forked from mindspore-Ecosystem/mindspore
add switch_simplify pass to a2
This commit is contained in:
chenfei
parent
0768df5144
commit
ff066a0df3
|
|
@ -285,6 +285,7 @@ OptPassGroupMap GetOptPassesA(const opt::irpass::OptimizeIRPassLib &irpass) {
|
|||
opt::OptPassConfig a_1 = GetOptPassA1(irpass);
|
||||
opt::OptPassConfig a_2 = opt::OptPassConfig(
|
||||
{
|
||||
irpass.switch_simplify_,
|
||||
irpass.cast_eliminate_,
|
||||
irpass.specialize_transform_,
|
||||
irpass.merge_addn_,
|
||||
|
|
|
|||
|
|
@ -25,7 +25,6 @@ from mindspore.train.model import Model
|
|||
from mindspore.ops.composite import GradOperation
|
||||
from mindspore.common import ParameterTuple
|
||||
|
||||
|
||||
context.set_context(mode=context.GRAPH_MODE)
|
||||
|
||||
|
||||
|
|
@ -87,11 +86,11 @@ def _count_unequal_element(data_expected, data_me, rtol, atol):
|
|||
assert data_expected.shape == data_me.shape
|
||||
total_count = len(data_expected.flatten())
|
||||
error = np.abs(data_expected - data_me)
|
||||
greater = np.greater(error, atol + np.abs(data_me)*rtol)
|
||||
greater = np.greater(error, atol + np.abs(data_me) * rtol)
|
||||
loss_count = np.count_nonzero(greater)
|
||||
assert (loss_count/total_count) < rtol, \
|
||||
"\ndata_expected_std:{0}\ndata_me_error:{1}\nloss:{2}".\
|
||||
format(data_expected[greater], data_me[greater], error[greater])
|
||||
assert (loss_count / total_count) < rtol, \
|
||||
"\ndata_expected_std:{0}\ndata_me_error:{1}\nloss:{2}". \
|
||||
format(data_expected[greater], data_me[greater], error[greater])
|
||||
|
||||
|
||||
def allclose_nparray(data_expected, data_me, rtol, atol, equal_nan=True):
|
||||
|
|
@ -115,10 +114,10 @@ class ControlGraphSupportNotEqual(Cell):
|
|||
else:
|
||||
out2 = input_data / input_data
|
||||
if x == z:
|
||||
out3_f = (lambda a: a+a)
|
||||
out3_f = (lambda a: a + a)
|
||||
out3 = out3_f(input_data)
|
||||
else:
|
||||
out3_f = (lambda a: a+a+a)
|
||||
out3_f = (lambda a: a + a + a)
|
||||
out3 = out3_f(input_data)
|
||||
return out, out2, out3
|
||||
|
||||
|
|
@ -175,15 +174,15 @@ class ControlBprop(Cell):
|
|||
else:
|
||||
out2 = input_data / input_data
|
||||
if x == z:
|
||||
out3_f = (lambda a: a+a)
|
||||
out3_f = (lambda a: a + a)
|
||||
out3 = out3_f(input_data)
|
||||
else:
|
||||
out3_f = (lambda a: a+a+a)
|
||||
out3_f = (lambda a: a + a + a)
|
||||
out3 = out3_f(input_data)
|
||||
return out, out2, out3
|
||||
|
||||
def bprop(self, x, y, z, input_data, out, dout):
|
||||
return x*2, y*3, z, input_data*5.1
|
||||
return x * 2, y * 3, z, input_data * 5.1
|
||||
|
||||
|
||||
@pytest.mark.level1
|
||||
|
|
@ -199,10 +198,10 @@ def test_ctrl_if_while_bprop_true():
|
|||
grad_net = GradOfAllInputs(net, sens_param=False)
|
||||
grad_net.set_train()
|
||||
grads = grad_net(Tensor(x), Tensor(y), Tensor(x), Tensor(input_data))
|
||||
allclose_nparray(x*2, grads[0].asnumpy(), 0.0000, 0.0000)
|
||||
allclose_nparray(y*3, grads[1].asnumpy(), 0.0000, 0.0000)
|
||||
allclose_nparray(x * 2, grads[0].asnumpy(), 0.0000, 0.0000)
|
||||
allclose_nparray(y * 3, grads[1].asnumpy(), 0.0000, 0.0000)
|
||||
allclose_nparray(x, grads[2].asnumpy(), 0.0000, 0.0000)
|
||||
allclose_nparray(input_data*5.1, grads[3].asnumpy(), 0.0000, 0.0000)
|
||||
allclose_nparray(input_data * 5.1, grads[3].asnumpy(), 0.0000, 0.0000)
|
||||
|
||||
|
||||
class TwoInput(Cell):
|
||||
|
|
@ -234,7 +233,7 @@ class InlineBpropTwoInput1(Cell):
|
|||
grads = self.grad(x, y)
|
||||
else:
|
||||
grads = self.grad(x, y)
|
||||
return grads[0]*2, grads[1]*2
|
||||
return grads[0] * 2, grads[1] * 2
|
||||
|
||||
|
||||
@pytest.mark.level1
|
||||
|
|
@ -248,8 +247,8 @@ def test_ctrl_if_while_bprop_inlinebprop_twoinput():
|
|||
grad_net = GradOfAllInputs(net, sens_param=False)
|
||||
grad_net.set_train()
|
||||
grads = grad_net(input1, input2)
|
||||
allclose_nparray(input1.asnumpy()*2, grads[1].asnumpy(), 0, 0)
|
||||
allclose_nparray(input2.asnumpy()*2, grads[0].asnumpy(), 0, 0)
|
||||
allclose_nparray(input1.asnumpy() * 2, grads[1].asnumpy(), 0, 0)
|
||||
allclose_nparray(input2.asnumpy() * 2, grads[0].asnumpy(), 0, 0)
|
||||
|
||||
|
||||
class ControlOneIfOneParaOneAddn(Cell):
|
||||
|
|
@ -467,7 +466,7 @@ class SideEffectPrintInHighOrdeAddnNet(Cell):
|
|||
@pytest.mark.platform_x86_ascend_training
|
||||
@pytest.mark.env_onecard
|
||||
def test_side_effect_high_order_print_in_high_order_net():
|
||||
print_file = os.getcwd()+"/test_side_effect_high_order_print_in_high_order_net.data"
|
||||
print_file = os.getcwd() + "/test_side_effect_high_order_print_in_high_order_net.data"
|
||||
context.set_context(print_file_path=print_file)
|
||||
net = SideEffectPrintInHighOrdeAddnNet()
|
||||
out1 = net(Tensor([9.0], ms.float32))
|
||||
|
|
@ -617,7 +616,7 @@ class HighGrad(Cell):
|
|||
def __init__(self, network, grad_list, sens_param=False, real_inputs_count=None):
|
||||
super().__init__()
|
||||
self.grads = [network]
|
||||
for i in range(len(grad_list)-1):
|
||||
for i in range(len(grad_list) - 1):
|
||||
_grad = grad_list[i](self.grads[i], sens_param=False)
|
||||
self.grads.append(_grad)
|
||||
self.final_grad = grad_list[-1](self.grads[-1],
|
||||
|
|
@ -676,10 +675,10 @@ class SideEffectControlFlowAssignDependWhileNet(Cell):
|
|||
return grad_out
|
||||
|
||||
|
||||
@pytest.mark.level1
|
||||
# Now the case can't pass because the GPU RT problem, so only run on Ascend current time.
|
||||
@pytest.mark.level0
|
||||
@pytest.mark.platform_arm_ascend_training
|
||||
@pytest.mark.platform_x86_gpu_training
|
||||
@pytest.mark.platform_x86_cpu
|
||||
@pytest.mark.platform_x86_ascend_training
|
||||
@pytest.mark.env_onecard
|
||||
def test_side_effect_grad_control_flow_assign_depend_while_net():
|
||||
context.set_context(mode=context.GRAPH_MODE)
|
||||
|
|
|
|||
Loading…
Reference in New Issue