def test_param_type(self):
program_translator = ProgramTranslator()
program_translator.enable(True)
x_data = np.random.random((20, 20)).astype('float32')
with fluid.dygraph.guard(fluid.CPUPlace()):
net = SimpleFcLayer(20)
x = fluid.dygraph.to_variable(x_data)
out = net(x)
program_cache = net.forward.program_cache
_, (concrete_program, _) = program_cache.last()
params = concrete_program.parameters
concrete_program.parameters = params[0]
# TypeError: Type of self._params should be list or tuple,
# but received <class 'paddle.fluid.framework.ParamBase'>.
with self.assertRaises(TypeError):
partial_program_from(concrete_program)
params[0] = "linear.w.0"
concrete_program.parameters = params
# TypeError: Type of self._params[0] should be framework.ParamBase,
# but received <type 'str'>.
with self.assertRaises(TypeError):
partial_program_from(concrete_program)
def test_save_inference_model(self):
fc_size = 20
x = np.random.random((fc_size, fc_size)).astype('float32')
layer = SimpleFcLayer(fc_size)
program_translator = ProgramTranslator.get_instance()
program_cache = ProgramTranslator().get_program_cache
adam = fluid.optimizer.SGD(learning_rate=0.001)
program_translator.set_optimizer(adam, 'mean')
for i in range(5):
out = layer(x)
main_program = ProgramTranslator.get_instance().main_program
expected_persistable_vars = set(
[layer._linear.weight.name, layer._linear.bias.name])
infer_model_dir = "./test_dy2stat_save_inference_model"
ProgramTranslator.get_instance().save_inference_model(infer_model_dir)
saved_var_names = set([
filename for filename in os.listdir(infer_model_dir)
if filename != '__model__'
])
self.assertEqual(saved_var_names, expected_persistable_vars)
def _run(self, func, x, with_exception, to_static):
ProgramTranslator().enable(to_static)
if with_exception:
with self.assertRaises(BaseException):
with fluid.dygraph.guard():
func(x)
else:
with fluid.dygraph.guard():
func(x)
def train(self, to_static=False):
program_translator = ProgramTranslator()
program_translator.enable(to_static)
with fluid.dygraph.guard():
dy_layer = self.dygraph_class()
x = fluid.dygraph.to_variable(self.data)
prediction = dy_layer(x)
if isinstance(prediction, (list, tuple)):
prediction = prediction[0]
return prediction.numpy()
def train(self, to_static=False):
prog_trans = ProgramTranslator()
prog_trans.enable(to_static)
loss_data = []
with fluid.dygraph.guard(self.place):
fluid.default_main_program().random_seed = SEED
fluid.default_startup_program().random_seed = SEED
mnist = MNIST()
adam = AdamOptimizer(learning_rate=0.001,
parameter_list=mnist.parameters())
for epoch in range(self.epoch_num):
start = time()
for batch_id, data in enumerate(self.train_reader()):
dy_x_data = np.array([
x[0].reshape(1, 28, 28) for x in data
]).astype('float32')
y_data = np.array([x[1] for x in data
]).astype('int64').reshape(-1, 1)
img = to_variable(dy_x_data)
label = to_variable(y_data)
label.stop_gradient = True
prediction, acc, avg_loss = mnist(img, label=label)
avg_loss.backward()
adam.minimize(avg_loss)
loss_data.append(avg_loss.numpy()[0])
# save checkpoint
mnist.clear_gradients()
if batch_id % 10 == 0:
print(
"Loss at epoch {} step {}: loss: {:}, acc: {}, cost: {}"
.format(epoch, batch_id, avg_loss.numpy(),
acc.numpy(),
time() - start))
start = time()
if batch_id == 50:
mnist.eval()
prediction, acc, avg_loss = mnist(img, label)
loss_data.append(avg_loss.numpy()[0])
# new save load check
self.check_jit_save_load(mnist, [dy_x_data], [img],
to_static, prediction)
break
return loss_data
def train_static(self):
main_program = fluid.Program()
loss_data = []
with fluid.program_guard(main_program):
static_net = self.dygraph_class()
adam = fluid.optimizer.AdamOptimizer(learning_rate=0.001)
# set optimizer
# TODO: Need a better interfaces to set optimizer.
program_translator = ProgramTranslator()
program_translator.set_optimizer(adam, 'avg_loss')
for batch_id in range(self.batch_num):
pred, avg_loss = static_net(self.data)
loss_data.append(np.array(avg_loss))
return loss_data
def test_save_load_same_result(self):
program_translator = ProgramTranslator()
x_data = np.random.randn(30, 10, 32).astype('float32')
batch_num = 3
with fluid.dygraph.guard(place):
program_translator.enable(True)
x = fluid.dygraph.to_variable(x_data)
net = Linear(32, 64)
adam = AdamOptimizer(learning_rate=0.1,
parameter_list=net.parameters())
for i in range(batch_num):
static_out, static_loss = net(x)
# Update parameters
static_loss.backward()
adam.minimize(static_loss)
net.clear_gradients()
# Save parameters
fluid.save_dygraph(net.state_dict(), self.model_path)
# minimize() will update parameter, call net() to get output and avg_loss.
# Switch into eval mode.
net.eval()
static_out, static_loss = net(x)
# load parameters into dygraph
with fluid.dygraph.guard(place):
dygraph_net = Linear(32, 64)
# Load parameters
model_dict, _ = fluid.load_dygraph(self.model_path)
dygraph_net.set_dict(model_dict)
# Switch into eval mode.
dygraph_net.eval()
x = fluid.dygraph.to_variable(x_data)
# predict output
program_translator.enable(False)
dygraph_out, dygraph_loss = dygraph_net(x)
self.assertTrue(np.allclose(dygraph_out.numpy(), static_out.numpy()))
self.assertTrue(np.allclose(dygraph_loss.numpy(), static_loss.numpy()))
def train(self, to_static=False):
prog_trans = ProgramTranslator()
prog_trans.enable(to_static)
with fluid.dygraph.guard(fluid.CPUPlace()):
dygraph_net = self.dygraph_class()
adam = fluid.optimizer.AdamOptimizer(
learning_rate=0.001, parameter_list=dygraph_net.parameters())
loss_data = []
for batch_id in range(self.batch_num):
input = fluid.dygraph.to_variable(self.data)
pred, avg_loss = dygraph_net(input)
loss_data.append(avg_loss.numpy())
avg_loss.backward()
adam.minimize(avg_loss)
dygraph_net.clear_gradients()
return loss_data
def check_save_inference_model(self,
model,
inputs,
gt_out,
feed=None,
fetch=None):
program_translator = ProgramTranslator()
expected_persistable_vars = set([p.name for p in model.parameters()])
infer_model_dir = "./test_dy2stat_save_inference_model"
program_translator.save_inference_model(infer_model_dir,
feed=feed,
fetch=fetch)
saved_var_names = set([
filename for filename in os.listdir(infer_model_dir)
if filename != '__model__'
])
self.assertEqual(saved_var_names, expected_persistable_vars)
# Check the correctness of the inference
infer_out = self.load_and_run_inference(infer_model_dir, inputs)
self.assertTrue(np.allclose(gt_out, infer_out))
def test_program_translator(self):
answer = "\
def dyfunc_with_if_else(x_v, label=None):\n\
\n\
def true_fn_1(x_v):\n\
x_v = x_v - 1\n\
return x_v\n\
\n\
def false_fn_1(x_v):\n\
x_v = x_v + 1\n\
return x_v\n\
x_v = fluid.layers.cond(fluid.layers.mean(x_v)[0] > 5, lambda :\n\
true_fn_1(x_v), lambda : false_fn_1(x_v))\n\
if label is not None:\n\
loss = fluid.layers.cross_entropy(x_v, label)\n\
return loss\n\
return x_v\n"
program_translator = ProgramTranslator()
code = program_translator.get_code(dyfunc_with_if_else)
self.assertEqual(answer, code)
# limitations under the License.
import time
import unittest
import numpy as np
import paddle.fluid as fluid
from paddle.fluid.dygraph.dygraph_to_static import ProgramTranslator
from paddle.fluid.dygraph.io import VARIABLE_FILENAME
from bert_dygraph_model import PretrainModelLayer
from bert_utils import get_bert_config, get_feed_data_reader
from predictor_utils import PredictorTools
program_translator = ProgramTranslator()
place = fluid.CUDAPlace(
0) if fluid.is_compiled_with_cuda() else fluid.CPUPlace()
SEED = 2020
STEP_NUM = 10
PRINT_STEP = 2
MODEL_SAVE_PATH = "./bert.inference.model"
DY_STATE_DICT_SAVE_PATH = "./bert.dygraph"
def train(bert_config, data_reader, to_static):
with fluid.dygraph.guard(place):
fluid.default_main_program().random_seed = SEED
fluid.default_startup_program().random_seed = SEED
data_loader = fluid.io.DataLoader.from_generator(capacity=50,
def setUp(self):
self.x = np.random.randn(30, 10, 32).astype('float32')
self.weight = np.random.randn(32, 64).astype('float32')
self.program_translator = ProgramTranslator()
def test_program_translator(self):
answer = get_source_code(StaticCode2.dyfunc_with_if_else)
program_translator = ProgramTranslator()
code = program_translator.get_code(dyfunc_with_if_else)
self.assertEqual(answer, code)
def test_decorator(self):
x_v = None
program_translator = ProgramTranslator()
code = program_translator.get_code(dyfunc_with_if_else)
answer = get_source_code(StaticCode1.dyfunc_with_if_else)
self.assertEqual(answer, code)
def test_code(self):
answer = get_source_code(StaticCode.dyfunc_assign)
program_translator = ProgramTranslator()
code = program_translator.get_code(dyfunc_assign)
self.assertEqual(answer, code)
def setUp(self):
self.program_translator = ProgramTranslator()
self.x = np.random.randn(10, 32).astype('float32')
