def test_api(self):
if ver.mkl() == "ON" and 'Linux' in platform.platform():
import paddle.fluid as fluid
dict_size = 20
data_t = fluid.layers.data(name='word',
shape=[1],
dtype='int64',
lod_level=1)
padding_idx = np.random.randint(1, 10)
out = fluid.contrib.fused_embedding_seq_pool(
input=data_t,
size=[dict_size, 32],
param_attr='w',
padding_idx=padding_idx,
is_sparse=False)
place = fluid.CPUPlace()
exe = fluid.Executor(place)
exe.run(fluid.default_startup_program())
# prepare input words' idx
x_tensor = fluid.core.LoDTensor()
idxs = np.random.randint(1, 10, (8)).astype("int64")
x_tensor.set(idxs, place)
x_tensor.set_recursive_sequence_lengths([[4, 4]])
ret = exe.run(feed={'word': x_tensor}, fetch_list=[out])
def test_check_grad(self):
# TODO(wangzhongpu): support lod in dygraph mode
if ver.mkl() == "ON" and 'Linux' in platform.platform():
self.attrs = {'is_sparse': False}
self.check_grad(['W'],
'Out',
no_grad_set=['Ids'],
check_dygraph=False)
def test_check_grad(self):
if ver.mkl() == "ON" and 'Linux' in platform.platform():
ids = np.squeeze(self.ids, axis=2)
padding_idx = np.random.choice(ids.flatten(), 1)[0]
self.attrs = {'padding_idx': int(padding_idx), 'is_sparse': False}
# TODO(wangzhongpu): support lod in dygraph mode
self.check_grad(
['W'], 'Out', no_grad_set=['Ids'], check_dygraph=False)
def test_check_output(self):
if ver.mkl() == "ON" and 'Linux' in platform.platform():
ids = np.squeeze(self.ids, axis=2)
padding_idx = np.random.choice(ids.flatten(), 1)[0]
output = list()
index = 0
for count in self.lod[0]:
arr = ids[index:count + index]
out = np.reshape(self.table[arr.flatten()],
[arr.shape[0], arr.shape[1], self.emb_size])
idx = np.argwhere(arr == padding_idx)
for item in idx:
out[item[0], item[1], :] = np.zeros(self.emb_size)
output.append(np.sum(out, 0))
index += count
self.outputs = {
'Out': np.reshape(
np.array(output), [len(self.lod[0]), 2 * self.emb_size])
}
self.attrs = {'padding_idx': int(padding_idx)}
self.check_output()
def test_check_grad(self):
if ver.mkl() == "ON" and 'Linux' in platform.platform():
ids = np.squeeze(self.ids, axis=2)
padding_idx = np.random.choice(ids.flatten(), 1)[0]
self.attrs = {'padding_idx': int(padding_idx), 'is_sparse': False}
self.check_grad(['W'], 'Out', no_grad_set=('Ids'))
def test_check_grad(self):
if ver.mkl() == "ON" and 'Linux' in platform.platform():
self.attrs = {'is_sparse': False}
self.check_grad(['W'], 'Out', no_grad_set=('Ids'))