with gzip.open('params_pass_%d.tar.gz' % event.pass_id, 'w') as f:
parameters.to_tar(f)
reader_test = data_provider.data_reader('data/train', 10000)
result = trainer.test(reader=paddle.batch(reader=reader_test,
batch_size=16),
feeding=feeding)
print "\nTest with Pass %d, %s" % (event.pass_id, result.metrics)
def event_handler(event):
if isinstance(event, paddle.event.EndIteration):
if event.batch_id % 50 == 0:
print ("\n pass %d, Batch: %d cost: %f" % (event.pass_id, event.batch_id, event.cost))
else:
sys.stdout.write('.')
sys.stdout.flush()
if isinstance(event, paddle.event.EndPass):
# save parameters
feeding = {'x': 0,
'y': 1}
with gzip.open('output/params_pass_%d.tar.gz' % event.pass_id, 'w') as f:
parameters.to_tar(f)
filepath = 'data/test.data'
result = trainer.test(
reader=paddle.batch(data_provider.data_reader(filepath), batch_size=16),
feeding=feeding)
print ("\nTest with Pass %d, cost: %s" % (event.pass_id, result.cost))
def event_handler(event):
if isinstance(event, paddle.event.EndIteration):
if event.batch_id % 100 == 0:
print "\n pass %d, Batch: %d cost: %f, %s" % (
event.pass_id, event.batch_id, event.cost, event.metrics)
else:
sys.stdout.write('.')
sys.stdout.flush()
if isinstance(event, paddle.event.EndPass):
# save parameters
feeding = {'image': 0, 'label': 1}
with gzip.open('output/params_pass_%d.tar.gz' % event.pass_id,
'w') as f:
parameters.to_tar(f)
filepath = '/Users/vic/Dev/DeepLearning/Paddle/VGG-CIFAR/Images/cifar-10-batches-py/test_batch'
result = trainer.test(reader=paddle.batch(data_provider.data_reader(
filepath, 0),
batch_size=128),
feeding=feeding)
print "\nTest with Pass %d, %s" % (event.pass_id, result.metrics)
def event_handler(event):
if isinstance(event, paddle.event.EndIteration):
if event.batch_id % 100 == 0:
print "\nPass: %d Batch: %d [Cost: %f ][%s]\n" % (event.pass_id, event.batch_id, event.cost, event.metrics)
else:
sys.stdout.write('.')
sys.stdout.flush()
if isinstance(event, paddle.event.EndPass):
#save parameters
with gzip.open('output/params.tar.gz', 'w') as f:
parameters.to_tar(f)
# test
feeding = {'image': 0,
'label': 1}
filepath = ""
result = trainer.test(reader=paddle.batch(reader=data_provider.data_reader(filepath, 0), batch_size=128),
feeding=feeding)
print "\nTest Result: [Cost: %f] [%s] " % (result.cost, result.metrics)
act=paddle.activation.Softmax())
label = paddle.layer.data(name='label',
type=paddle.data_type.integer_value(10))
cost = paddle.layer.classification_cost(input=output_layer, label=label)
parameters = paddle.parameters.create(cost)
print parameters.keys()
optimizer = paddle.optimizer.Momentum(
momentum=0.9,
regularization=paddle.optimizer.L2Regularization(rate=0.0002 * 128),
learning_rate=0.1 / 128.0,
learning_rate_decay_a=0.1,
learning_rate_decay_b=50000 * 100,
learning_rate_schedule='discexp')
trainer = paddle.trainer.SGD(cost=cost,
parameters=parameters,
update_equation=optimizer)
feeding = {'image': 0, 'label': 1}
file_path = '/Users/vic/Dev/DeepLearning/Paddle/VGG-CIFAR/Images/cifar-10-batches-py/data_batch_1'
reader = data_provider.data_reader(file_path, 0)
trainer.train(num_passes=100,
reader=paddle.batch(reader, batch_size=16),
event_handler=event_handler,
feeding=feeding)
parameters=parameters,
update_equation=momentum_optimizer)
feeding = {'image': 0, 'label': 1}
def event_handler(event):
if isinstance(event, paddle.event.EndIteration):
if event.batch_id % 100 == 0:
print "\nPass %d, Batch %d, Cost %f, %s" % (
event.pass_id, event.batch_id, event.cost, event.metrics)
else:
sys.stdout.write('.')
sys.stdout.flush()
if isinstance(event, paddle.event.EndPass):
# save parameters
with gzip.open('params_pass_%d.tar.gz' % event.pass_id, 'w') as f:
parameters.to_tar(f)
reader_test = data_provider.data_reader('data/train', 10000)
result = trainer.test(reader=paddle.batch(reader=reader_test,
batch_size=16),
feeding=feeding)
print "\nTest with Pass %d, %s" % (event.pass_id, result.metrics)
reader = data_provider.data_reader('data/train', 50000)
trainer.train(reader=paddle.batch(reader=reader, batch_size=16),
num_passes=50,
event_handler=event_handler,
feeding=feeding)
def train(x_, model_path, is_predict=False):
paddle.init(use_gpu=False, trainer_count=1)
# 步长
TIME_STEP = 10
x = paddle.layer.data(
name='x', type=paddle.data_type.dense_vector_sequence(TIME_STEP))
# 模型输出
output = network(x)
# 训练
if not is_predict:
label = paddle.layer.data(name='y',
type=paddle.data_type.dense_vector(dim=1))
# 计算损失
loss = paddle.layer.mse_cost(input=output, label=label)
parameters = paddle.parameters.create(loss)
# Adam 优化算法
optimizer = paddle.optimizer.Adam(
learning_rate=1e-3,
regularization=paddle.optimizer.L2Regularization(rate=8e-4))
trainer = paddle.trainer.SGD(cost=loss,
parameters=parameters,
update_equation=optimizer)
feeding = {'x': 0, 'y': 1}
# 训练事件处理
def event_handler(event):
if isinstance(event, paddle.event.EndIteration):
if event.batch_id % 50 == 0:
print("\n pass %d, Batch: %d cost: %f" %
(event.pass_id, event.batch_id, event.cost))
else:
sys.stdout.write('.')
sys.stdout.flush()
if isinstance(event, paddle.event.EndPass):
# save parameters
feeding = {'x': 0, 'y': 1}
with gzip.open('output/params_pass_%d.tar.gz' % event.pass_id,
'w') as f:
parameters.to_tar(f)
filepath = 'data/test.data'
# 测试数据
test_reader = data_provider.data_reader(filepath)
result = trainer.test(reader=paddle.batch(test_reader,
batch_size=16),
feeding=feeding)
print("\nTest with Pass %d, cost: %s" %
(event.pass_id, result.cost))
train_file_path = 'data/train.data'
reader = data_provider.data_reader(train_file_path)
trainer.train(paddle.batch(reader=reader, batch_size=128),
num_passes=200,
event_handler=event_handler,
feeding=feeding)
else:
# 进行预测
# 加载模型参数
with gzip.open(model_path, 'r') as openFile:
parameters = paddle.parameters.Parameters.from_tar(openFile)
# 使用infer进行预测
result = paddle.infer(input=x_,
parameters=parameters,
output_layer=output,
feeding={'x': 0})
return result