未验证 提交 b85af464 编写于 作者: M mhy-666 提交者: GitHub

remove layers.tensor.argmin/argmax/assign/cast/concat/sums (#49944)

上级 61a933ac
......@@ -112,7 +112,7 @@ class Categorical(distribution.Distribution):
self.dtype = logits.dtype
self.logits = self._to_tensor(logits)[0]
if self.dtype != convert_dtype(self.logits.dtype):
self.logits = tensor.cast(self.logits, dtype=self.dtype)
self.logits = paddle.cast(self.logits, dtype=self.dtype)
dist_sum = paddle.sum(self.logits, axis=-1, keepdim=True)
self._prob = self.logits / dist_sum
......
......@@ -200,7 +200,7 @@ class Distribution:
for arg in numpy_args:
arg_broadcasted, _ = np.broadcast_arrays(arg, tmp)
arg_variable = paddle.tensor.create_tensor(dtype=dtype)
tensor.assign(arg_broadcasted, arg_variable)
paddle.assign(arg_broadcasted, arg_variable)
variable_args.append(arg_variable)
return tuple(variable_args)
......@@ -235,7 +235,7 @@ class Distribution:
warnings.warn(
"dtype of input 'value' needs to be the same as parameters of distribution class. dtype of 'value' will be converted."
)
return tensor.cast(value, dtype=param.dtype)
return paddle.cast(value, dtype=param.dtype)
return value
def _probs_to_logits(self, probs, is_binary=False):
......
......@@ -132,8 +132,8 @@ class Normal(distribution.Distribution):
# pylint: disable=unbalanced-tuple-unpacking
self.loc, self.scale = self._to_tensor(loc, scale)
if self.dtype != convert_dtype(self.loc.dtype):
self.loc = tensor.cast(self.loc, dtype=self.dtype)
self.scale = tensor.cast(self.scale, dtype=self.dtype)
self.loc = paddle.cast(self.loc, dtype=self.dtype)
self.scale = paddle.cast(self.scale, dtype=self.dtype)
super().__init__(self.loc.shape)
@property
......
......@@ -137,8 +137,8 @@ class Uniform(distribution.Distribution):
# pylint: disable=unbalanced-tuple-unpacking
self.low, self.high = self._to_tensor(low, high)
if self.dtype != convert_dtype(self.low.dtype):
self.low = tensor.cast(self.low, dtype=self.dtype)
self.high = tensor.cast(self.high, dtype=self.dtype)
self.low = paddle.cast(self.low, dtype=self.dtype)
self.high = paddle.cast(self.high, dtype=self.dtype)
super().__init__(self.low.shape)
......@@ -218,8 +218,8 @@ class Uniform(distribution.Distribution):
name = self.name + '_log_prob'
lb_bool = self.low < value
ub_bool = value < self.high
lb = tensor.cast(lb_bool, dtype=value.dtype)
ub = tensor.cast(ub_bool, dtype=value.dtype)
lb = paddle.cast(lb_bool, dtype=value.dtype)
ub = paddle.cast(ub_bool, dtype=value.dtype)
return paddle.subtract(
paddle.log(lb * ub), paddle.log(self.high - self.low), name=name
)
......@@ -245,8 +245,8 @@ class Uniform(distribution.Distribution):
name = self.name + '_probs'
lb_bool = self.low < value
ub_bool = value < self.high
lb = tensor.cast(lb_bool, dtype=value.dtype)
ub = tensor.cast(ub_bool, dtype=value.dtype)
lb = paddle.cast(lb_bool, dtype=value.dtype)
ub = paddle.cast(ub_bool, dtype=value.dtype)
return paddle.divide((lb * ub), (self.high - self.low), name=name)
def entropy(self):
......
......@@ -96,7 +96,7 @@ class DecoupledWeightDecay:
[param, grad]
), framework.name_scope('weight decay'):
updated_param = paddle.subtract(x=param, y=scaled_param)
paddle.fluid.layers.assign(input=updated_param, output=param)
paddle.assign(updated_param, output=param)
optimize_ops = self.apply_optimize(
loss=loss,
......
......@@ -295,9 +295,9 @@ class TestAmpWithNonIterableDataLoader(unittest.TestCase):
)
with fluid.layers.control_flow.Switch() as switch:
with switch.case(label != zero_var):
fluid.layers.assign(input=zero_var, output=label)
paddle.assign(zero_var, output=label)
with switch.default():
fluid.layers.assign(input=one_var, output=label)
paddle.assign(one_var, output=label)
net = resnet_cifar10(image)
logits = paddle.static.nn.fc(
......
......@@ -182,7 +182,7 @@ class TestWeightDecay(unittest.TestCase):
for params in param_list:
updated_p = paddle.subtract(x=params[0], y=params[1])
fluid.layers.assign(input=updated_p, output=params[0])
paddle.assign(updated_p, output=params[0])
optimizer.apply_optimize(avg_cost, startup_prog, params_grads)
......
......@@ -65,7 +65,7 @@ def _coalesce_tensors(var_groups):
flattened_vars.append(
paddle.reshape(x=g_var, shape=[np.prod(g_var.shape)])
)
coalesced_grad = nn.concat(flattened_vars)
coalesced_grad = paddle.concat(flattened_vars)
coalesced_grads_and_grad_vars.append(
[coalesced_grad, grad_vars, g_var_shapes]
)
......
......@@ -1703,7 +1703,7 @@ class Variable(metaclass=VariableMetaClass):
tmp = fluid.dygraph.base.to_variable(x)
tmp.stop_gradient=False
inputs2.append(tmp)
ret2 = fluid.layers.sums(inputs2)
ret2 = paddle.add_n(inputs2)
loss2 = paddle.sum(ret2)
loss2.backward()
print(loss2.gradient())
......@@ -1751,7 +1751,7 @@ class Variable(metaclass=VariableMetaClass):
tmp = fluid.dygraph.base.to_variable(x)
tmp.stop_gradient=False
inputs2.append(tmp)
ret2 = fluid.layers.sums(inputs2)
ret2 = paddle.add_n(inputs2)
loss2 = paddle.sum(ret2)
loss2.backward()
print(loss2.gradient())
......
......@@ -144,7 +144,7 @@ def model():
input=lr_embbding, pool_type="sum"
)
merge_layer = fluid.layers.concat(input=[dnn_out, lr_pool], axis=1)
merge_layer = paddle.concat([dnn_out, lr_pool], axis=1)
predict = paddle.static.nn.fc(x=merge_layer, size=2, activation='softmax')
acc = paddle.static.accuracy(input=predict, label=label)
......
......@@ -15,7 +15,7 @@
from ..wrapped_decorator import signature_safe_contextmanager
from .layer_function_generator import templatedoc
from .tensor import assign, cast, fill_constant
from .tensor import fill_constant
from .. import core
from ..framework import (
Program,
......@@ -1058,7 +1058,7 @@ def assign_skip_lod_tensor_array(input, output):
if isinstance(output, Variable) and isinstance(
input, support_ret_buildin_type
):
assign(input, output)
paddle.assign(input, output)
else:
output = input
return
......@@ -1069,7 +1069,7 @@ def assign_skip_lod_tensor_array(input, output):
main_program.current_block().parent_idx
)
if parent_block and not parent_block._find_var_recursive(input.name):
assign(input, output)
paddle.assign(input, output)
else:
if (
isinstance(output, Variable)
......@@ -1081,7 +1081,7 @@ def assign_skip_lod_tensor_array(input, output):
input.shape, output.shape
)
)
assign(input, output)
paddle.assign(input, output)
# (TODO: Mine) There exists dependency (jit.dy2static.convert_operators). It will be removed later.
......@@ -1195,7 +1195,7 @@ def while_loop(cond, body, loop_vars, is_test=False, name=None):
)
now_cond = cond(*output_vars)
map_structure(assign_skip_lod_tensor_array, output_vars, loop_vars)
assign(now_cond, pre_cond)
paddle.assign(now_cond, pre_cond)
return loop_vars
......
......@@ -55,7 +55,7 @@ def _decay_step_counter(begin=0):
global_step = nn.autoincreased_step_counter(
counter_name='@LR_DECAY_COUNTER@', begin=begin, step=1
)
global_step = tensor.cast(global_step, 'float32')
global_step = paddle.cast(global_step, 'float32')
return global_step
......@@ -361,7 +361,7 @@ def polynomial_decay(
with control_flow.Switch() as switch:
with switch.case(global_step == zero_var):
tensor.assign(input=one_var, output=div_res)
paddle.assign(one_var, output=div_res)
decay_steps = decay_steps * div_res
else:
decay_steps_var = tensor.fill_constant(
......@@ -595,11 +595,11 @@ def linear_lr_warmup(learning_rate, warmup_steps, start_lr, end_lr):
decayed_lr = start_lr + linear_step * (
global_step / float(warmup_steps)
)
tensor.assign(decayed_lr, lr)
paddle.assign(decayed_lr, lr)
with switch.default():
if not isinstance(learning_rate, Variable):
learning_rate = tensor.fill_constant(
shape=[1], dtype=dtype, value=float(learning_rate)
)
tensor.assign(learning_rate, lr)
paddle.assign(learning_rate, lr)
return lr
......@@ -41,7 +41,7 @@ from .layer_function_generator import (
templatedoc,
_generate_doc_string_,
)
from .tensor import concat, assign, fill_constant, zeros
from .tensor import fill_constant, zeros
from . import utils
from .. import unique_name
from .. import core
......
......@@ -12,6 +12,7 @@
# See the License for the specific language governing permissions and
# limitations under the License.
import paddle
import collections
import copy
import numpy as np
......@@ -387,7 +388,7 @@ def _contain_var(list_or_tuple):
def get_shape_tensor_inputs(inputs, attrs, shape, op_type):
from .tensor import fill_constant, cast
from .tensor import fill_constant
def _get_attr_shape(list_shape):
attr_shape = []
......@@ -411,7 +412,7 @@ def get_shape_tensor_inputs(inputs, attrs, shape, op_type):
'(When type of shape in' + op_type + 'is list or tuple.)',
)
if convert_dtype(dim.dtype) == 'int64':
dim = cast(x=dim, dtype='int32')
dim = paddle.cast(x=dim, dtype='int32')
shape_tensor_list.append(dim)
else:
temp_out = fill_constant([1], 'int32', dim, force_cpu=True)
......@@ -428,7 +429,7 @@ def get_shape_tensor_inputs(inputs, attrs, shape, op_type):
'(When type of shape in' + op_type + ' is Variable.)',
)
if convert_dtype(shape.dtype) == 'int64':
shape = cast(shape, 'int32')
shape = paddle.cast(shape, 'int32')
inputs["ShapeTensor"] = shape
elif isinstance(shape, (list, tuple)):
attrs["shape"] = _get_attr_shape(shape)
......
......@@ -3920,14 +3920,14 @@ class ModelAverage(Optimizer):
self._get_accumulator('num_updates', param)
)
# backup param value to grad
layers.assign(input=param, output=grad)
paddle.assign(param, output=grad)
# param = (sum_1 + sum_2 + sum_3) / (num_accumulates + old_num_accumulates)
tmp = paddle.add_n([num_accumulates, old_num_accumulates])
sum = paddle.add_n([sum_1, sum_2, sum_3])
tmp = layers.cast(
tmp = paddle.cast(
x=tmp, dtype='float32' if self._dtype is None else self._dtype
)
sum = layers.cast(
sum = paddle.cast(
x=sum, dtype='float32' if self._dtype is None else self._dtype
)
paddle.assign(paddle.divide(sum, tmp), output=param)
......@@ -3935,7 +3935,7 @@ class ModelAverage(Optimizer):
def _add_average_restore_op(self, block, param_grad):
param = block._clone_variable(param_grad[0])
grad = block._clone_variable(param_grad[1])
layers.assign(input=grad, output=param)
paddle.assign(grad, output=param)
def _append_average_accumulate_op(self, param):
self.helper = LayerHelper("average_accumulate")
......@@ -4229,15 +4229,13 @@ class ExponentialMovingAverage:
param = block._clone_variable(param)
tmp = block._clone_variable(tmp)
ema = block._clone_variable(self._ema_vars[param.name])
layers.assign(input=param, output=tmp)
paddle.assign(param, output=tmp)
# bias correction
with layers.control_flow.Switch() as switch:
with switch.case(global_step > 0):
layers.assign(
output=param, input=ema / (1.0 - decay_pow)
)
paddle.assign(ema / (1.0 - decay_pow), output=param)
with switch.default():
layers.assign(output=param, input=ema)
paddle.assign(ema, output=param)
self.restore_program = Program()
block = self.restore_program.global_block()
......@@ -4245,7 +4243,7 @@ class ExponentialMovingAverage:
for param, tmp in self._params_tmps:
tmp = block._clone_variable(tmp)
param = block._clone_variable(param)
layers.assign(input=tmp, output=param)
paddle.assign(tmp, output=param)
def _get_ema_decay(self):
with default_main_program()._lr_schedule_guard():
......@@ -4261,9 +4259,9 @@ class ExponentialMovingAverage:
decay_t = (self._thres_steps + 1.0) / (self._thres_steps + 10.0)
with layers.control_flow.Switch() as switch:
with switch.case(decay_t < self._decay):
layers.tensor.assign(decay_t, decay_var)
paddle.assign(decay_t, decay_var)
with switch.default():
layers.tensor.assign(
paddle.assign(
np.array([self._decay], dtype=np.float32), decay_var
)
return decay_var
......@@ -4276,7 +4274,7 @@ class ExponentialMovingAverage:
dtype='int64',
persistable=True,
)
global_step = layers.cast(global_step, "float32")
global_step = paddle.cast(global_step, "float32")
decay_var = block._clone_variable(self._decay_var)
decay_pow_acc = paddle.pow(decay_var, global_step)
return decay_pow_acc, global_step
......@@ -4313,7 +4311,7 @@ class ExponentialMovingAverage:
ema_t = param_ema * self._decay_var + param * (
1 - self._decay_var
)
layers.assign(input=ema_t, output=param_ema)
paddle.assign(ema_t, output=param_ema)
# for fp16 params
for param_ema, master_ema in param_master_emas:
......@@ -7272,7 +7270,7 @@ class LookaheadOptimizer:
for param_name in params:
fast_var = main_block.var(param_name)
slow_var = param_to_slow[param_name]
layers.assign(input=fast_var, output=slow_var)
paddle.assign(fast_var, output=slow_var)
with switch.case(mod == zero_var):
for param_name in params:
fast_var = main_block.var(param_name)
......@@ -7283,8 +7281,8 @@ class LookaheadOptimizer:
slow_var, paddle.subtract(one_var, alpha)
),
)
layers.assign(input=tmp_var, output=slow_var)
layers.assign(input=tmp_var, output=fast_var)
paddle.assign(tmp_var, output=slow_var)
paddle.assign(tmp_var, output=fast_var)
with switch.default():
pass
return mini_out
......
......@@ -91,8 +91,8 @@ def get_usr_combined_features():
usr_job_fc = paddle.static.nn.fc(x=usr_job_emb, size=16)
concat_embed = layers.concat(
input=[usr_fc, usr_gender_fc, usr_age_fc, usr_job_fc], axis=1
concat_embed = paddle.concat(
[usr_fc, usr_gender_fc, usr_age_fc, usr_job_fc], axis=1
)
usr_combined_features = paddle.static.nn.fc(
......@@ -150,8 +150,8 @@ def get_mov_combined_features():
pool_type="sum",
)
concat_embed = layers.concat(
input=[mov_fc, mov_categories_hidden, mov_title_conv], axis=1
concat_embed = paddle.concat(
[mov_fc, mov_categories_hidden, mov_title_conv], axis=1
)
# FIXME(dzh) : need tanh operator
......
......@@ -87,8 +87,8 @@ def train(
param_attr='shared_w',
)
concat_embed = fluid.layers.concat(
input=[embed_first, embed_second, embed_third, embed_forth], axis=1
concat_embed = paddle.concat(
[embed_first, embed_second, embed_third, embed_forth], axis=1
)
hidden1 = paddle.static.nn.fc(
x=concat_embed, size=HIDDEN_SIZE, activation='sigmoid'
......
......@@ -58,7 +58,7 @@ def make_program():
)
where_1 = paddle.where(y > 1, y, out1)
paddle.fluid.layers.assign(where_1, where_0)
paddle.assign(where_1, where_0)
return main_program, start_program
......
......@@ -53,7 +53,7 @@ def net():
zero = fluid.layers.fill_constant(shape=[1], dtype='int64', value=0)
# test float16 value
fp16_zero = fluid.layers.cast(zero, dtype='float16')
fp16_zero = paddle.cast(zero, dtype='float16')
y = y + zero
......
......@@ -35,19 +35,11 @@ class TestCollectiveSendRecv(TestCollectiveRunnerBase):
)
tindata.desc.set_need_check_feed(False)
if self.rank == 0:
data1 = fluid.layers.assign(
np.array([[0, 1, 2]], dtype='float32')
)
data2 = fluid.layers.assign(
np.array([[3, 4, 5]], dtype='float32')
)
data1 = paddle.assign(np.array([[0, 1, 2]], dtype='float32'))
data2 = paddle.assign(np.array([[3, 4, 5]], dtype='float32'))
elif self.rank == 1:
data1 = fluid.layers.assign(
np.array([[3, 4, 5]], dtype='float32')
)
data2 = fluid.layers.assign(
np.array([[0, 1, 2]], dtype='float32')
)
data1 = paddle.assign(np.array([[3, 4, 5]], dtype='float32'))
data2 = paddle.assign(np.array([[0, 1, 2]], dtype='float32'))
tensor_array = paddle.tensor.create_array(dtype='float32')
i = fluid.layers.fill_constant(shape=[1], dtype='int64', value=0)
paddle.tensor.array_write(data1, i, tensor_array)
......
......@@ -122,19 +122,19 @@ class TestHybridParallelInferenceHelperClass(unittest.TestCase):
# update cond and assign to cond_int, we will sync cond_int
paddle.assign(paddle.less_than(x=step_idx, y=max_len), cond)
layers.assign(layers.cast(cond, dtype="int32"), cond_int)
paddle.assign(paddle.cast(cond, dtype="int32"), cond_int)
with paddle.fluid.device_guard(f'{device}:all'):
# the code below must at end of while block and exists in device:all
layers.assign(layers.cast(cond_int, dtype='bool'), cond)
paddle.assign(paddle.cast(cond_int, dtype='bool'), cond)
with paddle.fluid.device_guard(f'{device}:all'):
out = paddle.tensor.create_array(data.dtype)
layers.assign(data, out)
paddle.assign(data, out)
with paddle.fluid.device_guard(f'{device}:all'):
# use a empty lod_tensor_array to clear lod_tensor_array
layers.assign(paddle.tensor.create_array(data.dtype), data)
paddle.assign(paddle.tensor.create_array(data.dtype), data)
helper = HybridParallelInferenceHelper(
startup_program,
......
......@@ -95,7 +95,7 @@ class TestDistCTR2x2(TestDistRunnerBase):
input=lr_embbding, pool_type="sum"
)
merge_layer = fluid.layers.concat(input=[dnn_out, lr_pool], axis=1)
merge_layer = paddle.concat([dnn_out, lr_pool], axis=1)
predict = paddle.static.nn.fc(
x=merge_layer, size=2, activation='softmax'
......
......@@ -144,7 +144,7 @@ class TestDistCTR2x2(FleetDistRunnerBase):
input=lr_embbding, pool_type="sum"
)
merge_layer = fluid.layers.concat(input=[dnn_out, lr_pool], axis=1)
merge_layer = paddle.concat([dnn_out, lr_pool], axis=1)
predict = paddle.static.nn.fc(
x=merge_layer, size=2, activation='softmax'
......
......@@ -116,8 +116,8 @@ class TestHeterPipelinePsCTR2x2(FleetDistHeterRunnerBase):
dnn_out = fc
with fluid.device_guard("cpu"):
merge_layer = fluid.layers.concat(input=[dnn_out, lr_pool], axis=1)
label = fluid.layers.cast(label, dtype="int64")
merge_layer = paddle.concat([dnn_out, lr_pool], axis=1)
label = paddle.cast(label, dtype="int64")
predict = paddle.static.nn.fc(
x=merge_layer, size=2, activation='softmax'
)
......
......@@ -55,7 +55,7 @@ def fake_simnet_reader():
def get_acc(cos_q_nt, cos_q_pt, batch_size):
cond = paddle.less_than(cos_q_nt, cos_q_pt)
cond = fluid.layers.cast(cond, dtype='float64')
cond = paddle.cast(cond, dtype='float64')
cond_3 = paddle.sum(cond)
acc = paddle.divide(
cond_3,
......
......@@ -134,7 +134,8 @@ class TestDistCTR2x2(FleetDistRunnerBase):
lr_pool = paddle.static.nn.sequence_lod.sequence_pool(
input=lr_embbding, pool_type="sum"
)
merge_layer = fluid.layers.concat(input=[dnn_out, lr_pool], axis=1)
merge_layer = paddle.concat([dnn_out, lr_pool], axis=1)
predict = paddle.static.nn.fc(
x=merge_layer, size=2, activation='softmax'
)
......
......@@ -1193,8 +1193,8 @@ def multi_head_attention(
q, k, v = __compute_qkv(queries, keys, values, n_head, d_key, d_value)
if cache is not None: # use cache and concat time steps
k = cache["k"] = layers.concat([cache["k"], k], axis=1)
v = cache["v"] = layers.concat([cache["v"], v], axis=1)
k = cache["k"] = paddle.concat([cache["k"], k], axis=1)
v = cache["v"] = paddle.concat([cache["v"], v], axis=1)
q = __split_heads(q, n_head)
k = __split_heads(k, n_head)
......@@ -1858,11 +1858,11 @@ def fast_decode(
# update states
layers.array_write(selected_ids, i=step_idx, array=ids)
layers.array_write(selected_scores, i=step_idx, array=scores)
layers.assign(pre_src_attn_bias, trg_src_attn_bias)
layers.assign(pre_enc_output, enc_output)
paddle.assign(pre_src_attn_bias, trg_src_attn_bias)
paddle.assign(pre_enc_output, enc_output)
for i in range(n_layer):
layers.assign(pre_caches[i]["k"], caches[i]["k"])
layers.assign(pre_caches[i]["v"], caches[i]["v"])
paddle.assign(pre_caches[i]["k"], caches[i]["k"])
paddle.assign(pre_caches[i]["v"], caches[i]["v"])
length_cond = paddle.less_than(x=step_idx, y=max_len)
finish_cond = paddle.logical_not(layers.is_empty(x=selected_ids))
paddle.logical_and(x=length_cond, y=finish_cond, out=cond)
......
......@@ -75,8 +75,8 @@ class TestDistWord2vec2x2(TestDistRunnerBase):
),
)
concat_embed = fluid.layers.concat(
input=[embed_first, embed_second, embed_third, embed_forth],
concat_embed = paddle.concat(
[embed_first, embed_second, embed_third, embed_forth],
axis=1,
)
hidden1 = paddle.static.nn.fc(
......
......@@ -384,7 +384,7 @@ class PretrainModelLayer(Layer):
mask_pos,
labels,
):
mask_pos = fluid.layers.cast(x=mask_pos, dtype='int32')
mask_pos = paddle.cast(x=mask_pos, dtype='int32')
enc_output, next_sent_feat = self.bert_layer(
src_ids, position_ids, sentence_ids, input_mask
......
......@@ -18,7 +18,7 @@ from seq2seq_utils import Seq2SeqModelHyperParams as args
import paddle
import paddle.fluid as fluid
from paddle.fluid import ParamAttr, layers
from paddle.fluid import ParamAttr
from paddle.fluid.dygraph import Layer
from paddle.fluid.dygraph.base import to_variable
from paddle.jit.api import to_static
......@@ -67,7 +67,7 @@ class BasicLSTMUnit(Layer):
)
def forward(self, input, pre_hidden, pre_cell):
concat_input_hidden = layers.concat([input, pre_hidden], 1)
concat_input_hidden = paddle.concat([input, pre_hidden], 1)
gate_input = paddle.matmul(x=concat_input_hidden, y=self._weight)
gate_input = paddle.add(gate_input, self._bias)
......@@ -488,12 +488,12 @@ class BaseModel(fluid.dygraph.Layer):
self._gather(x, beam_indices, batch_pos) for x in new_dec_cell
]
next_finished = self._gather(beam_finished, beam_indices, batch_pos)
next_finished = fluid.layers.cast(next_finished, "bool")
next_finished = paddle.cast(next_finished, "bool")
next_finished = paddle.logical_or(
next_finished,
paddle.equal(token_indices, end_token_tensor),
)
next_finished = fluid.layers.cast(next_finished, "float32")
next_finished = paddle.cast(next_finished, "float32")
dec_hidden, dec_cell = new_dec_hidden, new_dec_cell
beam_finished = next_finished
......@@ -808,7 +808,7 @@ class AttentionModel(fluid.dygraph.Layer):
for step_idx in range(max_seq_len):
j = step_idx + 0
step_input = tar_emb[j]
step_input = fluid.layers.concat([step_input, input_feed], 1)
step_input = paddle.concat([step_input, input_feed], 1)
new_dec_hidden, new_dec_cell = [], []
for i in range(self.num_layers):
new_hidden, new_cell = self.dec_units[i](
......@@ -826,7 +826,7 @@ class AttentionModel(fluid.dygraph.Layer):
step_input = new_hidden
dec_att = self.attention(step_input, enc_outputs, enc_padding_mask)
dec_att = paddle.squeeze(dec_att, [1])
concat_att_out = fluid.layers.concat([dec_att, step_input], 1)
concat_att_out = paddle.concat([dec_att, step_input], 1)
out = self.concat_fc(concat_att_out)
input_feed = out
dec_output.append(out)
......
......@@ -97,7 +97,7 @@ class ConcatLayer:
"""
operation
"""
concat = fluid.layers.concat(inputs, axis=self.axis)
concat = paddle.concat(inputs, axis=self.axis)
return concat
......
......@@ -68,7 +68,7 @@ class TestAST2Func(unittest.TestCase):
x_data = np.random.random([10, 16]).astype('float32')
main_program = fluid.Program()
with fluid.program_guard(main_program):
x_v = fluid.layers.assign(x_data)
x_v = paddle.assign(x_data)
true_ret = func(x_v)
test_ret = self._ast2func(func)(x_v)
exe = fluid.Executor(fluid.CPUPlace())
......
......@@ -252,7 +252,7 @@ class TestDygraphBasicApi(unittest.TestCase):
main_program = fluid.Program()
main_program.random_seed = SEED
with fluid.program_guard(main_program, startup_program):
data = fluid.layers.assign(self.input)
data = paddle.assign(self.input)
static_out = dygraph_to_static_func(self.dygraph_func)(data)
exe = fluid.Executor(fluid.CPUPlace())
......
......@@ -310,7 +310,7 @@ def bmn_loss_func(
self_bm_mask = paddle.static.create_global_var(
shape=[dscale, tscale], value=0, dtype=DATATYPE, persistable=True
)
fluid.layers.assign(bm_mask, self_bm_mask)
paddle.assign(bm_mask, self_bm_mask)
self_bm_mask.stop_gradient = True
return self_bm_mask
......@@ -319,9 +319,9 @@ def bmn_loss_func(
pred_score = paddle.reshape(x=pred_score, shape=[-1])
gt_label = paddle.reshape(x=gt_label, shape=[-1])
gt_label.stop_gradient = True
pmask = fluid.layers.cast(x=(gt_label > 0.5), dtype=DATATYPE)
num_entries = fluid.layers.cast(paddle.shape(pmask), dtype=DATATYPE)
num_positive = fluid.layers.cast(paddle.sum(pmask), dtype=DATATYPE)
pmask = paddle.cast(x=(gt_label > 0.5), dtype=DATATYPE)
num_entries = paddle.cast(paddle.shape(pmask), dtype=DATATYPE)
num_positive = paddle.cast(paddle.sum(pmask), dtype=DATATYPE)
ratio = num_entries / num_positive
coef_0 = 0.5 * ratio / (ratio - 1)
coef_1 = 0.5 * ratio
......@@ -345,34 +345,34 @@ def bmn_loss_func(
gt_iou_map = paddle.multiply(gt_iou_map, mask)
u_hmask = fluid.layers.cast(x=gt_iou_map > 0.7, dtype=DATATYPE)
u_hmask = paddle.cast(x=gt_iou_map > 0.7, dtype=DATATYPE)
u_mmask = paddle.logical_and(gt_iou_map <= 0.7, gt_iou_map > 0.3)
u_mmask = fluid.layers.cast(x=u_mmask, dtype=DATATYPE)
u_mmask = paddle.cast(x=u_mmask, dtype=DATATYPE)
u_lmask = paddle.logical_and(gt_iou_map <= 0.3, gt_iou_map >= 0.0)
u_lmask = fluid.layers.cast(x=u_lmask, dtype=DATATYPE)
u_lmask = paddle.cast(x=u_lmask, dtype=DATATYPE)
u_lmask = paddle.multiply(u_lmask, mask)
num_h = fluid.layers.cast(paddle.sum(u_hmask), dtype=DATATYPE)
num_m = fluid.layers.cast(paddle.sum(u_mmask), dtype=DATATYPE)
num_l = fluid.layers.cast(paddle.sum(u_lmask), dtype=DATATYPE)
num_h = paddle.cast(paddle.sum(u_hmask), dtype=DATATYPE)
num_m = paddle.cast(paddle.sum(u_mmask), dtype=DATATYPE)
num_l = paddle.cast(paddle.sum(u_lmask), dtype=DATATYPE)
r_m = num_h / num_m
u_smmask = fluid.layers.assign(
u_smmask = paddle.assign(
local_random.uniform(
0.0, 1.0, [gt_iou_map.shape[1], gt_iou_map.shape[2]]
).astype(DATATYPE)
)
u_smmask = paddle.multiply(u_mmask, u_smmask)
u_smmask = fluid.layers.cast(x=(u_smmask > (1.0 - r_m)), dtype=DATATYPE)
u_smmask = paddle.cast(x=(u_smmask > (1.0 - r_m)), dtype=DATATYPE)
r_l = num_h / num_l
u_slmask = fluid.layers.assign(
u_slmask = paddle.assign(
local_random.uniform(
0.0, 1.0, [gt_iou_map.shape[1], gt_iou_map.shape[2]]
).astype(DATATYPE)
)
u_slmask = paddle.multiply(u_lmask, u_slmask)
u_slmask = fluid.layers.cast(x=(u_slmask > (1.0 - r_l)), dtype=DATATYPE)
u_slmask = paddle.cast(x=(u_slmask > (1.0 - r_l)), dtype=DATATYPE)
weights = u_hmask + u_smmask + u_slmask
weights.stop_gradient = True
......@@ -385,8 +385,8 @@ def bmn_loss_func(
def pem_cls_loss_func(pred_score, gt_iou_map, mask):
gt_iou_map = paddle.multiply(gt_iou_map, mask)
gt_iou_map.stop_gradient = True
pmask = fluid.layers.cast(x=(gt_iou_map > 0.9), dtype=DATATYPE)
nmask = fluid.layers.cast(x=(gt_iou_map <= 0.9), dtype=DATATYPE)
pmask = paddle.cast(x=(gt_iou_map > 0.9), dtype=DATATYPE)
nmask = paddle.cast(x=(gt_iou_map <= 0.9), dtype=DATATYPE)
nmask = paddle.multiply(nmask, mask)
num_positive = paddle.sum(pmask)
......
......@@ -141,7 +141,7 @@ class MyConvLayer(fluid.dygraph.Layer):
@paddle.jit.to_static
def dymethod(self, x_v):
x_v = fluid.layers.assign(x_v)
x_v = paddle.assign(x_v)
return x_v
......
......@@ -86,7 +86,7 @@ class DynamicGRU(fluid.dygraph.Layer):
if self.is_reverse:
res = res[::-1]
res = fluid.layers.concat(res, axis=1)
res = paddle.concat(res, axis=1)
return res
......@@ -154,7 +154,7 @@ class BiGRU(fluid.dygraph.Layer):
res_pre_gru_r = self.pre_gru_r(input_feature)
res_gru_r = self.gru_r(res_pre_gru_r)
bi_merge = fluid.layers.concat(input=[res_gru, res_gru_r], axis=-1)
bi_merge = paddle.concat([res_gru, res_gru_r], axis=-1)
return bi_merge
......
......@@ -94,7 +94,7 @@ def test_list_append_in_for_loop_with_concat(x, iter_num):
) # TODO(liym27): Delete it if the type of parameter iter_num can be resolved
for i in range(iter_num):
a.append(x)
a = fluid.layers.concat(a, axis=0)
a = paddle.concat(a, axis=0)
return a
......
......@@ -89,7 +89,7 @@ class SimpleLSTMRNN(fluid.Layer):
weight_1 = self.weight_1_arr[k]
bias = self.bias_arr[k]
nn = fluid.layers.concat([step_input, pre_hidden], 1)
nn = paddle.concat([step_input, pre_hidden], 1)
gate_input = paddle.matmul(x=nn, y=weight_1)
gate_input = paddle.add(gate_input, bias)
......@@ -111,16 +111,16 @@ class SimpleLSTMRNN(fluid.Layer):
mode='upscale_in_train',
)
res.append(step_input)
real_res = fluid.layers.concat(res, 1)
real_res = paddle.concat(res, 1)
real_res = paddle.reshape(
real_res, [-1, self._num_steps, self._hidden_size]
)
last_hidden = fluid.layers.concat(hidden_array, 1)
last_hidden = paddle.concat(hidden_array, 1)
last_hidden = paddle.reshape(
last_hidden, shape=[-1, self._num_layers, self._hidden_size]
)
last_hidden = paddle.transpose(x=last_hidden, perm=[1, 0, 2])
last_cell = fluid.layers.concat(cell_array, 1)
last_cell = paddle.concat(cell_array, 1)
last_cell = paddle.reshape(
last_cell, shape=[-1, self._num_layers, self._hidden_size]
)
......
......@@ -152,7 +152,7 @@ def train(args, place, to_static):
cur_loss = paddle.multiply(_R, log_prob)
policy_loss.append(cur_loss)
policy_loss = fluid.layers.concat(policy_loss)
policy_loss = paddle.concat(policy_loss)
policy_loss = paddle.sum(policy_loss)
policy_loss.backward()
......
......@@ -248,8 +248,8 @@ class BiGRU(fluid.dygraph.Layer):
gru_backward = self._gru_backward(fc_1)
gru_forward_tanh = paddle.tanh(gru_forward)
gru_backward_tanh = paddle.tanh(gru_backward)
encoded_vector = fluid.layers.concat(
input=[gru_forward_tanh, gru_backward_tanh], axis=2
encoded_vector = paddle.concat(
[gru_forward_tanh, gru_backward_tanh], axis=2
)
encoded_vector = paddle.max(encoded_vector, axis=1)
fc_2 = self._fc2(encoded_vector)
......
......@@ -146,8 +146,8 @@ class MultiHeadAttention(Layer):
if cache is not None:
cache_k, cache_v = cache["k"], cache["v"]
k = layers.concat([cache_k, k], axis=2)
v = layers.concat([cache_v, v], axis=2)
k = paddle.concat([cache_k, k], axis=2)
v = paddle.concat([cache_v, v], axis=2)
cache["k"], cache["v"] = k, v
# scale dot product attention
product = paddle.matmul(x=q, y=k, transpose_y=True)
......@@ -774,7 +774,7 @@ class Transformer(Layer):
return res
def mask_probs(probs, finished, noend_mask_tensor):
finished = layers.cast(finished, dtype=probs.dtype)
finished = paddle.cast(finished, dtype=probs.dtype)
probs = paddle.multiply(
paddle.expand(
paddle.unsqueeze(finished, [2]),
......
......@@ -207,7 +207,7 @@ class Upsample(fluid.dygraph.Layer):
shape_nchw = paddle.shape(inputs)
shape_hw = paddle.slice(shape_nchw, axes=[0], starts=[2], ends=[4])
shape_hw.stop_gradient = True
in_shape = fluid.layers.cast(shape_hw, dtype='int32')
in_shape = paddle.cast(shape_hw, dtype='int32')
out_shape = in_shape * self.scale
out_shape.stop_gradient = True
......@@ -295,9 +295,7 @@ class YOLOv3(fluid.dygraph.Layer):
blocks = self.block(inputs)
for i, block in enumerate(blocks):
if i > 0:
block = fluid.layers.concat(
input=[route, block], axis=1 # noqa: F821
)
block = paddle.concat([route, block], axis=1) # noqa: F821
route, tip = self.yolo_blocks[i](block)
block_out = self.block_outputs[i](tip)
self.outputs.append(block_out)
......@@ -349,8 +347,8 @@ class YOLOv3(fluid.dygraph.Layer):
if not self.is_train:
# get pred
yolo_boxes = fluid.layers.concat(self.boxes, axis=1)
yolo_scores = fluid.layers.concat(self.scores, axis=2)
yolo_boxes = paddle.concat(self.boxes, axis=1)
yolo_scores = paddle.concat(self.scores, axis=2)
pred = _legacy_C_ops.multiclass_nms(
bboxes=yolo_boxes,
......
......@@ -107,8 +107,8 @@ def net(batch_size=4, lr=0.01):
)
dnn_out = fc
merge_layer = fluid.layers.concat(input=[dnn_out, lr_pool], axis=1)
label = fluid.layers.cast(label, dtype="int64")
merge_layer = paddle.concat([dnn_out, lr_pool], axis=1)
label = paddle.cast(label, dtype="int64")
predict = paddle.static.nn.fc(
x=merge_layer, size=2, activation='softmax'
)
......
......@@ -24,10 +24,10 @@ from paddle.fluid.incubate.fleet.parameter_server.distribute_transpiler import (
input_x = paddle.static.data(name="x", shape=[-1, 32], dtype='float32')
input_y = paddle.static.data(name="y", shape=[-1, 1], dtype='int64')
input_y = fluid.layers.cast(input_y, dtype="float32")
input_y = paddle.cast(input_y, dtype="float32")
with fluid.device_guard("gpu"):
input_y = fluid.layers.cast(input_y, dtype="int64")
input_y = paddle.cast(input_y, dtype="int64")
cost = mlp(input_x, input_y)
optimizer = fluid.optimizer.Adagrad(learning_rate=0.01)
......
......@@ -47,7 +47,7 @@ class TestBase(IPUOpTest):
x = paddle.static.data(
name=self.feed_list[0], shape=self.feed_shape[0], dtype='float32'
)
out = paddle.fluid.layers.argmax(x, **self.attrs)
out = paddle.argmax(x, **self.attrs)
self.fetch_list = [out.name]
def run_model(self, exec_mode):
......
......@@ -47,7 +47,7 @@ class TestBase(IPUOpTest):
x = paddle.static.data(
name=self.feed_list[0], shape=self.feed_shape[0], dtype='float32'
)
out = paddle.fluid.layers.argmin(x, **self.attrs)
out = paddle.argmin(x, **self.attrs)
self.fetch_list = [out.name]
def run_model(self, exec_mode):
......
......@@ -56,7 +56,7 @@ class TestBase(IPUOpTest):
y = paddle.static.data(
name=self.feed_list[1], shape=self.feed_shape[1], dtype='float32'
)
out = paddle.fluid.layers.concat([x, y], **self.attrs)
out = paddle.concat([x, y], **self.attrs)
self.fetch_list = [out.name]
def run_model(self, exec_mode):
......
......@@ -115,7 +115,7 @@ class SlicePluginTRTTestInt32(SlicePluginTRTTest):
starts = self.params_starts
ends = self.params_ends
slice_out = paddle.slice(data, axes=axes, starts=starts, ends=ends)
cast_out = fluid.layers.cast(slice_out, 'float32')
cast_out = paddle.cast(slice_out, 'float32')
out = nn.batch_norm(cast_out, is_test=True)
self.feeds = {
......@@ -140,7 +140,7 @@ class StaticSlicePluginTRTTestInt32(SlicePluginTRTTest):
starts = self.params_starts
ends = self.params_ends
slice_out = paddle.slice(data, axes=axes, starts=starts, ends=ends)
cast_out = fluid.layers.cast(slice_out, 'float32')
cast_out = paddle.cast(slice_out, 'float32')
out = nn.batch_norm(cast_out, is_test=True)
self.feeds = {
......
......@@ -61,7 +61,7 @@ class TensorRTSubgraphPassConcatTest(InferencePassTest):
data2 = fluid.data(
name="data2", shape=[-1, 3, 64, 64], dtype="float32"
)
concat_out = fluid.layers.concat([data1, data2], axis=2)
concat_out = paddle.concat([data1, data2], axis=2)
out = nn.batch_norm(concat_out, is_test=True)
self.feeds = {
"data1": np.random.random([1, 3, 64, 64]).astype("float32"),
......
......@@ -38,7 +38,7 @@ class TransposeFlattenConcatFusePassTRTTest(InferencePassTest):
flatt1 = paddle.flatten(trans1, 1, -1)
flatt2 = paddle.flatten(trans2, 1, -1)
concat_out = fluid.layers.concat([flatt1, flatt2], axis=1)
concat_out = paddle.concat([flatt1, flatt2], axis=1)
# There is no parameters for above structure.
# Hence, append a batch_norm to avoid failure caused by load_combined.
reshape_out = paddle.reshape(concat_out, [-1, 0, 1, 1])
......
......@@ -113,7 +113,7 @@ class FusionGroupPassInplaceTest(FusionGroupPassTest):
# subgraph with 3 op node
tmp_0 = self.feed_vars[0] - self.feed_vars[1]
tmp_1 = tmp_0 * self.feed_vars[2]
tmp_2 = layers.assign(tmp_1, output=tmp_0)
tmp_2 = paddle.assign(tmp_1, output=tmp_0)
tmp_3 = paddle.matmul(tmp_2, self.feed_vars[3])
self.num_fused_ops = 1
......@@ -138,17 +138,17 @@ class FusionGroupPassTestCastAndFP16(FusionGroupPassTest):
# subgraph with 2 op nodes
tmp_0 = self.feed_vars[0] * self.feed_vars[1]
tmp_1 = layers.cast(tmp_0, dtype="float16")
tmp_1 = paddle.cast(tmp_0, dtype="float16")
zero = layers.fill_constant(shape=[128], dtype="float16", value=0)
# TODO(xreki): fix precision problem when using softmax of float16.
# tmp_2 = layers.softmax(tmp_1)
tmp_2 = paddle.add(tmp_1, zero)
tmp_3 = paddle.matmul(tmp_0, self.feed_vars[2])
# subgraph with 4 op nodes
tmp_3 = layers.cast(tmp_2, dtype="float16")
tmp_3 = paddle.cast(tmp_2, dtype="float16")
tmp_4 = paddle.nn.functional.relu(tmp_1 + tmp_3)
tmp_5 = layers.cast(tmp_4, dtype=dtype)
tmp_3 = layers.cast(tmp_2, dtype=dtype)
tmp_5 = paddle.cast(tmp_4, dtype=dtype)
tmp_3 = paddle.cast(tmp_2, dtype=dtype)
self.append_gradients(tmp_5)
......@@ -185,8 +185,8 @@ class FusionGroupPassCastTest(FusionGroupPassTest):
self.feed_vars = self._prepare_feed_vars([2, 2], dtype, 2)
tmp_0 = paddle.add(self.feed_vars[0], self.feed_vars[1])
tmp_1 = layers.cast(tmp_0, dtype="float64")
tmp_2 = layers.cast(tmp_1, dtype="float32")
tmp_1 = paddle.cast(tmp_0, dtype="float64")
tmp_2 = paddle.cast(tmp_1, dtype="float32")
self.append_gradients(tmp_2)
......
......@@ -84,7 +84,7 @@ class TestSyncBatchNormOpTraining(TestSyncBatchNormRunnerBase):
use_cudnn=use_cudnn,
)
if self.bn_dtype == np.float16:
conv = fluid.layers.cast(conv, 'float16')
conv = paddle.cast(conv, 'float16')
bn = paddle.static.nn.batch_norm(
conv,
param_attr=fluid.ParamAttr(name='bn_scale'),
......@@ -95,7 +95,7 @@ class TestSyncBatchNormOpTraining(TestSyncBatchNormRunnerBase):
is_test=only_forward,
)
if self.bn_dtype == np.float16:
bn = fluid.layers.cast(bn, 'float32')
bn = paddle.cast(bn, 'float32')
sigmoid = paddle.nn.functional.sigmoid(bn)
out = paddle.sum(sigmoid)
# if not sync_bn:
......
......@@ -139,7 +139,7 @@ class TestCastOpError(unittest.TestCase):
x1 = fluid.create_lod_tensor(
np.array([[-1]]), [[1]], fluid.MLUPlace(0)
)
self.assertRaises(TypeError, fluid.layers.cast, x1, 'int32')
self.assertRaises(TypeError, paddle.cast, x1, 'int32')
if __name__ == '__main__':
......
......@@ -187,7 +187,7 @@ class TestOneHotOpApi(unittest.TestCase):
self._run(depth)
def test_api_with_depthTensor(self):
depth = fluid.layers.assign(input=np.array([10], dtype=np.int32))
depth = paddle.assign(np.array([10], dtype=np.int32))
self._run(depth)
def test_api_with_dygraph(self):
......
......@@ -344,7 +344,7 @@ class TestWhereDygraphAPI(unittest.TestCase):
y = paddle.where(x)
self.assertEqual(type(y), tuple)
self.assertEqual(len(y), 2)
z = fluid.layers.concat(list(y), axis=1)
z = paddle.concat(list(y), axis=1)
exe = fluid.Executor(paddle.device.MLUPlace(0))
(res,) = exe.run(
feed={'x': data}, fetch_list=[z.name], return_numpy=False
......@@ -357,7 +357,7 @@ class TestWhereDygraphAPI(unittest.TestCase):
y = paddle.where(x)
self.assertEqual(type(y), tuple)
self.assertEqual(len(y), 1)
z = fluid.layers.concat(list(y), axis=1)
z = paddle.concat(list(y), axis=1)
exe = fluid.Executor(paddle.device.MLUPlace(0))
(res,) = exe.run(
feed={'x': data}, fetch_list=[z.name], return_numpy=False
......
......@@ -94,7 +94,7 @@ class TestAssignApi(unittest.TestCase):
main_program = fluid.Program()
with fluid.program_guard(main_program):
x = paddle.tensor.create_tensor(dtype=self.dtype)
layers.assign(input=self.value, output=x)
paddle.assign(self.value, output=x)
exe = fluid.Executor(self.place)
[fetched_x] = exe.run(main_program, feed={}, fetch_list=[x])
......
......@@ -169,7 +169,7 @@ class TestConcatAPIWithLoDTensorArray(unittest.TestCase):
if use_fluid_api:
self.program = fluid.Program()
with fluid.program_guard(self.program):
input = fluid.layers.assign(self.x)
input = paddle.assign(self.x)
tensor_array = paddle.tensor.create_array(dtype='float32')
zero = fluid.layers.fill_constant(
shape=[1], value=0, dtype="int64"
......@@ -178,7 +178,7 @@ class TestConcatAPIWithLoDTensorArray(unittest.TestCase):
for i in range(self.iter_num):
paddle.tensor.array_write(input, zero + i, tensor_array)
self.out_var = fluid.layers.concat(tensor_array, axis=self.axis)
self.out_var = paddle.concat(tensor_array, axis=self.axis)
else:
self.program = paddle.static.Program()
with paddle.static.program_guard(self.program):
......
......@@ -210,7 +210,7 @@ class TestOneHotOpApi(unittest.TestCase):
self._run(depth)
def test_api_with_depthTensor(self):
depth = fluid.layers.assign(input=np.array([10], dtype=np.int32))
depth = paddle.assign(np.array([10], dtype=np.int32))
self._run(depth)
def test_api_with_dygraph(self):
......
......@@ -137,7 +137,7 @@ class TestStackAPIWithLoDTensorArray(unittest.TestCase):
def set_program(self):
self.program = fluid.Program()
with fluid.program_guard(self.program):
input = fluid.layers.assign(self.x)
input = paddle.assign(self.x)
tensor_array = paddle.tensor.create_array(dtype='float32')
zero = fluid.layers.fill_constant(shape=[1], value=0, dtype="int64")
......@@ -175,7 +175,7 @@ class TestTensorStackAPIWithLoDTensorArray(unittest.TestCase):
def set_program(self):
self.program = fluid.Program()
with fluid.program_guard(self.program):
input = fluid.layers.assign(self.x)
input = paddle.assign(self.x)
tensor_array = paddle.tensor.create_array(dtype='float32')
zero = fluid.layers.fill_constant(shape=[1], value=0, dtype="int64")
......
......@@ -38,7 +38,7 @@ class TestWhileOp(unittest.TestCase):
)
# fill_constant npu op doesn't support int64
i = layers.zeros(shape=[1], dtype='int32')
i = layers.cast(i, 'int64')
i = paddle.cast(i, 'int64')
i.stop_gradient = True
init = layers.zeros(shape=[10], dtype='float32')
mem_array = paddle.tensor.array_write(x=init, i=i)
......@@ -48,28 +48,28 @@ class TestWhileOp(unittest.TestCase):
i = paddle.increment(i)
paddle.tensor.array_write(d2, i, array=data_array)
i = layers.zeros(shape=[1], dtype='int32')
i = layers.cast(i, 'int64')
i = paddle.cast(i, 'int64')
i.stop_gradient = True
array_len = layers.fill_constant(shape=[1], dtype='int32', value=5)
array_len = layers.cast(array_len, 'int64')
array_len = paddle.cast(array_len, 'int64')
array_len.stop_gradient = True
cond = paddle.ones(shape=[1], dtype='int32')
cond = layers.cast(cond, 'bool')
cond = paddle.cast(cond, 'bool')
j = layers.fill_constant(shape=[1], dtype='int32', value=1)
j = layers.cast(j, 'int64')
j = paddle.cast(j, 'int64')
j.stop_gradient = True
array_len2 = layers.fill_constant(shape=[1], dtype='int32', value=3)
array_len2 = layers.cast(array_len2, 'int64')
array_len2 = paddle.cast(array_len2, 'int64')
array_len2.stop_gradient = True
cond2 = paddle.logical_or(x=j, y=array_len2)
cond2 = paddle.ones(shape=[1], dtype='int32')
cond2 = layers.cast(cond2, 'bool')
cond2 = paddle.cast(cond2, 'bool')
while_op = paddle.static.nn.control_flow.While(cond=cond)
while_op2 = paddle.static.nn.control_flow.While(cond=cond2)
with while_op.block():
d = paddle.tensor.array_read(array=data_array, i=i)
prev = paddle.tensor.array_read(array=mem_array, i=i)
result = layers.sums(input=[d, prev])
result = paddle.add_n([d, prev])
i = paddle.increment(x=i)
paddle.tensor.array_write(result, i=i, array=mem_array)
......@@ -78,7 +78,7 @@ class TestWhileOp(unittest.TestCase):
with while_op2.block():
d2 = paddle.tensor.array_read(array=data_array, i=j)
prev2 = paddle.tensor.array_read(array=mem_array, i=j)
result2 = layers.sums(input=[d2, prev2])
result2 = paddle.add_n([d2, prev2])
j = paddle.increment(x=j)
paddle.tensor.array_write(result2, i=j, array=mem_array)
......
......@@ -17,11 +17,10 @@ import unittest
import numpy as np
import paddle
sys.path.append("../")
from op_test import OpTest
import paddle
import paddle.fluid as fluid
import paddle.fluid.core as core
......@@ -156,9 +155,8 @@ class TestSequencePadOpError(unittest.TestCase):
def test_x_variable():
# the input x type must be Variable
x = np.random.random((2, 4)).astype("float32")
pad_value = fluid.layers.assign(
input=np.array([0.0], dtype=np.float32)
)
pad_value = paddle.assign(np.array([0.0], dtype=np.float32))
paddle.static.nn.sequence_lod.sequence_pad(x=x, pad_value=pad_value)
self.assertRaises(TypeError, test_x_variable)
......@@ -178,9 +176,8 @@ class TestSequencePadOpError(unittest.TestCase):
x2 = paddle.static.data(
name='x2', shape=[-1, 10, 5], dtype='int16', lod_level=1
)
pad_value2 = fluid.layers.assign(
input=np.array([0.0], dtype=np.int32)
)
pad_value2 = paddle.assign(np.array([0.0], dtype=np.int32))
paddle.static.nn.sequence_lod.sequence_pad(
x=x2, pad_value=pad_value2
)
......@@ -189,7 +186,8 @@ class TestSequencePadOpError(unittest.TestCase):
def test_length_dtype(self):
x = fluid.data(name='x', shape=[10, 5], dtype='float32', lod_level=1)
pad_value = fluid.layers.assign(input=np.array([0.0], dtype=np.float32))
pad_value = paddle.assign(np.array([0.0], dtype=np.float32))
out, length = paddle.static.nn.sequence_lod.sequence_pad(
x=x, pad_value=pad_value
)
......
......@@ -35,8 +35,8 @@ class TestMultiplyWrite(TestCompatibility):
inp1 = paddle.full((1,), 2)
inp2 = paddle.full((1,), 3)
paddle.fluid.layers.assign(inp1, out)
paddle.fluid.layers.assign(inp2, out)
paddle.assign(inp1, out)
paddle.assign(inp2, out)
return main_program, startup_program, out
def setUp(self):
......
......@@ -47,13 +47,13 @@ def _test_read_write(x):
mean_a1 = paddle.mean(a1)
mean_a2 = paddle.mean(a2)
a_sum = layers.sums(input=[mean_a0, mean_a1, mean_a2])
a_sum = paddle.add_n([mean_a0, mean_a1, mean_a2])
mean_x0 = paddle.mean(x[0])
mean_x1 = paddle.mean(x[1])
mean_x2 = paddle.mean(x[2])
x_sum = layers.sums(input=[mean_x0, mean_x1, mean_x2])
x_sum = paddle.add_n([mean_x0, mean_x1, mean_x2])
return a_sum, x_sum
......@@ -81,7 +81,7 @@ class TestArrayReadWrite(unittest.TestCase):
)
self.assertEqual(outs[0], outs[1])
total_sum = layers.sums(input=[a_sum, x_sum])
total_sum = paddle.add_n([a_sum, x_sum])
total_sum_scaled = paddle.scale(x=total_sum, scale=1 / 6.0)
append_backward(total_sum_scaled)
......@@ -116,9 +116,7 @@ class TestArrayReadWrite(unittest.TestCase):
a_sum_dygraph, x_sum_dygraph = _test_read_write(x_dygraph)
self.assertEqual(a_sum_dygraph, x_sum_dygraph)
total_sum_dygraph = layers.sums(
input=[a_sum_dygraph, x_sum_dygraph]
)
total_sum_dygraph = paddle.add_n([a_sum_dygraph, x_sum_dygraph])
total_sum_scaled_dygraph = paddle.scale(
x=total_sum_dygraph, scale=1 / 6.0
)
......
......@@ -78,7 +78,7 @@ class TestAssignOpWithLoDTensorArray(unittest.TestCase):
z = paddle.add(x=x, y=y)
i = fluid.layers.fill_constant(shape=[1], dtype='int64', value=0)
init_array = paddle.tensor.array_write(x=z, i=i)
array = fluid.layers.assign(init_array)
array = paddle.assign(init_array)
sums = paddle.tensor.array_read(array=init_array, i=i)
mean = paddle.mean(sums)
append_backward(mean)
......@@ -110,10 +110,10 @@ class TestAssignOpError(unittest.TestCase):
x1 = fluid.create_lod_tensor(
np.array([[-1]]), [[1]], fluid.CPUPlace()
)
self.assertRaises(TypeError, fluid.layers.assign, x1)
self.assertRaises(TypeError, paddle.assign, x1)
# When the type of input is numpy.ndarray, the dtype of input must be float32, int32.
x2 = np.array([[2.5, 2.5]], dtype='uint8')
self.assertRaises(TypeError, fluid.layers.assign, x2)
self.assertRaises(TypeError, paddle.assign, x2)
paddle.disable_static()
......@@ -252,7 +252,7 @@ class TestAssignOpErrorApi(unittest.TestCase):
class TestAssignDoubleGradCheck(unittest.TestCase):
def assign_wrapper(self, x):
return paddle.fluid.layers.assign(x[0])
return paddle.assign(x[0])
@prog_scope()
def func(self, place):
......@@ -262,7 +262,7 @@ class TestAssignDoubleGradCheck(unittest.TestCase):
data = paddle.static.data('data', [3, 4, 5], dtype)
data.persistable = True
out = paddle.fluid.layers.assign(data)
out = paddle.assign(data)
data_arr = np.random.uniform(-1, 1, data.shape).astype(dtype)
gradient_checker.double_grad_check(
......@@ -283,7 +283,7 @@ class TestAssignDoubleGradCheck(unittest.TestCase):
class TestAssignTripleGradCheck(unittest.TestCase):
def assign_wrapper(self, x):
return paddle.fluid.layers.assign(x[0])
return paddle.assign(x[0])
@prog_scope()
def func(self, place):
......@@ -293,7 +293,7 @@ class TestAssignTripleGradCheck(unittest.TestCase):
data = paddle.static.data('data', [3, 4, 5], dtype)
data.persistable = True
out = paddle.fluid.layers.assign(data)
out = paddle.assign(data)
data_arr = np.random.uniform(-1, 1, data.shape).astype(dtype)
gradient_checker.triple_grad_check(
......
......@@ -20,7 +20,6 @@ import op_test
import paddle
import paddle.fluid as fluid
import paddle.fluid.framework as framework
import paddle.fluid.layers as layers
paddle.enable_static()
......@@ -84,7 +83,7 @@ class TestAssignApi(unittest.TestCase):
main_program = fluid.Program()
with fluid.program_guard(main_program):
x = paddle.tensor.create_tensor(dtype=self.dtype)
layers.assign(input=self.value, output=x)
paddle.assign(self.value, output=x)
exe = fluid.Executor(self.place)
[fetched_x] = exe.run(main_program, feed={}, fetch_list=[x])
......
......@@ -114,7 +114,7 @@ class TestCastOpError(unittest.TestCase):
x1 = fluid.create_lod_tensor(
np.array([[-1]]), [[1]], fluid.CPUPlace()
)
self.assertRaises(TypeError, fluid.layers.cast, x1, 'int32')
self.assertRaises(TypeError, paddle.cast, x1, 'int32')
class TestCastOpEager(unittest.TestCase):
......
......@@ -49,7 +49,8 @@ class TestCommunicatorGeoEnd2End(unittest.TestCase):
pool = paddle.static.nn.sequence_lod.sequence_pool(
input=emb, pool_type="sum"
)
z = fluid.layers.concat(input=[x, pool], axis=1)
z = paddle.concat([x, pool], axis=1)
y_predict = paddle.static.nn.fc(x=z, size=1)
y = paddle.static.data(name='y', shape=[-1, 1], dtype='float32')
cost = paddle.nn.functional.square_error_cost(input=y_predict, label=y)
......
......@@ -450,8 +450,8 @@ class API_TestElementwise_Equal(unittest.TestCase):
def test_api(self):
paddle.enable_static()
with fluid.program_guard(fluid.Program(), fluid.Program()):
label = fluid.layers.assign(np.array([3, 3], dtype="int32"))
limit = fluid.layers.assign(np.array([3, 2], dtype="int32"))
label = paddle.assign(np.array([3, 3], dtype="int32"))
limit = paddle.assign(np.array([3, 2], dtype="int32"))
out = paddle.equal(x=label, y=limit)
place = fluid.CPUPlace()
exe = fluid.Executor(place)
......@@ -459,8 +459,8 @@ class API_TestElementwise_Equal(unittest.TestCase):
self.assertEqual((res == np.array([True, False])).all(), True)
with fluid.program_guard(fluid.Program(), fluid.Program()):
label = fluid.layers.assign(np.array([3, 3], dtype="int32"))
limit = fluid.layers.assign(np.array([3, 3], dtype="int32"))
label = paddle.assign(np.array([3, 3], dtype="int32"))
limit = paddle.assign(np.array([3, 3], dtype="int32"))
out = paddle.equal(x=label, y=limit)
place = fluid.CPUPlace()
exe = fluid.Executor(place)
......@@ -474,8 +474,8 @@ class TestCompareOpPlace(unittest.TestCase):
place = paddle.CPUPlace()
if core.is_compiled_with_cuda():
place = paddle.CUDAPlace(0)
label = fluid.layers.assign(np.array([3, 3], dtype="int32"))
limit = fluid.layers.assign(np.array([3, 2], dtype="int32"))
label = paddle.assign(np.array([3, 3], dtype="int32"))
limit = paddle.assign(np.array([3, 2], dtype="int32"))
out = paddle.less_than(label, limit)
exe = fluid.Executor(place)
(res,) = exe.run(fetch_list=[out])
......
......@@ -18,7 +18,6 @@ import numpy as np
import op_test
import paddle
import paddle.fluid as fluid
def create_test_not_equal_class(op_type, typename, callback):
......@@ -107,8 +106,8 @@ for _type_name in {'float32', 'float64', 'int32', 'int64', 'bool'}:
class TestEqualReduceAPI(unittest.TestCase):
def test_name(self):
x = fluid.layers.assign(np.array([3, 4], dtype="int32"))
y = fluid.layers.assign(np.array([3, 4], dtype="int32"))
x = paddle.assign(np.array([3, 4], dtype="int32"))
y = paddle.assign(np.array([3, 4], dtype="int32"))
out = paddle.equal_all(x, y, name='equal_res')
assert 'equal_res' in out.name
......
......@@ -249,8 +249,10 @@ class TestConcatOpError(unittest.TestCase):
def test_errors(self):
with program_guard(Program(), Program()):
# The input type of concat_op should be list.
x1 = paddle.static.data(shape=[-1, 4], dtype='int32', name='x1')
fluid.layers.concat(x1)
paddle.concat(x1)
# The item in input must be Variable.
x2 = fluid.create_lod_tensor(
np.array([[-1]]), [[1]], fluid.CPUPlace()
......@@ -258,24 +260,25 @@ class TestConcatOpError(unittest.TestCase):
x3 = fluid.create_lod_tensor(
np.array([[-1]]), [[1]], fluid.CPUPlace()
)
self.assertRaises(TypeError, fluid.layers.concat, [x2])
self.assertRaises(TypeError, paddle.concat, [x2])
# The input dtype of concat_op must be float16, float32, float64, int32, int64.
x4 = paddle.static.data(shape=[-1, 4], dtype='uint8', name='x4')
x5 = paddle.static.data(shape=[-1, 4], dtype='uint8', name='x5')
self.assertRaises(TypeError, fluid.layers.concat, [x4, x5])
self.assertRaises(TypeError, paddle.concat, [x4, x5])
x6 = paddle.static.data(shape=[-1, 4], dtype='float16', name='x6')
x7 = paddle.static.data(shape=[-1, 4], dtype='float16', name='x7')
x8 = paddle.static.data(shape=[-1, 4], dtype='float32', name='x8')
fluid.layers.concat([x6, x7])
paddle.concat([x6, x7])
# The type of axis in concat_op should be int or Variable.
def test_axis_type():
fluid.layers.concat([x6, x7], 3.2)
paddle.concat([x6, x7], 3.2)
self.assertRaises(TypeError, test_axis_type)
def test_input_same_dtype():
fluid.layers.concat([x7, x8])
paddle.concat([x7, x8])
self.assertRaises(TypeError, test_input_same_dtype)
......@@ -284,7 +287,7 @@ class TestConcatAPI(unittest.TestCase):
def test_fluid_api(self):
paddle.enable_static()
x_1 = fluid.data(shape=[None, 1, 4, 5], dtype='int32', name='x_1')
fluid.layers.concat([x_1, x_1], 0)
paddle.concat([x_1, x_1], 0)
input_2 = np.random.random([2, 1, 4, 5]).astype("int32")
input_3 = np.random.random([2, 2, 4, 5]).astype("int32")
......@@ -292,9 +295,9 @@ class TestConcatAPI(unittest.TestCase):
x_3 = fluid.data(shape=[2, 2, 4, 5], dtype='int32', name='x_3')
positive_1_int32 = fluid.layers.fill_constant([1], "int32", 1)
positive_1_int64 = fluid.layers.fill_constant([1], "int64", 1)
out_1 = fluid.layers.concat(input=[x_2, x_3], axis=1)
out_2 = fluid.layers.concat(input=[x_2, x_3], axis=positive_1_int32)
out_3 = fluid.layers.concat(input=[x_2, x_3], axis=positive_1_int64)
out_1 = paddle.concat([x_2, x_3], axis=1)
out_2 = paddle.concat([x_2, x_3], axis=positive_1_int32)
out_3 = paddle.concat([x_2, x_3], axis=positive_1_int64)
exe = fluid.Executor(place=fluid.CPUPlace())
[res_1, res_2, res_3] = exe.run(
......@@ -344,7 +347,7 @@ class TestConcatAPI(unittest.TestCase):
x1 = paddle.to_tensor(in1)
x2 = paddle.to_tensor(in2)
x3 = paddle.to_tensor(in3)
out1 = fluid.layers.concat(input=[x1, x2, x3], axis=-1)
out1 = paddle.concat([x1, x2, x3], axis=-1)
out2 = paddle.concat(x=[x1, x2], axis=0)
np_out1 = np.concatenate([in1, in2, in3], axis=-1)
np_out2 = np.concatenate([in1, in2], axis=0)
......@@ -365,7 +368,7 @@ class TestConcatAPI(unittest.TestCase):
# The input dtype of concat_op must be float16, float32, float64, int32, int64.
x4 = paddle.fluid.data(shape=[4], dtype='uint8', name='x4')
x5 = paddle.fluid.data(shape=[4], dtype='uint8', name='x5')
self.assertRaises(TypeError, fluid.layers.concat, [x4, x5])
self.assertRaises(TypeError, paddle.concat, [x4, x5])
# The type of axis in concat_op should be int or Variable.
x6 = paddle.static.data(shape=[-1, 4], dtype='float16', name='x6')
......@@ -405,7 +408,7 @@ class TestConcatAPIWithLoDTensorArray(unittest.TestCase):
if use_fluid_api:
self.program = fluid.Program()
with fluid.program_guard(self.program):
input = fluid.layers.assign(self.x)
input = paddle.assign(self.x)
tensor_array = paddle.tensor.create_array(dtype='float32')
zero = fluid.layers.fill_constant(
shape=[1], value=0, dtype="int64"
......@@ -414,7 +417,7 @@ class TestConcatAPIWithLoDTensorArray(unittest.TestCase):
for i in range(self.iter_num):
paddle.tensor.array_write(input, zero + i, tensor_array)
self.out_var = fluid.layers.concat(tensor_array, axis=self.axis)
self.out_var = paddle.concat(tensor_array, axis=self.axis)
else:
self.program = paddle.static.Program()
with paddle.static.program_guard(self.program):
......
......@@ -36,7 +36,7 @@ class ConditionalBlockTest(unittest.TestCase):
out = paddle.tensor.create_tensor(dtype='float32')
with cond.block():
hidden = paddle.static.nn.fc(x=data, size=10)
layers.assign(hidden, out)
paddle.assign(hidden, out)
cpu = core.CPUPlace()
exe = Executor(cpu)
......
......@@ -949,7 +949,7 @@ class TestDatasetWithFetchHandler(unittest.TestCase):
data = paddle.static.data(
name=slot, shape=[-1, 1], dtype="int64", lod_level=1
)
var = fluid.layers.cast(x=data, dtype='float32')
var = paddle.cast(x=data, dtype='float32')
pool = paddle.static.nn.sequence_lod.sequence_pool(
input=var, pool_type='AVERAGE'
)
......@@ -957,7 +957,7 @@ class TestDatasetWithFetchHandler(unittest.TestCase):
slots_vars.append(data)
poolings.append(pool)
concated = fluid.layers.concat(poolings, axis=1)
concated = paddle.concat(poolings, axis=1)
fc = paddle.static.nn.fc(x=concated, activation='tanh', size=32)
return slots_vars, fc
......
......@@ -95,7 +95,7 @@ class TestDistFleetHeterProgram(unittest.TestCase):
sparse_embed_seq = list(map(embedding_layer, inputs[1:-1]))
concated = fluid.layers.concat(sparse_embed_seq + inputs[0:1], axis=1)
concated = paddle.concat(sparse_embed_seq + inputs[0:1], axis=1)
with fluid.device_guard("gpu"):
fc1 = paddle.static.nn.fc(
......@@ -149,7 +149,7 @@ class TestDistFleetHeterProgram(unittest.TestCase):
)
with fluid.device_guard("gpu"):
labels = fluid.layers.cast(inputs[-1], dtype="int64")
labels = paddle.cast(inputs[-1], dtype="int64")
cost = paddle.nn.functional.cross_entropy(
input=predict, label=labels, reduction='none', use_softmax=False
)
......
......@@ -37,7 +37,7 @@ class TestPSMinimize(unittest.TestCase):
def net(self):
def get_acc(cos_q_nt, cos_q_pt, batch_size):
cond = paddle.less_than(cos_q_nt, cos_q_pt)
cond = fluid.layers.cast(cond, dtype='float64')
cond = paddle.cast(cond, dtype='float64')
cond_3 = paddle.sum(cond)
acc = paddle.divide(
cond_3,
......
......@@ -37,7 +37,7 @@ class TestPSPassWithBow(unittest.TestCase):
def net(self):
def get_acc(cos_q_nt, cos_q_pt, batch_size):
cond = paddle.less_than(cos_q_nt, cos_q_pt)
cond = fluid.layers.cast(cond, dtype='float64')
cond = paddle.cast(cond, dtype='float64')
cond_3 = paddle.sum(cond)
acc = paddle.divide(
cond_3,
......
......@@ -37,7 +37,7 @@ class TestPSPassWithBow(unittest.TestCase):
def net(self):
def get_acc(cos_q_nt, cos_q_pt, batch_size):
cond = paddle.less_than(cos_q_nt, cos_q_pt)
cond = fluid.layers.cast(cond, dtype='float64')
cond = paddle.cast(cond, dtype='float64')
cond_3 = paddle.sum(cond)
acc = paddle.divide(
cond_3,
......
......@@ -40,7 +40,7 @@ class TestPSPassWithBow(unittest.TestCase):
def net(self):
def get_acc(cos_q_nt, cos_q_pt, batch_size):
cond = paddle.less_than(cos_q_nt, cos_q_pt)
cond = fluid.layers.cast(cond, dtype='float64')
cond = paddle.cast(cond, dtype='float64')
cond_3 = paddle.sum(cond)
acc = paddle.divide(
cond_3,
......
......@@ -41,7 +41,7 @@ class TestPSPassWithBow(unittest.TestCase):
def net(self):
def get_acc(cos_q_nt, cos_q_pt, batch_size):
cond = paddle.less_than(cos_q_nt, cos_q_pt)
cond = fluid.layers.cast(cond, dtype='float64')
cond = paddle.cast(cond, dtype='float64')
cond_3 = paddle.sum(cond)
acc = paddle.divide(
cond_3,
......
......@@ -40,7 +40,7 @@ class TestPSPassWithBow(unittest.TestCase):
def net(self):
def get_acc(cos_q_nt, cos_q_pt, batch_size):
cond = paddle.less_than(cos_q_nt, cos_q_pt)
cond = fluid.layers.cast(cond, dtype='float64')
cond = paddle.cast(cond, dtype='float64')
cond_3 = paddle.sum(cond)
acc = paddle.divide(
cond_3,
......
......@@ -37,7 +37,7 @@ class TestPSPassWithBow(unittest.TestCase):
def net(self):
def get_acc(cos_q_nt, cos_q_pt, batch_size):
cond = paddle.less_than(cos_q_nt, cos_q_pt)
cond = fluid.layers.cast(cond, dtype='float64')
cond = paddle.cast(cond, dtype='float64')
cond_3 = paddle.sum(cond)
acc = paddle.divide(
cond_3,
......
......@@ -37,7 +37,7 @@ class TestPSPassWithBow(unittest.TestCase):
def net(self):
def get_acc(cos_q_nt, cos_q_pt, batch_size):
cond = paddle.less_than(cos_q_nt, cos_q_pt)
cond = fluid.layers.cast(cond, dtype='float64')
cond = paddle.cast(cond, dtype='float64')
cond_3 = paddle.sum(cond)
acc = paddle.divide(
cond_3,
......
......@@ -37,7 +37,7 @@ class TestPSPassWithBow(unittest.TestCase):
def net(self):
def get_acc(cos_q_nt, cos_q_pt, batch_size):
cond = paddle.less_than(cos_q_nt, cos_q_pt)
cond = fluid.layers.cast(cond, dtype='float64')
cond = paddle.cast(cond, dtype='float64')
cond_3 = paddle.sum(cond)
acc = paddle.divide(
cond_3,
......
......@@ -37,7 +37,7 @@ class TestPSPassWithBow(unittest.TestCase):
def net(self):
def get_acc(cos_q_nt, cos_q_pt, batch_size):
cond = paddle.less_than(cos_q_nt, cos_q_pt)
cond = fluid.layers.cast(cond, dtype='float64')
cond = paddle.cast(cond, dtype='float64')
cond_3 = paddle.sum(cond)
acc = paddle.divide(
cond_3,
......
......@@ -252,7 +252,7 @@ class TestDistMnistAsync2x2WithGauss(TestFleetBase):
lr_pool = paddle.static.nn.sequence_lod.sequence_pool(
input=lr_embbding, pool_type="sum"
)
merge_layer = fluid.layers.concat(input=[dnn_out, lr_pool], axis=1)
merge_layer = paddle.concat([dnn_out, lr_pool], axis=1)
predict = paddle.static.nn.fc(
x=merge_layer, size=2, activation='softmax'
)
......
......@@ -35,7 +35,7 @@ class TestSPMT(unittest.TestCase):
def net(self):
def get_acc(cos_q_nt, cos_q_pt, batch_size):
cond = paddle.less_than(cos_q_nt, cos_q_pt)
cond = fluid.layers.cast(cond, dtype='float64')
cond = paddle.cast(cond, dtype='float64')
cond_3 = paddle.sum(cond)
acc = paddle.divide(
cond_3,
......
......@@ -733,9 +733,7 @@ class TestDistLookupTableBase(TranspilerTest):
title_emb = emb_pool(title_ids, self.lookup_table_name, is_distributed)
brand_emb = emb_pool(brand_ids, self.lookup_table_name, is_distributed)
profile_emb = emb_pool(profile_ids, "profile_emb", False)
fc0 = fluid.layers.concat(
input=[title_emb, brand_emb, profile_emb], axis=1
)
fc0 = paddle.concat([title_emb, brand_emb, profile_emb], axis=1)
predict = paddle.static.nn.fc(
x=fc0,
size=2,
......
......@@ -29,7 +29,7 @@ def build_and_run_program(place, batch_size, beam_size, stop_gradient=False):
fluid.default_main_program().random_seed = 1
np.random.seed(2)
x = layers.assign(
x = paddle.assign(
np.random.rand(batch_size, beam_size, 32).astype("float32")
)
indices = fluid.data(shape=[None, beam_size], dtype="int64", name="indices")
......@@ -43,9 +43,9 @@ def build_and_run_program(place, batch_size, beam_size, stop_gradient=False):
while_op = paddle.static.nn.control_flow.While(cond)
scores = paddle.tensor.array_write(x, step_idx)
with while_op.block():
bs = layers.cast(paddle.shape(x)[0], "int64")
bs = paddle.cast(paddle.shape(x)[0], "int64")
for _ in range(20):
bs = layers.cast(bs, 'int64')
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]),
......@@ -57,7 +57,7 @@ def build_and_run_program(place, batch_size, beam_size, stop_gradient=False):
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)
layers.assign(length_cond, cond)
paddle.assign(length_cond, cond)
out = tensor_array_to_tensor(scores, axis=0, use_stack=True)[0]
loss = paddle.mean(out)
......
......@@ -164,7 +164,7 @@ def lm_model(
weight_1 = weight_1_arr[k]
bias = bias_arr[k]
nn = layers.concat([input, pre_hidden], 1)
nn = paddle.concat([input, pre_hidden], 1)
gate_input = paddle.matmul(x=nn, y=weight_1)
gate_input = paddle.add(gate_input, bias)
......@@ -230,8 +230,8 @@ def lm_model(
)
last_cell_array.append(last_c)
real_res = paddle.transpose(x=real_res, perm=[1, 0, 2])
last_hidden = layers.concat(last_hidden_array, 0)
last_cell = layers.concat(last_cell_array, 0)
last_hidden = paddle.concat(last_hidden_array, 0)
last_cell = paddle.concat(last_cell_array, 0)
return real_res, last_hidden, last_cell
......@@ -288,7 +288,7 @@ def lm_model(
weight_1 = weight_1_arr[k]
bias = bias_arr[k]
nn = layers.concat([input, pre_hidden], 1)
nn = paddle.concat([input, pre_hidden], 1)
gate_input = paddle.matmul(x=nn, y=weight_1)
gate_input = paddle.add(gate_input, bias)
......@@ -314,19 +314,19 @@ def lm_model(
res.append(input)
last_hidden = layers.concat(hidden_array, 1)
last_hidden = paddle.concat(hidden_array, 1)
last_hidden = paddle.reshape(
last_hidden, shape=[-1, num_layers, hidden_size]
)
last_hidden = paddle.transpose(x=last_hidden, perm=[1, 0, 2])
last_cell = layers.concat(cell_array, 1)
last_cell = paddle.concat(cell_array, 1)
last_cell = paddle.reshape(
last_cell, shape=[-1, num_layers, hidden_size]
)
last_cell = paddle.transpose(x=last_cell, perm=[1, 0, 2])
real_res = layers.concat(res, 0)
real_res = paddle.concat(res, 0)
real_res = paddle.reshape(real_res, shape=[len, -1, hidden_size])
real_res = paddle.transpose(x=real_res, perm=[1, 0, 2])
......@@ -439,8 +439,8 @@ def lm_model(
# can be used directly in next batch. This can avoid the fetching of
# last_hidden and last_cell and feeding of init_hidden and init_cell in
# each training step.
layers.assign(input=last_cell, output=init_cell)
layers.assign(input=last_hidden, output=init_hidden)
paddle.assign(last_cell, output=init_cell)
paddle.assign(last_hidden, output=init_hidden)
feeding_list = ['x', 'y', 'init_hidden', 'init_cell']
return loss, last_hidden, last_cell, feeding_list
......
......@@ -427,7 +427,7 @@ class EagerDeletionRecurrentOpMultipleMemoryTest(EagerDeletionRecurrentOpTest1):
mem1 = paddle.scale(x=h_pre1, scale=1.0)
mem2 = paddle.scale(x=h_pre2, scale=1.0)
out = layers.sums(input=[mem1, x_t, mem2])
out = paddle.add_n([mem1, x_t, mem2])
rnn.update_memory(h_pre1, mem1)
rnn.update_memory(h_pre2, mem2)
......
......@@ -104,7 +104,7 @@ class TestEagerDeletionWhileOpBase(unittest.TestCase):
prev = paddle.tensor.array_read(array=mem_array, i=i)
d = paddle.reshape(d, shape=[10])
prev = paddle.reshape(prev, shape=[10])
result = layers.sums(input=[d, prev])
result = paddle.add_n([d, prev])
i = paddle.increment(x=i)
paddle.tensor.array_write(result, i=i, array=mem_array)
......@@ -114,7 +114,7 @@ class TestEagerDeletionWhileOpBase(unittest.TestCase):
prev2 = paddle.tensor.array_read(array=mem_array, i=j)
d2 = paddle.reshape(d2, shape=[10])
prev2 = paddle.reshape(prev2, shape=[10])
result2 = layers.sums(input=[d2, prev2])
result2 = paddle.add_n([d2, prev2])
j = paddle.increment(x=j)
paddle.tensor.array_write(result2, i=j, array=mem_array)
......
......@@ -51,7 +51,7 @@ class TestEmbeddingIdStopGradientBase(unittest.TestCase):
with fluid.scope_guard(scope):
x_1 = fluid.data(name='x1', shape=[4, 1], dtype='int64')
x_2 = fluid.data(name='x2', shape=[4, 1], dtype='int64')
x = fluid.layers.concat([x_1, x_2], axis=-1)
x = paddle.concat([x_1, x_2], axis=-1)
for _ in range(self.reshape_times):
x = paddle.reshape(x, [-1, 1])
......
......@@ -18,7 +18,6 @@ import numpy as np
import paddle
import paddle.fluid as fluid
import paddle.fluid.layers as layers
class TestFetchVar(unittest.TestCase):
......@@ -30,7 +29,7 @@ class TestFetchVar(unittest.TestCase):
x = paddle.tensor.create_tensor(
dtype="int32", persistable=True, name="x"
)
layers.assign(input=self.val, output=x)
paddle.assign(self.val, output=x)
exe = fluid.Executor(fluid.CPUPlace())
exe.run(fluid.default_main_program(), feed={}, fetch_list=[])
fetched_x = fluid.executor._fetch_var("x")
......
......@@ -78,7 +78,7 @@ class TestFleet1(unittest.TestCase):
dtype="int64",
lod_level=1,
)
label_cast = fluid.layers.cast(label, dtype='float32')
label_cast = paddle.cast(label, dtype='float32')
cost = paddle.nn.functional.log_loss(fc, label_cast)
try:
adam = fluid.optimizer.Adam(learning_rate=0.000005)
......
......@@ -72,7 +72,7 @@ class TestFleet1(unittest.TestCase):
dtype="int64",
lod_level=1,
)
label_cast = fluid.layers.cast(label, dtype='float32')
label_cast = paddle.cast(label, dtype='float32')
cost = paddle.nn.functional.log_loss(fc, label_cast)
try:
adam = fluid.optimizer.Adam(learning_rate=0.000005)
......
......@@ -89,7 +89,7 @@ class TestCloudRoleMaker(unittest.TestCase):
label = paddle.static.data(
name="click", shape=[-1, 1], dtype="int64", lod_level=1
)
label_cast = fluid.layers.cast(label, dtype='float32')
label_cast = paddle.cast(label, dtype='float32')
cost = paddle.nn.functional.log_loss(fc, label_cast)
try:
adam = fluid.optimizer.Adam(learning_rate=0.000005)
......
......@@ -70,7 +70,7 @@ class TestCloudRoleMaker2(unittest.TestCase):
label = paddle.static.data(
name="click", shape=[-1, 1], dtype="int64", lod_level=1
)
label_cast = fluid.layers.cast(label, dtype='float32')
label_cast = paddle.cast(label, dtype='float32')
cost = paddle.nn.functional.log_loss(fc, label_cast)
try:
adam = fluid.optimizer.Adam(learning_rate=0.000005)
......
......@@ -63,7 +63,7 @@ class TestCloudRoleMaker(unittest.TestCase):
label = paddle.static.data(
name="click", shape=[-1, 1], dtype="int64", lod_level=1
)
label_cast = fluid.layers.cast(label, dtype='float32')
label_cast = paddle.cast(label, dtype='float32')
cost = paddle.nn.functional.log_loss(fc, label_cast)
try:
adam = fluid.optimizer.Adam(learning_rate=0.000005)
......
......@@ -66,7 +66,7 @@ class TestFleet1(unittest.TestCase):
label = paddle.static.data(
name="click", shape=[-1, 1], dtype="int64", lod_level=1
)
label_cast = fluid.layers.cast(label, dtype='float32')
label_cast = paddle.cast(label, dtype='float32')
cost = paddle.nn.functional.log_loss(fc, label_cast)
strategy = {}
......
......@@ -88,8 +88,8 @@ class AutoPruneLayer2(fluid.Layer):
def forward(self, x, label):
feature = self.linear(x)
label = self.linear2(label)
label = fluid.layers.cast(label, dtype="float32")
label = fluid.layers.cast(label, dtype='int64')
label = paddle.cast(label, dtype="float32")
label = paddle.cast(label, dtype='int64')
# Note that the label is not persistable in paddle.nn.functional.cross_entropy.
loss = paddle.nn.functional.cross_entropy(
input=feature, label=label, reduction='none', use_softmax=False
......@@ -244,7 +244,7 @@ class TestImperativeAutoPrune(unittest.TestCase):
out1 = linear(a)
out2 = linear2(b)
out1.stop_gradient = True
out = fluid.layers.concat(input=[out1, out2, c], axis=1)
out = paddle.concat([out1, out2, c], axis=1)
out.backward()
self.assertIsNone(linear.weight.gradient())
self.assertIsNone(out1.gradient())
......@@ -262,7 +262,7 @@ class TestImperativeAutoPrune(unittest.TestCase):
out1 = linear(a)
out2 = linear2(b)
out1.stop_gradient = True
out = fluid.layers.concat(input=[out1, out2, c], axis=1)
out = paddle.concat([out1, out2, c], axis=1)
out.backward()
self.assertIsNone(linear.weight.gradient())
self.assertIsNone(out1.gradient())
......@@ -338,7 +338,7 @@ class TestImperativeAutoPrune(unittest.TestCase):
out1 = linear(a)
out2 = linear2(b)
out1.stop_gradient = True
out = fluid.layers.concat(input=[out1, out2, c], axis=1)
out = paddle.concat([out1, out2, c], axis=1)
# TODO(jiabin): In Eager Mode we don't actually need sort_sum_gradient, this test should be removed when we don't support fluid anymore.
fluid.set_flags({'FLAGS_sort_sum_gradient': True})
out.backward()
......@@ -413,8 +413,8 @@ class TestImperativeAutoPrune(unittest.TestCase):
linear = paddle.nn.Linear(1, 1)
label = fluid.dygraph.to_variable(value1).astype("float32")
label = linear(label)
label = fluid.layers.cast(label, dtype="float32")
label = fluid.layers.cast(label, dtype='int64')
label = paddle.cast(label, dtype="float32")
label = paddle.cast(label, dtype='int64')
out = paddle.nn.functional.one_hot(label, 100)
loss = paddle.mean(out)
loss.backward()
......
此差异已折叠。
此差异已折叠。
此差异已折叠。
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册