105 lines
3.6 KiB
Python
105 lines
3.6 KiB
Python
# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
|
|
#
|
|
# Licensed under the Apache License, Version 2.0 (the "License");
|
|
# you may not use this file except in compliance with the License.
|
|
# You may obtain a copy of the License at
|
|
#
|
|
# http://www.apache.org/licenses/LICENSE-2.0
|
|
#
|
|
# Unless required by applicable law or agreed to in writing, software
|
|
# distributed under the License is distributed on an "AS IS" BASIS,
|
|
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
# See the License for the specific language governing permissions and
|
|
# limitations under the License.
|
|
|
|
import unittest
|
|
|
|
import numpy as np
|
|
from op_test import get_places
|
|
|
|
import paddle
|
|
from paddle import base
|
|
from paddle.tensor.manipulation import tensor_array_to_tensor
|
|
|
|
paddle.enable_static()
|
|
|
|
|
|
def build_and_run_program(place, batch_size, beam_size, stop_gradient=False):
|
|
paddle.seed(1)
|
|
np.random.seed(2)
|
|
|
|
x = paddle.assign(
|
|
np.random.rand(batch_size, beam_size, 2).astype("float32")
|
|
)
|
|
indices = paddle.static.data(
|
|
shape=[None, beam_size], dtype="int64", name="indices"
|
|
)
|
|
step_idx = paddle.tensor.fill_constant(
|
|
shape=[1], dtype="int64", value=0, force_cpu=True
|
|
)
|
|
max_len = paddle.tensor.fill_constant(
|
|
shape=[1], dtype="int64", value=10, force_cpu=True
|
|
)
|
|
cond = paddle.less_than(x=step_idx, y=max_len)
|
|
while_op = paddle.static.nn.control_flow.While(cond)
|
|
scores = paddle.tensor.array_write(x, step_idx)
|
|
with while_op.block():
|
|
bs = paddle.cast(paddle.shape(x)[0], "int64")
|
|
for _ in range(2):
|
|
bs = paddle.cast(bs, 'int64')
|
|
bs.stop_gradient = stop_gradient
|
|
batch_pos = paddle.expand(
|
|
paddle.unsqueeze(paddle.arange(0, bs, 1, dtype=bs.dtype), [1]),
|
|
[-1, beam_size],
|
|
)
|
|
topk_coordinates = paddle.stack([batch_pos, indices], axis=2)
|
|
topk_coordinates.stop_gradient = stop_gradient
|
|
score = paddle.gather_nd(x, topk_coordinates)
|
|
paddle.increment(x=step_idx, value=1.0)
|
|
paddle.tensor.array_write(score, i=step_idx, array=scores)
|
|
length_cond = paddle.less_than(x=step_idx, y=max_len)
|
|
paddle.assign(length_cond, cond)
|
|
scores.stop_gradient = True
|
|
out = tensor_array_to_tensor(scores, axis=0, use_stack=True)[0]
|
|
loss = paddle.mean(out)
|
|
opt = paddle.optimizer.Adam(0.01)
|
|
opt.minimize(loss)
|
|
exe = base.Executor(place)
|
|
data = np.random.random_integers(
|
|
low=0, high=beam_size - 1, size=(batch_size, beam_size)
|
|
).astype("int64")
|
|
exe.run(paddle.static.default_startup_program())
|
|
(loss_val,) = exe.run(feed={"indices": data}, fetch_list=[loss])
|
|
|
|
return loss_val
|
|
|
|
|
|
class TestDynRNNStopGradient(unittest.TestCase):
|
|
def setUp(self):
|
|
self.batch_size = 2
|
|
self.beam_size = 2
|
|
|
|
def run_main(self, place):
|
|
with paddle.pir_utils.IrGuard():
|
|
main_program = paddle.static.Program()
|
|
startup_program = paddle.static.Program()
|
|
with (
|
|
paddle.static.program_guard(main_program, startup_program),
|
|
base.scope_guard(base.Scope()),
|
|
):
|
|
value1 = build_and_run_program(
|
|
place, self.batch_size, self.beam_size, False
|
|
)
|
|
value2 = build_and_run_program(
|
|
place, self.batch_size, self.beam_size, True
|
|
)
|
|
np.testing.assert_array_equal(value1, value2)
|
|
|
|
def test_check_main(self):
|
|
for p in get_places():
|
|
self.run_main(p)
|
|
|
|
|
|
if __name__ == '__main__':
|
|
unittest.main()
|