fix memory usage
This commit is contained in:
lucasdelimanogueira
parent
e670559fe4
commit
e6d58121f1
16 changed files with 754 additions and 559 deletions
BIN
build/cuda.cu.o
BIN
build/cuda.cu.o
Binary file not shown.
BIN
build/tensor.o
BIN
build/tensor.o
Binary file not shown.
28
mprofile_20240605162109.dat
Normal file
28
mprofile_20240605162109.dat
Normal file
|
|
@ -0,0 +1,28 @@
|
|||
CMDLINE /usr/bin/python3 train_singlegpu.py
|
||||
MEM 1.109375 1717615269.6916
|
||||
MEM 14.167969 1717615269.7921
|
||||
MEM 21.085938 1717615269.8925
|
||||
MEM 26.617188 1717615269.9930
|
||||
MEM 32.312500 1717615270.0934
|
||||
MEM 38.429688 1717615270.1939
|
||||
MEM 40.683594 1717615270.2943
|
||||
MEM 46.445312 1717615270.3947
|
||||
MEM 54.089844 1717615270.4953
|
||||
MEM 58.347656 1717615270.5959
|
||||
MEM 63.308594 1717615270.6964
|
||||
MEM 65.226562 1717615270.7970
|
||||
MEM 70.898438 1717615270.8975
|
||||
MEM 76.828125 1717615270.9980
|
||||
MEM 82.242188 1717615271.0986
|
||||
MEM 87.656250 1717615271.1991
|
||||
MEM 93.328125 1717615271.2996
|
||||
MEM 98.742188 1717615271.4001
|
||||
MEM 104.671875 1717615271.5006
|
||||
MEM 135.867188 1717615271.6011
|
||||
MEM 117.457031 1717615271.7016
|
||||
MEM 125.246094 1717615271.8022
|
||||
MEM 125.503906 1717615271.9026
|
||||
MEM 125.503906 1717615272.0029
|
||||
MEM 127.074219 1717615272.1032
|
||||
MEM 127.074219 1717615272.2036
|
||||
MEM 73.593750 1717615272.3040
|
||||
257
mprofile_20240605162319.dat
Normal file
257
mprofile_20240605162319.dat
Normal file
|
|
@ -0,0 +1,257 @@
|
|||
CMDLINE /usr/bin/python3 train_singlegpu.py
|
||||
MEM 1.171875 1717615399.9716
|
||||
MEM 18.777344 1717615400.0720
|
||||
MEM 25.433594 1717615400.1725
|
||||
MEM 30.570312 1717615400.2729
|
||||
MEM 36.207031 1717615400.3734
|
||||
MEM 40.167969 1717615400.4738
|
||||
MEM 46.121094 1717615400.5743
|
||||
MEM 56.503906 1717615400.6747
|
||||
MEM 63.429688 1717615400.7751
|
||||
MEM 70.007812 1717615400.8755
|
||||
MEM 84.187500 1717615400.9758
|
||||
MEM 98.109375 1717615401.0761
|
||||
MEM 140.132812 1717615401.1765
|
||||
MEM 125.375000 1717615401.2769
|
||||
MEM 125.632812 1717615401.3772
|
||||
MEM 126.433594 1717615401.4775
|
||||
MEM 127.203125 1717615401.5778
|
||||
MEM 127.203125 1717615401.6781
|
||||
MEM 127.316406 1717615401.7785
|
||||
MEM 127.316406 1717615401.8787
|
||||
MEM 134.515625 1717615401.9790
|
||||
MEM 134.773438 1717615402.0793
|
||||
MEM 136.054688 1717615402.1796
|
||||
MEM 140.953125 1717615402.2799
|
||||
MEM 140.953125 1717615402.3802
|
||||
MEM 146.867188 1717615402.4805
|
||||
MEM 147.125000 1717615402.5809
|
||||
MEM 153.562500 1717615402.6813
|
||||
MEM 154.335938 1717615402.7817
|
||||
MEM 154.335938 1717615402.8820
|
||||
MEM 154.335938 1717615402.9825
|
||||
MEM 154.335938 1717615403.0828
|
||||
MEM 154.335938 1717615403.1832
|
||||
MEM 154.335938 1717615403.2836
|
||||
MEM 161.039062 1717615403.3839
|
||||
MEM 161.812500 1717615403.4842
|
||||
MEM 162.328125 1717615403.5845
|
||||
MEM 169.273438 1717615403.6848
|
||||
MEM 169.531250 1717615403.7852
|
||||
MEM 176.992188 1717615403.8855
|
||||
MEM 176.992188 1717615403.9857
|
||||
MEM 184.449219 1717615404.0860
|
||||
MEM 184.449219 1717615404.1863
|
||||
MEM 186.488281 1717615404.2866
|
||||
MEM 191.898438 1717615404.3869
|
||||
MEM 192.156250 1717615404.4872
|
||||
MEM 199.632812 1717615404.5875
|
||||
MEM 199.632812 1717615404.6878
|
||||
MEM 207.109375 1717615404.7881
|
||||
MEM 207.109375 1717615404.8884
|
||||
MEM 213.296875 1717615404.9887
|
||||
MEM 214.761719 1717615405.0890
|
||||
MEM 214.957031 1717615405.1894
|
||||
MEM 222.171875 1717615405.2897
|
||||
MEM 222.429688 1717615405.3900
|
||||
MEM 229.648438 1717615405.4903
|
||||
MEM 229.906250 1717615405.5906
|
||||
MEM 236.093750 1717615405.6910
|
||||
MEM 237.382812 1717615405.7913
|
||||
MEM 237.898438 1717615405.8916
|
||||
MEM 244.859375 1717615405.9919
|
||||
MEM 245.117188 1717615406.0922
|
||||
MEM 252.335938 1717615406.1925
|
||||
MEM 252.593750 1717615406.2928
|
||||
MEM 258.781250 1717615406.3931
|
||||
MEM 260.070312 1717615406.4934
|
||||
MEM 260.585938 1717615406.5937
|
||||
MEM 267.546875 1717615406.6940
|
||||
MEM 267.804688 1717615406.7944
|
||||
MEM 275.023438 1717615406.8947
|
||||
MEM 275.281250 1717615406.9949
|
||||
MEM 281.726562 1717615407.0953
|
||||
MEM 282.757812 1717615407.1956
|
||||
MEM 283.273438 1717615407.2959
|
||||
MEM 290.234375 1717615407.3962
|
||||
MEM 290.234375 1717615407.4965
|
||||
MEM 297.710938 1717615407.5969
|
||||
MEM 297.710938 1717615407.6972
|
||||
MEM 298.226562 1717615407.7976
|
||||
MEM 305.187500 1717615407.8979
|
||||
MEM 305.445312 1717615407.9983
|
||||
MEM 305.445312 1717615408.0987
|
||||
MEM 312.921875 1717615408.1990
|
||||
MEM 312.921875 1717615408.2993
|
||||
MEM 319.109375 1717615408.3996
|
||||
MEM 320.398438 1717615408.4999
|
||||
MEM 320.656250 1717615408.6002
|
||||
MEM 327.875000 1717615408.7005
|
||||
MEM 328.132812 1717615408.8008
|
||||
MEM 335.609375 1717615408.9011
|
||||
MEM 335.609375 1717615409.0014
|
||||
MEM 341.796875 1717615409.1018
|
||||
MEM 343.085938 1717615409.2021
|
||||
MEM 343.601562 1717615409.3024
|
||||
MEM 350.562500 1717615409.4026
|
||||
MEM 350.820312 1717615409.5029
|
||||
MEM 358.039062 1717615409.6032
|
||||
MEM 358.296875 1717615409.7035
|
||||
MEM 365.000000 1717615409.8039
|
||||
MEM 365.773438 1717615409.9042
|
||||
MEM 366.289062 1717615410.0044
|
||||
MEM 373.250000 1717615410.1047
|
||||
MEM 373.507812 1717615410.2051
|
||||
MEM 380.726562 1717615410.3055
|
||||
MEM 380.984375 1717615410.4057
|
||||
MEM 388.460938 1717615410.5060
|
||||
MEM 388.460938 1717615410.6064
|
||||
MEM 390.523438 1717615410.7066
|
||||
MEM 395.937500 1717615410.8069
|
||||
MEM 396.195312 1717615410.9072
|
||||
MEM 403.414062 1717615411.0075
|
||||
MEM 403.671875 1717615411.1078
|
||||
MEM 411.148438 1717615411.2081
|
||||
MEM 411.148438 1717615411.3084
|
||||
MEM 417.335938 1717615411.4088
|
||||
MEM 418.625000 1717615411.5091
|
||||
MEM 418.882812 1717615411.6094
|
||||
MEM 426.101562 1717615411.7097
|
||||
MEM 426.359375 1717615411.8100
|
||||
MEM 433.578125 1717615411.9103
|
||||
MEM 433.835938 1717615412.0106
|
||||
MEM 440.023438 1717615412.1110
|
||||
MEM 441.312500 1717615412.2113
|
||||
MEM 441.828125 1717615412.3116
|
||||
MEM 448.789062 1717615412.4119
|
||||
MEM 449.046875 1717615412.5122
|
||||
MEM 456.265625 1717615412.6125
|
||||
MEM 456.523438 1717615412.7127
|
||||
MEM 462.968750 1717615412.8131
|
||||
MEM 464.000000 1717615412.9133
|
||||
MEM 464.515625 1717615413.0136
|
||||
MEM 471.476562 1717615413.1139
|
||||
MEM 471.734375 1717615413.2142
|
||||
MEM 478.949219 1717615413.3145
|
||||
MEM 479.207031 1717615413.4148
|
||||
MEM 486.683594 1717615413.5151
|
||||
MEM 486.683594 1717615413.6154
|
||||
MEM 487.457031 1717615413.7157
|
||||
MEM 494.160156 1717615413.8160
|
||||
MEM 494.417969 1717615413.9163
|
||||
MEM 501.636719 1717615414.0166
|
||||
MEM 501.894531 1717615414.1169
|
||||
MEM 509.113281 1717615414.2172
|
||||
MEM 509.371094 1717615414.3175
|
||||
MEM 509.886719 1717615414.4178
|
||||
MEM 516.847656 1717615414.5181
|
||||
MEM 516.847656 1717615414.6184
|
||||
MEM 524.324219 1717615414.7187
|
||||
MEM 524.582031 1717615414.8190
|
||||
MEM 531.027344 1717615414.9193
|
||||
MEM 532.058594 1717615415.0196
|
||||
MEM 532.574219 1717615415.1199
|
||||
MEM 539.535156 1717615415.2202
|
||||
MEM 539.535156 1717615415.3205
|
||||
MEM 547.011719 1717615415.4209
|
||||
MEM 547.269531 1717615415.5211
|
||||
MEM 554.488281 1717615415.6214
|
||||
MEM 554.746094 1717615415.7217
|
||||
MEM 556.808594 1717615415.8220
|
||||
MEM 562.222656 1717615415.9223
|
||||
MEM 562.222656 1717615416.0226
|
||||
MEM 569.699219 1717615416.1229
|
||||
MEM 569.957031 1717615416.2232
|
||||
MEM 577.175781 1717615416.3235
|
||||
MEM 577.433594 1717615416.4238
|
||||
MEM 582.332031 1717615416.5241
|
||||
MEM 584.910156 1717615416.6244
|
||||
MEM 584.910156 1717615416.7248
|
||||
MEM 592.386719 1717615416.8251
|
||||
MEM 592.386719 1717615416.9253
|
||||
MEM 599.863281 1717615417.0257
|
||||
MEM 600.121094 1717615417.1259
|
||||
MEM 603.472656 1717615417.2262
|
||||
MEM 606.523438 1717615417.3266
|
||||
MEM 606.523438 1717615417.4269
|
||||
MEM 606.523438 1717615417.5272
|
||||
MEM 606.523438 1717615417.6275
|
||||
MEM 606.523438 1717615417.7278
|
||||
MEM 606.523438 1717615417.8281
|
||||
MEM 606.523438 1717615417.9284
|
||||
MEM 606.523438 1717615418.0287
|
||||
MEM 606.523438 1717615418.1290
|
||||
MEM 606.523438 1717615418.2293
|
||||
MEM 606.523438 1717615418.3296
|
||||
MEM 606.523438 1717615418.4299
|
||||
MEM 606.523438 1717615418.5301
|
||||
MEM 606.523438 1717615418.6305
|
||||
MEM 606.523438 1717615418.7308
|
||||
MEM 606.523438 1717615418.8311
|
||||
MEM 606.523438 1717615418.9314
|
||||
MEM 606.523438 1717615419.0317
|
||||
MEM 606.523438 1717615419.1320
|
||||
MEM 606.523438 1717615419.2323
|
||||
MEM 606.523438 1717615419.3327
|
||||
MEM 606.523438 1717615419.4330
|
||||
MEM 606.523438 1717615419.5333
|
||||
MEM 606.523438 1717615419.6336
|
||||
MEM 606.523438 1717615419.7340
|
||||
MEM 606.523438 1717615419.8345
|
||||
MEM 606.523438 1717615419.9348
|
||||
MEM 606.523438 1717615420.0351
|
||||
MEM 606.523438 1717615420.1354
|
||||
MEM 606.523438 1717615420.2357
|
||||
MEM 606.523438 1717615420.3360
|
||||
MEM 606.523438 1717615420.4363
|
||||
MEM 606.523438 1717615420.5366
|
||||
MEM 606.523438 1717615420.6370
|
||||
MEM 606.523438 1717615420.7373
|
||||
MEM 606.523438 1717615420.8377
|
||||
MEM 606.523438 1717615420.9381
|
||||
MEM 606.523438 1717615421.0384
|
||||
MEM 606.523438 1717615421.1387
|
||||
MEM 606.523438 1717615421.2390
|
||||
MEM 606.523438 1717615421.3393
|
||||
MEM 606.523438 1717615421.4396
|
||||
MEM 606.523438 1717615421.5399
|
||||
MEM 606.523438 1717615421.6402
|
||||
MEM 606.523438 1717615421.7405
|
||||
MEM 606.523438 1717615421.8408
|
||||
MEM 606.523438 1717615421.9411
|
||||
MEM 606.523438 1717615422.0415
|
||||
MEM 606.523438 1717615422.1418
|
||||
MEM 612.191406 1717615422.2421
|
||||
MEM 612.191406 1717615422.3425
|
||||
MEM 612.707031 1717615422.4428
|
||||
MEM 618.636719 1717615422.5431
|
||||
MEM 618.636719 1717615422.6434
|
||||
MEM 625.597656 1717615422.7437
|
||||
MEM 625.597656 1717615422.8441
|
||||
MEM 631.785156 1717615422.9444
|
||||
MEM 632.816406 1717615423.0447
|
||||
MEM 633.074219 1717615423.1450
|
||||
MEM 640.292969 1717615423.2453
|
||||
MEM 640.292969 1717615423.3456
|
||||
MEM 647.769531 1717615423.4459
|
||||
MEM 647.769531 1717615423.5461
|
||||
MEM 654.214844 1717615423.6465
|
||||
MEM 655.246094 1717615423.7468
|
||||
MEM 655.761719 1717615423.8471
|
||||
MEM 662.722656 1717615423.9474
|
||||
MEM 662.980469 1717615424.0477
|
||||
MEM 670.199219 1717615424.1480
|
||||
MEM 670.457031 1717615424.2483
|
||||
MEM 677.160156 1717615424.3486
|
||||
MEM 677.933594 1717615424.4489
|
||||
MEM 678.449219 1717615424.5492
|
||||
MEM 685.410156 1717615424.6495
|
||||
MEM 685.410156 1717615424.7498
|
||||
MEM 692.886719 1717615424.8501
|
||||
MEM 692.886719 1717615424.9504
|
||||
MEM 700.363281 1717615425.0506
|
||||
MEM 700.363281 1717615425.1509
|
||||
MEM 700.878906 1717615425.2512
|
||||
MEM 707.839844 1717615425.3515
|
||||
MEM 708.097656 1717615425.4518
|
||||
MEM 652.933594 1717615425.5522
|
||||
Binary file not shown.
|
|
@ -24,9 +24,8 @@ __host__ void cpu_to_cuda(Tensor* tensor, int device_id) {
|
|||
|
||||
tensor->data = data_tmp;
|
||||
|
||||
const char* device_str = "cuda";
|
||||
tensor->device = (char*)malloc(strlen(device_str) + 1);
|
||||
strcpy(tensor->device, device_str);
|
||||
tensor->device = (char*)malloc(strlen("cuda") + 1);
|
||||
strcpy(tensor->device, "cuda");
|
||||
}
|
||||
|
||||
__host__ void cuda_to_cpu(Tensor* tensor) {
|
||||
|
|
@ -37,9 +36,8 @@ __host__ void cuda_to_cpu(Tensor* tensor) {
|
|||
|
||||
tensor->data = data_tmp;
|
||||
|
||||
const char* device_str = "cpu";
|
||||
tensor->device = (char*)malloc(strlen(device_str) + 1);
|
||||
strcpy(tensor->device, device_str);
|
||||
tensor->device = (char*)malloc(strlen("cpu") + 1);
|
||||
strcpy(tensor->device, "cpu");
|
||||
}
|
||||
|
||||
__host__ void free_cuda(float* data) {
|
||||
|
|
|
|||
File diff suppressed because it is too large
Load diff
|
|
@ -13,6 +13,8 @@ typedef struct {
|
|||
extern "C" {
|
||||
Tensor* create_tensor(float* data, int* shape, int ndim, char* device);
|
||||
void delete_tensor(Tensor* tensor);
|
||||
void delete_strides(Tensor* tensor);
|
||||
void delete_device(Tensor* tensor);
|
||||
float get_item(Tensor* tensor, int* indices);
|
||||
Tensor* add_tensor(Tensor* tensor1, Tensor* tensor2);
|
||||
Tensor* sum_tensor(Tensor* tensor, int axis, bool keepdims);
|
||||
|
|
|
|||
Binary file not shown.
Binary file not shown.
|
|
@ -11,7 +11,7 @@ class Loss(Module, ABC):
|
|||
|
||||
def forward(self, predictions, labels):
|
||||
raise NotImplementedError
|
||||
|
||||
|
||||
def __call__(self, *inputs):
|
||||
return self.forward(*inputs)
|
||||
|
||||
|
|
|
|||
Binary file not shown.
|
|
@ -9,6 +9,7 @@ class SGD(Optimizer):
|
|||
self.momentum = momentum
|
||||
self._cache = {'velocity': [p.zeros_like() for (_, _, p) in self.parameters]}
|
||||
|
||||
|
||||
def step(self):
|
||||
for i, (module, name, _) in enumerate(self.parameters):
|
||||
parameter = getattr(module, name)
|
||||
|
|
|
|||
|
|
@ -26,10 +26,10 @@ class Tensor:
|
|||
|
||||
self.shape = shape.copy()
|
||||
|
||||
self.data_ctype = (ctypes.c_float * len(data))(*data.copy())
|
||||
self.shape_ctype = (ctypes.c_int * len(shape))(*shape.copy())
|
||||
self.ndim_ctype = ctypes.c_int(len(shape))
|
||||
self.device_ctype = device.encode('utf-8')
|
||||
self._data_ctype = (ctypes.c_float * len(data))(*data.copy())
|
||||
self._shape_ctype = (ctypes.c_int * len(shape))(*shape.copy())
|
||||
self._ndim_ctype = ctypes.c_int(len(shape))
|
||||
self._device_ctype = device.encode('utf-8')
|
||||
|
||||
self.ndim = len(shape)
|
||||
self.device = device
|
||||
|
|
@ -47,15 +47,11 @@ class Tensor:
|
|||
Tensor._C.create_tensor.restype = ctypes.POINTER(CTensor)
|
||||
|
||||
self.tensor = Tensor._C.create_tensor(
|
||||
self.data_ctype,
|
||||
self.shape_ctype,
|
||||
self.ndim_ctype,
|
||||
self.device_ctype
|
||||
self._data_ctype,
|
||||
self._shape_ctype,
|
||||
self._ndim_ctype,
|
||||
self._device_ctype
|
||||
)
|
||||
|
||||
del self.data_ctype
|
||||
del self.shape_ctype
|
||||
del self.device_ctype
|
||||
|
||||
else:
|
||||
self.tensor = None,
|
||||
|
|
@ -85,7 +81,21 @@ class Tensor:
|
|||
return flat_data, shape
|
||||
|
||||
def __del__(self):
|
||||
if self.tensor is not None:
|
||||
|
||||
if hasattr(self, '_data_ctype') and self._data_ctype is not None:
|
||||
# tensor created by user (ctypes) will be deleted by python garbage collector
|
||||
# only strides need to be deallocated manually because it is created inside C code
|
||||
Tensor._C.delete_strides.argtypes = [ctypes.POINTER(CTensor)]
|
||||
Tensor._C.delete_strides.restype = None
|
||||
Tensor._C.delete_strides(self.tensor)
|
||||
|
||||
Tensor._C.delete_device.argtypes = [ctypes.POINTER(CTensor)]
|
||||
Tensor._C.delete_device.restype = None
|
||||
Tensor._C.delete_device(self.tensor)
|
||||
|
||||
elif self.tensor is not None:
|
||||
# tensor created during operations must be deallocated
|
||||
|
||||
Tensor._C.delete_tensor.argtypes = [ctypes.POINTER(CTensor)]
|
||||
Tensor._C.delete_tensor.restype = None
|
||||
Tensor._C.delete_tensor(self.tensor)
|
||||
|
|
|
|||
BIN
profiling_results.prof
Normal file
BIN
profiling_results.prof
Normal file
Binary file not shown.
97
train_singlegpu.py
Normal file
97
train_singlegpu.py
Normal file
|
|
@ -0,0 +1,97 @@
|
|||
import norch
|
||||
import norch.nn as nn
|
||||
import norch.optim as optim
|
||||
from norch.norchvision import transforms as T
|
||||
import random
|
||||
random.seed(1)
|
||||
from memory_profiler import profile
|
||||
|
||||
@profile
|
||||
def main():
|
||||
|
||||
|
||||
BATCH_SIZE = 32
|
||||
device = "cpu"
|
||||
epochs = 10
|
||||
|
||||
transform = T.Compose(
|
||||
[
|
||||
T.ToTensor(),
|
||||
T.Reshape([-1, 784, 1])
|
||||
]
|
||||
)
|
||||
|
||||
target_transform = T.Compose(
|
||||
[
|
||||
T.ToTensor()
|
||||
]
|
||||
)
|
||||
|
||||
|
||||
print("Loading data")
|
||||
train_data, test_data = norch.norchvision.datasets.MNIST.splits(transform=transform, target_transform=target_transform)
|
||||
train_loader = norch.utils.data.DataLoader(train_data, batch_size=BATCH_SIZE)
|
||||
|
||||
class MyModel(nn.Module):
|
||||
def __init__(self):
|
||||
super(MyModel, self).__init__()
|
||||
self.fc1 = nn.Linear(784, 30)
|
||||
self.sigmoid1 = nn.Sigmoid()
|
||||
self.fc2 = nn.Linear(30, 10)
|
||||
self.sigmoid2 = nn.Sigmoid()
|
||||
|
||||
def forward(self, x):
|
||||
out = self.fc1(x)
|
||||
out = self.sigmoid1(out)
|
||||
out = self.fc2(out)
|
||||
out = self.sigmoid2(out)
|
||||
|
||||
return out
|
||||
|
||||
print("Creating model")
|
||||
model = MyModel().to(device)
|
||||
criterion = nn.CrossEntropyLoss()
|
||||
optimizer = optim.SGD(model.parameters(), lr=0.01)
|
||||
loss_list = []
|
||||
|
||||
print("Starting training")
|
||||
for epoch in range(epochs):
|
||||
|
||||
avg_loss = 0
|
||||
num_steps = 0
|
||||
|
||||
for idx, batch in enumerate(train_loader):
|
||||
|
||||
if idx % 100 == 0 and idx > 0:
|
||||
print(f"Epoch: {epoch}/{epochs} - Step: {idx} / {len(train_loader)}")
|
||||
break
|
||||
|
||||
inputs, target = batch
|
||||
|
||||
inputs = inputs.to(device)
|
||||
target = target.to(device)
|
||||
|
||||
outputs = model(inputs)
|
||||
loss = criterion(outputs, target)
|
||||
|
||||
optimizer.zero_grad()
|
||||
|
||||
loss.backward()
|
||||
|
||||
optimizer.step()
|
||||
|
||||
avg_loss += loss[0]
|
||||
num_steps += 1
|
||||
|
||||
avg_loss = avg_loss / num_steps
|
||||
print(f'Epoch [{epoch + 1}/{epochs}], Loss: {avg_loss:.4f}')
|
||||
loss_list.append(avg_loss)
|
||||
|
||||
break
|
||||
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
main()
|
||||
|
||||
|
||||
Loading…
Add table
Add a link
Reference in a new issue