m53297601
/
mindspore
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

delete unused arguments in test cases

Browse Source
pull/1424/head
Yi Huaijie 5 years ago
parent 8cfc05e4cf
commit 0b137e5312
10 changed files with 158 additions and 86 deletions
Show Stats Download Patch File Download Diff File

30
tests/ut/python/parallel/test_arithmetic.py
Unescape View File

@ -493,15 +493,35 @@ def test_assign_sub():
1.1, dtype=np.float32)),
name="assignsub_weight")
def construct(self, x, y, z):
def construct(self, x):
out = self.mul(x, self.mul_weight)
out = self.assign_sub(self.assignsub_weight, out)
return out
class SubNetWithLoss(nn.Cell):
def __init__(self, network):
super(SubNetWithLoss, self).__init__()
self.loss = VirtualLoss()
self.network = network
def construct(self, x):
predict = self.network(x,)
return self.loss(predict)
class SubGradWrap(nn.Cell):
def __init__(self, network):
super(SubGradWrap, self).__init__()
self.network = network
def construct(self, x):
return C.grad_all(self.network)(x)
def compile_sub_net(net, x):
net.set_auto_parallel()
_executor.compile(net, x)
context.set_auto_parallel_context(device_num=64, global_rank=15)
context.set_auto_parallel_context(parallel_mode="semi_auto_parallel")
net = GradWrap(NetWithLoss(Net()))
net = SubGradWrap(SubNetWithLoss(Net()))
x = Tensor(np.ones([128, 32]), dtype=ms.float32)
y = Tensor(np.ones([128, 32]), dtype=ms.float32)
z = Tensor(np.ones([128, 32]), dtype=ms.float32)
compile_net(net, x, y, z)
compile_sub_net(net, x)

13
tests/ut/python/parallel/test_auto_parallel_parameter_cast.py
Unescape View File

@ -20,7 +20,6 @@ from mindspore import Tensor, Parameter
from mindspore import context
from mindspore.common import dtype as mstype
from mindspore.common.api import _executor
from mindspore.ops import composite as C
from mindspore.ops import operations as P
from mindspore.parallel import set_algo_parameters
from mindspore.parallel._utils import _reset_op_id as reset_op_id
@ -33,8 +32,8 @@ class NetWithLoss(nn.Cell):
self.loss = VirtualLoss()
self.network = network
def construct(self, x, y, z, w):
predict = self.network(x, y, z, w)
def construct(self, x, y):
predict = self.network(x, y)
return self.loss(predict)
@ -49,9 +48,9 @@ def test_common_parameter():
self.cast1 = P.Cast()
self.cast2 = P.Cast()
def construct(self, x, y, z, w):
def construct(self, x, y):
m1_result = self.matmul1(x, self.cast1(self.weight1, mstype.float32))
m2_result = self.matmul2(z, self.cast2(self.weight1, mstype.float32))
m2_result = self.matmul2(y, self.cast2(self.weight1, mstype.float32))
m3_result = self.matmul3(m2_result, m1_result)
return m3_result
@ -62,15 +61,13 @@ def test_common_parameter():
set_algo_parameters(elementwise_op_strategy_follow=True)
x = Tensor(np.ones([64, 64]), dtype=ms.float32)
y = Tensor(np.ones([64, 64]), dtype=ms.float32)
z = Tensor(np.ones([64, 64]), dtype=ms.float32)
w = Tensor(np.ones([64, 64]), dtype=ms.float32)
net = NetWithLoss(Net())
context.set_auto_parallel_context(parallel_mode="auto_parallel")
net.set_auto_parallel()
reset_op_id()
_executor.compile(net, x, y, z, w, phase='train')
_executor.compile(net, x, y, phase='train')
strategies = _executor._get_strategy(net)
expected_strategies = {'Default/network-Net/MatMul-op1': [[8, 1], [1, 1]],
'Default/network-Net/MatMul-op3': [[8, 1], [1, 1]],

11
tests/ut/python/parallel/test_dataset_interface.py
Unescape View File

@ -135,7 +135,11 @@ def test_dataset_interface_sens_shape_not_equal_loss():
sens = Tensor(np.ones([256, 1024]), dtype=ms.float32)
try:
loss_scale_manager_sens(strategy1, sens)
except BaseException:
except ValueError:
pass
except TypeError:
pass
except RuntimeError:
pass
@ -153,7 +157,7 @@ def test_input_not_in_parameter_layotu_dict():
self.matmul_weight = Parameter(Tensor(np.ones([128, 256]), dtype=ms.float32), name="weight")
self.transpose1 = P.Transpose().set_strategy(strategy1)
def construct(self, x, b):
def construct(self, x):
x = self.matmul(x, self.matmul_weight)
x = self.transpose1(x, (1, 0))
return x
@ -163,7 +167,6 @@ def test_input_not_in_parameter_layotu_dict():
context.reset_auto_parallel_context()
context.set_auto_parallel_context(parallel_mode=ParallelMode.SEMI_AUTO_PARALLEL, device_num=device_num)
predict = Tensor(np.ones([32 * device_num, 128]), dtype=ms.float32)
b = Tensor(np.ones([32 * device_num, 128]), dtype=ms.float32)
net = Net(strategy1)
net.set_train()
net(predict, b)
net(predict)

23
tests/ut/python/parallel/test_different_type_for_div_op.py
Unescape View File

@ -28,13 +28,13 @@ class GradWrap(nn.Cell):
super(GradWrap, self).__init__()
self.network = network
def construct(self, x, y, bias):
return C.grad_all(self.network)(x, y, bias)
def construct(self, x, y):
return C.grad_all(self.network)(x, y)
def compile_net(net, x, y, bias):
def compile_net(net, x, y):
net.set_auto_parallel()
_executor.compile(net, x, y, bias)
_executor.compile(net, x, y)
def test_sum_as_loss_float16():
@ -44,7 +44,7 @@ def test_sum_as_loss_float16():
self.fc_nobias = P.MatMul(transpose_b=True).set_strategy(strategy0)
self.reduce_sum = P.ReduceSum(keep_dims=False).set_strategy(strategy1)
def construct(self, x, y, bias):
def construct(self, x, y):
out = self.fc_nobias(x, y)
out = self.reduce_sum(out, (0, 1))
return out
@ -57,8 +57,7 @@ def test_sum_as_loss_float16():
x = Tensor(np.ones([64, 32]), dtype=ms.float16)
y = Tensor(np.ones([64, 32]), dtype=ms.float16)
bias = Tensor(np.ones([64]), dtype=ms.float16)
compile_net(net, x, y, bias)
compile_net(net, x, y)
def test_sum_as_loss_float32():
@ -68,7 +67,7 @@ def test_sum_as_loss_float32():
self.fc_nobias = P.MatMul(transpose_b=True).set_strategy(strategy0)
self.reduce_sum = P.ReduceSum(keep_dims=False).set_strategy(strategy1)
def construct(self, x, y, bias):
def construct(self, x, y):
out = self.fc_nobias(x, y)
out = self.reduce_sum(out, (0, 1))
return out
@ -81,8 +80,7 @@ def test_sum_as_loss_float32():
x = Tensor(np.ones([64, 32]), dtype=ms.float32)
y = Tensor(np.ones([64, 32]), dtype=ms.float32)
bias = Tensor(np.ones([64]), dtype=ms.float32)
compile_net(net, x, y, bias)
compile_net(net, x, y)
def test_sum_as_loss_int32():
@ -92,7 +90,7 @@ def test_sum_as_loss_int32():
self.fc_nobias = P.MatMul(transpose_b=True).set_strategy(strategy0)
self.reduce_sum = P.ReduceSum(keep_dims=False).set_strategy(strategy1)
def construct(self, x, y, bias):
def construct(self, x, y):
out = self.fc_nobias(x, y)
out = self.reduce_sum(out, (0, 1))
return out
@ -105,5 +103,4 @@ def test_sum_as_loss_int32():
x = Tensor(np.ones([64, 32]), dtype=ms.int32)
y = Tensor(np.ones([64, 32]), dtype=ms.int32)
bias = Tensor(np.ones([64]), dtype=ms.int32)
compile_net(net, x, y, bias)
compile_net(net, x, y)

12
tests/ut/python/parallel/test_onehot.py
Unescape View File

@ -104,7 +104,11 @@ def test_onehot_batch_parallel_invalid_strategy():
strategy4 = ((16, 1), (16, 1))
try:
compile_graph(strategy1, strategy2, strategy3, strategy4)
except BaseException:
except ValueError:
pass
except TypeError:
pass
except RuntimeError:
pass
@ -144,7 +148,11 @@ def test_onehot_batch_parallel_invalid_strategy_axis0():
strategy4 = ((16, 1), (16, 1))
try:
compile_graph(strategy1, strategy2, strategy3, strategy4, onthot_axis=0)
except BaseException:
except ValueError:
pass
except TypeError:
pass
except RuntimeError:
pass

87
tests/ut/python/parallel/test_reduce_method_info.py
Unescape View File

@ -24,6 +24,17 @@ from mindspore.ops import operations as P
from tests.ut.python.ops.test_math_ops import VirtualLoss
class NetWithLossNoBias(nn.Cell):
def __init__(self, network):
super(NetWithLossNoBias, self).__init__()
self.loss = VirtualLoss()
self.network = network
def construct(self, x, y):
predict = self.network(x, y)
return self.loss(predict)
class NetWithLoss(nn.Cell):
def __init__(self, network):
super(NetWithLoss, self).__init__()
@ -35,6 +46,15 @@ class NetWithLoss(nn.Cell):
return self.loss(predict)
class GradWrapNoBias(nn.Cell):
def __init__(self, network):
super(GradWrapNoBias, self).__init__()
self.network = network
def construct(self, x, y):
return C.grad_all(self.network)(x, y)
class GradWrap(nn.Cell):
def __init__(self, network):
super(GradWrap, self).__init__()
@ -44,6 +64,11 @@ class GradWrap(nn.Cell):
return C.grad_all(self.network)(x, y, b)
def compile_net_no_bias(net, x, y):
net.set_auto_parallel()
_executor.compile(net, x, y)
def compile_net(net, x, y, b):
net.set_auto_parallel()
_executor.compile(net, x, y, b)
@ -165,7 +190,7 @@ def test_sum_mul5():
self.mul1 = P.Mul().set_strategy(strategy1)
self.reduce_sum = P.ReduceSum(keep_dims=True).set_strategy(strategy2)
def construct(self, x, y, b):
def construct(self, x, y):
out = self.mul1(x, y)
out = self.reduce_sum(out, 0)
return out
@ -173,13 +198,12 @@ def test_sum_mul5():
context.set_auto_parallel_context(device_num=64, global_rank=0)
strategy1 = ((1, 8, 8), (1, 8, 8))
strategy2 = ((2, 4, 1),)
net = GradWrap(NetWithLoss(Net(strategy1, strategy2)))
net = GradWrapNoBias(NetWithLossNoBias(Net(strategy1, strategy2)))
context.set_auto_parallel_context(parallel_mode="semi_auto_parallel")
x = Tensor(np.ones([128, 32, 64]), dtype=ms.float32)
y = Tensor(np.ones([128, 32, 64]), dtype=ms.float32)
b = Tensor(np.ones([1, 32, 64]), dtype=ms.float32)
compile_net(net, x, y, b)
compile_net_no_bias(net, x, y)
def test_sum_mul6():
@ -189,7 +213,7 @@ def test_sum_mul6():
self.mul1 = P.Mul().set_strategy(strategy1)
self.reduce_sum = P.ReduceSum(keep_dims=True).set_strategy(strategy2)
def construct(self, x, y, b):
def construct(self, x, y):
out = self.mul1(x, y)
out = self.reduce_sum(out, 1)
return out
@ -197,13 +221,12 @@ def test_sum_mul6():
context.set_auto_parallel_context(device_num=64, global_rank=0)
strategy1 = ((1, 8, 8), (1, 8, 8))
strategy2 = ((2, 1, 4),)
net = GradWrap(NetWithLoss(Net(strategy1, strategy2)))
net = GradWrapNoBias(NetWithLossNoBias(Net(strategy1, strategy2)))
context.set_auto_parallel_context(parallel_mode="semi_auto_parallel")
x = Tensor(np.ones([128, 32, 64]), dtype=ms.float32)
y = Tensor(np.ones([128, 32, 64]), dtype=ms.float32)
b = Tensor(np.ones([128, 1, 64]), dtype=ms.float32)
compile_net(net, x, y, b)
compile_net_no_bias(net, x, y)
def test_sum_mul7():
@ -213,7 +236,7 @@ def test_sum_mul7():
self.mul1 = P.Mul().set_strategy(strategy1)
self.reduce_sum = P.ReduceSum(keep_dims=True).set_strategy(strategy2)
def construct(self, x, y, b):
def construct(self, x, y):
out = self.mul1(x, y)
out = self.reduce_sum(out, (0, 1))
return out
@ -221,13 +244,12 @@ def test_sum_mul7():
context.set_auto_parallel_context(device_num=64, global_rank=0)
strategy1 = ((1, 8, 8), (1, 8, 8))
strategy2 = ((2, 4, 1),)
net = GradWrap(NetWithLoss(Net(strategy1, strategy2)))
net = GradWrapNoBias(NetWithLossNoBias(Net(strategy1, strategy2)))
context.set_auto_parallel_context(parallel_mode="semi_auto_parallel")
x = Tensor(np.ones([128, 32, 64]), dtype=ms.float32)
y = Tensor(np.ones([128, 32, 64]), dtype=ms.float32)
b = Tensor(np.ones([1, 64]), dtype=ms.float32)
compile_net(net, x, y, b)
compile_net_no_bias(net, x, y)
def test_max_mul():
@ -347,6 +369,12 @@ def gen_inputs_and_compile_net(net):
compile_net(net, x, y, b)
def gen_inputs_and_compile_net_no_bias(net):
x = Tensor(np.ones([128, 64, 64]), dtype=ms.float32)
y = Tensor(np.ones([128, 64, 64]), dtype=ms.float32)
compile_net_no_bias(net, x, y)
def tobefixed_test_arg_max_with_value_mul_semi_axis_parallel():
context.set_auto_parallel_context(device_num=8, global_rank=0)
strategy1 = ((1, 4, 2), (1, 4, 2))
@ -414,7 +442,7 @@ class ArgMinWithValueNet2(nn.Cell):
self.arg_min_with_value = P.ArgMinWithValue(keep_dims=True, axis=-1).set_strategy(strategy2)
self.relu = P.ReLU().set_strategy(strategy3)
def construct(self, x, y, b):
def construct(self, x, y):
out = self.mul1(x, y)
_, out = self.arg_min_with_value(out)
out = self.relu(out)
@ -426,9 +454,9 @@ def tobefixed_test_arg_min_with_value_mul_semi_axis_parallel2():
strategy1 = ((1, 4, 2), (1, 4, 2))
strategy2 = ((4, 1, 2),)
strategy3 = ((2, 4, 1),)
net = GradWrap(NetWithLoss(ArgMinWithValueNet2(strategy1, strategy2, strategy3)))
net = GradWrapNoBias(NetWithLossNoBias(ArgMinWithValueNet2(strategy1, strategy2, strategy3)))
context.set_auto_parallel_context(parallel_mode="semi_auto_parallel")
gen_inputs_and_compile_net(net)
gen_inputs_and_compile_net_no_bias(net)
def test_arg_min_with_value_mul_semi2():
@ -436,9 +464,9 @@ def test_arg_min_with_value_mul_semi2():
strategy1 = ((1, 4, 2), (1, 4, 2))
strategy2 = ((4, 1, 1),)
strategy3 = ((2, 4, 1),)
net = GradWrap(NetWithLoss(ArgMinWithValueNet2(strategy1, strategy2, strategy3)))
net = GradWrapNoBias(NetWithLossNoBias(ArgMinWithValueNet2(strategy1, strategy2, strategy3)))
context.set_auto_parallel_context(parallel_mode="semi_auto_parallel")
gen_inputs_and_compile_net(net)
gen_inputs_and_compile_net_no_bias(net)
def test_arg_min_with_value_mul_auto2():
@ -446,9 +474,9 @@ def test_arg_min_with_value_mul_auto2():
strategy1 = None
strategy2 = None
strategy3 = None
net = GradWrap(NetWithLoss(ArgMinWithValueNet2(strategy1, strategy2, strategy3)))
net = GradWrapNoBias(NetWithLossNoBias(ArgMinWithValueNet2(strategy1, strategy2, strategy3)))
context.set_auto_parallel_context(parallel_mode="auto_parallel")
gen_inputs_and_compile_net(net)
gen_inputs_and_compile_net_no_bias(net)
def test_cross_batch():
@ -459,7 +487,7 @@ def test_cross_batch():
self.reduce_sum = P.ReduceSum(keep_dims=False).set_strategy(strategy2)
self.reduce_mean = P.ReduceMean(keep_dims=False).set_strategy(strategy3).add_prim_attr("cross_batch", True)
def construct(self, x, y, b):
def construct(self, x, y):
out = self.mul1(x, y)
out = self.reduce_sum(out, -1)
out = self.reduce_mean(out, 0)
@ -469,13 +497,12 @@ def test_cross_batch():
strategy1 = ((4, 2), (4, 2))
strategy2 = ((2, 1),)
strategy3 = ((8,),)
net = GradWrap(NetWithLoss(Net(strategy1, strategy2, strategy3)))
net = GradWrapNoBias(NetWithLossNoBias(Net(strategy1, strategy2, strategy3)))
context.set_auto_parallel_context(parallel_mode="semi_auto_parallel")
x = Tensor(np.ones([32, 64]), dtype=ms.float32)
y = Tensor(np.ones([32, 64]), dtype=ms.float32)
b = Tensor(np.ones([32, 64]), dtype=ms.float32)
compile_net(net, x, y, b)
compile_net_no_bias(net, x, y)
def test_cross_batch2():
@ -486,7 +513,7 @@ def test_cross_batch2():
self.reduce_mean = P.ReduceMean(keep_dims=False).set_strategy(strategy2)
self.reduce_sum = P.ReduceSum(keep_dims=False).set_strategy(strategy3).add_prim_attr("cross_batch", True)
def construct(self, x, y, b):
def construct(self, x, y):
out = self.mul1(x, y)
out = self.reduce_mean(out, -1)
out = self.reduce_sum(out, 0)
@ -496,13 +523,12 @@ def test_cross_batch2():
strategy1 = ((4, 2), (4, 2))
strategy2 = ((2, 1),)
strategy3 = ((8,),)
net = GradWrap(NetWithLoss(Net(strategy1, strategy2, strategy3)))
net = GradWrapNoBias(NetWithLossNoBias(Net(strategy1, strategy2, strategy3)))
context.set_auto_parallel_context(parallel_mode="semi_auto_parallel")
x = Tensor(np.ones([32, 64]), dtype=ms.float32)
y = Tensor(np.ones([32, 64]), dtype=ms.float32)
b = Tensor(np.ones([32, 64]), dtype=ms.float32)
compile_net(net, x, y, b)
compile_net_no_bias(net, x, y)
def test_cross_batch_auto():
@ -513,20 +539,19 @@ def test_cross_batch_auto():
self.reduce_mean = P.ReduceMean(keep_dims=False)
self.reduce_sum = P.ReduceSum(keep_dims=False).add_prim_attr("cross_batch", True)
def construct(self, x, y, b):
def construct(self, x, y):
out = self.mul1(x, y)
out = self.reduce_mean(out, -1)
out = self.reduce_sum(out, 0)
return out
context.set_auto_parallel_context(device_num=8, global_rank=0)
net = GradWrap(NetWithLoss(Net()))
net = GradWrapNoBias(NetWithLossNoBias(Net()))
context.set_auto_parallel_context(parallel_mode="auto_parallel")
x = Tensor(np.ones([32, 64]), dtype=ms.float32)
y = Tensor(np.ones([32, 64]), dtype=ms.float32)
b = Tensor(np.ones([32, 64]), dtype=ms.float32)
compile_net(net, x, y, b)
compile_net_no_bias(net, x, y)
def test_max_empty_tuple():

24
tests/ut/python/parallel/test_reshape.py
Unescape View File

@ -114,7 +114,11 @@ def test_reshape1_strategy_1():
strategy_loss = ((8, 1), (8, 1))
try:
reshape_common(ParallelMode.SEMI_AUTO_PARALLEL, strategy0, strategy1, strategy2, strategy_loss)
except BaseException:
except ValueError:
pass
except TypeError:
pass
except RuntimeError:
pass
@ -125,7 +129,11 @@ def test_reshape1_strategy_2():
strategy_loss = ((8, 1), (8, 1))
try:
reshape_common(ParallelMode.AUTO_PARALLEL, strategy0, strategy1, strategy2, strategy_loss)
except BaseException:
except ValueError:
pass
except TypeError:
pass
except RuntimeError:
pass
@ -347,14 +355,22 @@ def test_reshape_net3_2():
def test_reshape_net4_1():
try:
reshape_net2(ReshapeNet4(((1, 8), (8, 1))))
except BaseException:
except ValueError:
pass
except TypeError:
pass
except RuntimeError:
pass
def test_reshape_net4_2():
try:
reshape_net2(ReshapeNet4(((1, 8), (8, 2))))
except BaseException:
except ValueError:
pass
except TypeError:
pass
except RuntimeError:
pass

9
tests/ut/python/parallel/test_scalar_loss.py
Unescape View File

@ -29,8 +29,8 @@ class GradWrap(nn.Cell):
super(GradWrap, self).__init__()
self.network = network
def construct(self, x, y, bias):
return C.grad_all(self.network)(x, y, bias)
def construct(self, x, y):
return C.grad_all(self.network)(x, y)
def test_sum_as_loss():
@ -41,7 +41,7 @@ def test_sum_as_loss():
self.reduce_sum = P.ReduceSum(keep_dims=False).set_strategy(strategy1)
self.mul = P.Mul().set_strategy(strategy=((), ()))
def construct(self, x, y, bias):
def construct(self, x, y):
out = self.fc_nobias(x, y)
out = self.reduce_sum(out, (0, 1))
out = self.mul(out, F.scalar_to_array(2.0))
@ -57,5 +57,4 @@ def test_sum_as_loss():
x = Tensor(np.ones([64, 32]), dtype=ms.float32)
y = Tensor(np.ones([64, 32]), dtype=ms.float32)
bias = Tensor(np.ones([64]), dtype=ms.float32)
_executor.compile(net, x, y, bias)
_executor.compile(net, x, y)

17
tests/ut/python/parallel/test_split_grad_sens.py
Unescape View File

@ -52,11 +52,21 @@ class GradWrap3(nn.Cell):
def construct(self, x, y, bias):
return C.grad_all(self.network)(x, y, bias)
class GradWrap4(nn.Cell):
def __init__(self, network):
super(GradWrap4, self).__init__()
self.network = network
def construct(self, x, y):
return C.grad_all(self.network)(x, y)
def compile_net(net, x, y, b):
net.set_auto_parallel()
_executor.compile(net, x, y, b)
def compile_net_no_bias(net, x, y):
net.set_auto_parallel()
_executor.compile(net, x, y)
def test_no_grad():
class Net(nn.Cell):
@ -144,7 +154,7 @@ def test_grad_sens_scalar_broadcast():
self.fc_nobias = P.MatMul(transpose_b=True).set_strategy(strategy0)
self.reduce_sum = P.ReduceSum(keep_dims=False).set_strategy(strategy1)
def construct(self, x, y, bias):
def construct(self, x, y):
out = self.fc_nobias(x, y)
out = self.reduce_sum(out, (0, 1))
return out
@ -152,10 +162,9 @@ def test_grad_sens_scalar_broadcast():
context.set_auto_parallel_context(device_num=16, global_rank=0)
strategy0 = ((4, 1), (4, 1))
strategy1 = ((4, 1),)
net = GradWrap3(Net(strategy0, strategy1))
net = GradWrap4(Net(strategy0, strategy1))
context.set_auto_parallel_context(parallel_mode="semi_auto_parallel")
x = Tensor(np.ones([64, 32]), dtype=ms.float32)
y = Tensor(np.ones([64, 32]), dtype=ms.float32)
bias = Tensor(np.ones([64]), dtype=ms.float32)
compile_net(net, x, y, bias)
compile_net_no_bias(net, x, y)

18
tests/ut/python/parallel/test_sum_as_loss.py
Unescape View File

@ -28,13 +28,13 @@ class GradWrap(nn.Cell):
super(GradWrap, self).__init__()
self.network = network
def construct(self, x, y, bias):
return C.grad_all(self.network)(x, y, bias)
def construct(self, x, y):
return C.grad_all(self.network)(x, y)
def compile_net(net, x, y, bias):
def compile_net(net, x, y):
net.set_auto_parallel()
_executor.compile(net, x, y, bias)
_executor.compile(net, x, y)
def test_sum_as_loss():
@ -44,7 +44,7 @@ def test_sum_as_loss():
self.fc_nobias = P.MatMul(transpose_b=True).set_strategy(strategy0)
self.reduce_sum = P.ReduceSum(keep_dims=False).set_strategy(strategy1)
def construct(self, x, y, bias):
def construct(self, x, y):
out = self.fc_nobias(x, y)
out = self.reduce_sum(out, (0, 1))
return out
@ -57,8 +57,7 @@ def test_sum_as_loss():
x = Tensor(np.ones([64, 32]), dtype=ms.float32)
y = Tensor(np.ones([64, 32]), dtype=ms.float32)
bias = Tensor(np.ones([64]), dtype=ms.float32)
compile_net(net, x, y, bias)
compile_net(net, x, y)
def test_sum_as_loss2():
@ -68,7 +67,7 @@ def test_sum_as_loss2():
self.fc_nobias = P.MatMul(transpose_b=True).set_strategy(strategy0)
self.reduce_sum = P.ReduceSum(keep_dims=False).set_strategy(strategy1)
def construct(self, x, y, bias):
def construct(self, x, y):
out = self.fc_nobias(x, y)
out = self.reduce_sum(out, (0, 1))
return out
@ -81,5 +80,4 @@ def test_sum_as_loss2():
x = Tensor(np.ones([64, 32]), dtype=ms.float32)
y = Tensor(np.ones([64, 32]), dtype=ms.float32)
bias = Tensor(np.ones([64]), dtype=ms.float32)
compile_net(net, x, y, bias)
compile_net(net, x, y)

Write Preview
Loading…
Cancel
Save