def erf_generic(x):
select = P.Select()
less = P.Less()
abs_cal = P.Abs()
return select(less(abs_cal(x), 1), erf_f32_generic(x),
1 - erfc_f32_generic(x))
def __init__(self,
scale=1.0,
shift=0.0,
name='ScalarAffine'):
"""
Constructor of ScalarAffine Bijector.
"""
param = dict(locals())
validator.check_value_type(
'scale', scale, [int, float], type(self).__name__)
validator.check_value_type(
'shift', shift, [int, float], type(self).__name__)
super(ScalarAffine, self).__init__(
is_constant_jacobian=True,
is_injective=True,
name=name,
dtype=None,
param=param)
self._scale = cast_to_tensor(scale)
self._shift = cast_to_tensor(shift)
self.abs = P.Abs()
self.oneslike = P.OnesLike()
self.dtypeop = P.DType()
self.cast = P.Cast()
self.log = log_generic
def erfc_generic(x):
select = P.Select()
greater = P.Greater()
abs_cal = P.Abs()
return select(greater(abs_cal(x), 1), erfc_f32_generic(x),
1 - erf_f32_generic(x))
def __init__(self, sharpness=1.0, name='Softplus'):
"""
Constructor of Softplus Bijector.
"""
param = dict(locals())
param['param_dict'] = {'sharpness': sharpness}
super(Softplus, self).__init__(name=name, dtype=None, param=param)
self._sharpness = self._add_parameter(sharpness, 'sharpness')
self.exp = exp_generic
self.log = log_generic
self.expm1 = P.Expm1()
self.abs = P.Abs()
self.dtypeop = P.DType()
self.cast = P.Cast()
self.fill = P.Fill()
self.greater = P.Greater()
self.less = P.Less()
self.log_sigmoid = LogSigmoid()
self.logicalor = P.LogicalOr()
self.select = P.Select()
self.shape = P.Shape()
self.sigmoid = P.Sigmoid()
self.softplus = self._softplus
self.inverse_softplus = self._inverse_softplus
self.threshold = np.log(np.finfo(np.float32).eps) + 1
self.tiny = np.exp(self.threshold)
def erfc_f32_generic(x):
"""
Calculate erfc for dtype of f32
"""
k_maxlog = 88.72283905206835
k_erfc_pcoefficient = [
+2.326819970068386e-2, -1.387039388740657e-1, +3.687424674597105e-1,
-5.824733027278666e-1, +6.210004621745983e-1, -4.944515323274145e-1,
+3.404879937665872e-1, -2.741127028184656e-1, +5.638259427386472e-1
]
k_erfc_rcoefficient = [
-1.047766399936249e+1, +1.297719955372516e+1, -7.495518717768503e+0,
+2.921019019210786e+0, -1.015265279202700e+0, +4.218463358204948e-1,
-2.820767439740514e-1, +5.641895067754075e-1
]
abs_cal = P.Abs()
select = P.Select()
less = P.Less()
fill = P.Fill()
dtype = P.DType()
shape = P.Shape()
abs_x = abs_cal(x)
z = exp_generic(-x * x)
q = 1 / abs_x
y = q * q
poly1 = _evaluate_polynomial(y, k_erfc_pcoefficient)
poly2 = _evaluate_polynomial(y, k_erfc_rcoefficient)
p = select(less(abs_x, 2.0), poly1, poly2)
y = z * q * p
zeros = fill(dtype(x), shape(x), 0)
y_clamp = select(less(z, -k_maxlog), zeros, y)
return select(less(x, 0), 2.0 - y_clamp, y_clamp)
def __init__(self, sharpness=1.0, name='Softplus'):
"""
Constructor of Softplus Bijector.
"""
param = dict(locals())
validator.check_value_type('sharpness', sharpness, [int, float],
type(self).__name__)
super(Softplus, self).__init__(name=name, param=param)
self._sharpness = cast_to_tensor(sharpness)
self.exp = exp_generic
self.log = log_generic
self.expm1 = expm1_generic
self.abs = P.Abs()
self.dtypeop = P.DType()
self.fill = P.Fill()
self.greater = P.Greater()
self.less = P.Less()
self.log_sigmoid = LogSigmoid()
self.logicalor = P.LogicalOr()
self.select = P.Select()
self.shape = P.Shape()
self.sigmoid = P.Sigmoid()
self.softplus = self._softplus
self.inverse_softplus = self._inverse_softplus
self.threshold = np.log(np.finfo(np.float32).eps) + 1
self.tiny = np.exp(self.threshold)
def __init__(self):
super(DiGamma, self).__init__()
# const numbers
self.k_lanczos_gamma = 7
self.k_base_lanczos_coeff = 0.99999999999980993227684700473478
self.k_lanczos_coefficients = [676.520368121885098567009190444019,
-1259.13921672240287047156078755283,
771.3234287776530788486528258894,
-176.61502916214059906584551354,
12.507343278686904814458936853,
-0.13857109526572011689554707,
9.984369578019570859563e-6,
1.50563273514931155834e-7]
self.nan = np.nan
self.pi = np.pi
self.lanczos_gamma_plus_one_half = self.k_lanczos_gamma + 0.5
self.log_lanczos_gamma_plus_one_half = np.log(self.lanczos_gamma_plus_one_half)
# operations
self.log1p = P.Log1p()
self.abs = P.Abs()
self.shape = P.Shape()
self.dtype = P.DType()
self.fill = P.Fill()
self.floor = P.Floor()
self.equal = P.Equal()
self.less = P.Less()
self.select = P.Select()
self.sin = P.Sin()
self.cos = P.Cos()
self.logicaland = P.LogicalAnd()
def _abs_max(gradients):
"""
Transform gradients to saliency through abs then take max along
channels.
"""
gradients = op.Abs()(gradients)
saliency = op.ReduceMax(keep_dims=True)(gradients, axis=1)
return saliency
def erfc_f64_generic(x):
"""
Calculate erfc for dtype of f64
"""
k_maxlog = 7.09782712893383996843e2
k_erfc_pcoefficient = [
2.46196981473530512524e-10, 5.64189564831068821977e-1,
7.46321056442269912687e0, 4.86371970985681366614e1,
1.96520832956077098242e2, 5.26445194995477358631e2,
9.34528527171957607540e2, 1.02755188689515710272e3,
5.57535335369399327526e2
]
k_erfc_qcoefficient = [
1.00000000000000000000e0, 1.32281951154744992508e1,
8.67072140885989742329e1, 3.54937778887819891062e2,
9.75708501743205489753e2, 1.82390916687909736289e3,
2.24633760818710981792e3, 1.65666309194161350182e3,
5.57535340817727675546e2
]
k_erfc_rcoefficient = [
5.64189583547755073984e-1, 1.27536670759978104416e0,
5.01905042251180477414e0, 6.16021097993053585195e0,
7.40974269950448939160e0, 2.97886665372100240670e0
]
k_erfc_scoefficient = [
1.00000000000000000000e0, 2.26052863220117276590e0,
9.39603524938001434673e0, 1.20489539808096656605e1,
1.70814450747565897222e1, 9.60896809063285878198e0,
3.36907645100081516050e02
]
abs_cal = P.Abs()
select = P.Select()
less = P.Less()
fill = P.Fill()
dtype = P.DType()
shape = P.Shape()
abs_x = abs_cal(x)
z = -x * x
exp_z = exp_generic(z)
temp1 = exp_z * _evaluate_polynomial(
abs_x, k_erfc_pcoefficient) / _evaluate_polynomial(
abs_x, k_erfc_qcoefficient)
temp2 = exp_z * _evaluate_polynomial(
abs_x, k_erfc_rcoefficient) / _evaluate_polynomial(
abs_x, k_erfc_scoefficient)
y = select(less(abs_x, 8.0), temp1, temp2)
zeros = fill(dtype(x), shape(x), 0)
y_clamp = select(less(z, k_maxlog), zeros, y)
poly2 = _evaluate_polynomial(y, k_erfc_rcoefficient)
p = select(less(abs_x, 2.0), poly1, poly2)
y = z * q * p
zeros = fill(dtype(x), shape(x), 0)
y_clamp = select(less(z, -k_maxlog), zeros, y)
return select(less(x, 0), 2.0 - y_clamp, y_clamp)
def __init__(self, scale):
super(L1Regularizer, self).__init__()
Validator.check_value_type("scale", scale, [int, float], self.cls_name)
if scale <= 0:
raise ValueError("scale should be a number which greater than 0")
if math.isinf(scale) or math.isnan(scale):
raise ValueError("scale can not be INF or NAN")
self.abs = P.Abs()
self.reduce_sum = P.ReduceSum()
self.scale = Tensor(scale, dtype=mstype.float32)
def construct(self, x):
"""
Construct method.
"""
output_hm, output_wh, output_off, output_kps = self.centerface_network(x)
output_hm_nms, _ = self.maxpool2d(output_hm)
abs_error = P.Abs()(output_hm - output_hm_nms)
abs_out = P.Abs()(output_hm)
error = abs_error / (abs_out + 1e-12)
# cannot use P.Equal()(output_hm, output_hm_nms), since maxpooling output has 0.1% error
keep = P.Select()(P.LessEqual()(error, 1e-3), \
P.Fill()(ms.float32, P.Shape()(error), 1.0), \
P.Fill()(ms.float32, P.Shape()(error), 0.0))
output_hm = output_hm * keep
# get topK and index
scores = self.reshape(output_hm, (self.test_batch, -1))
topk_scores, topk_inds = self.topk(scores, self.k)
return topk_scores, output_wh, output_off, output_kps, topk_inds
def abs_max(gradients):
"""
Transform gradients to saliency through abs then take max along channels.
Args:
gradients (_Tensor): Gradients which will be transformed to saliency map.
Returns:
_Tensor, saliency map integrated from gradients.
"""
gradients = op.Abs()(gradients)
saliency = op.ReduceMax(keep_dims=True)(gradients, axis=1)
return saliency
def __init__(self, scale=1.0, shift=0.0, name='ScalarAffine'):
"""
Constructor of scalar affine bijector.
"""
param = dict(locals())
validator.check_value_type('scale', scale, [int, float], name)
validator.check_value_type('shift', shift, [int, float], name)
self._scale = cast_to_tensor(scale)
self._shift = cast_to_tensor(shift)
super(ScalarAffine, self).__init__(is_constant_jacobian=True,
is_injective=True,
name=name,
dtype=None,
param=param)
self.abs = P.Abs()
self.log = log_by_step
self.checktensor = CheckTensor()
def __init__(self,
num_bits=2,
compute_type=mstype.float32,
clip_value=1.0,
per_channel=False):
self.num_bits = num_bits
self.compute_type = compute_type
self.clip_value = clip_value
self.per_channel = per_channel
self.clamp = C.clip_by_value
self.abs = P.Abs()
self.sum = P.ReduceSum()
self.nelement = F.size
self.div = P.Div()
self.cast = P.Cast()
self.max = P.ReduceMax()
self.min = P.ReduceMin()
self.floor = P.Floor()
def __init__(self,
num_bits=8,
compute_type=mstype.float32,
clip_value=1.0,
per_channel=False):
super(QuantizeWeightCell, self).__init__()
self.num_bits = num_bits
self.compute_type = compute_type
self.clip_value = clip_value
self.per_channel = per_channel
self.clamp = C.clip_by_value
self.abs = P.Abs()
self.sum = P.ReduceSum()
self.nelement = F.size
self.div = P.Div()
self.cast = P.Cast()
self.max = P.ReduceMax()
self.min = P.ReduceMin()
self.round = P.Round()
def __init__(self, scale=1.0, shift=0.0, name='ScalarAffine'):
"""
Constructor of ScalarAffine Bijector.
"""
param = dict(locals())
param['param_dict'] = {'scale': scale, 'shift': shift}
super(ScalarAffine, self).__init__(is_constant_jacobian=True,
is_injective=True,
name=name,
dtype=None,
param=param)
self._scale = self._add_parameter(scale, 'scale')
self._shift = self._add_parameter(shift, 'shift')
self.abs = P.Abs()
self.oneslike = P.OnesLike()
self.dtypeop = P.DType()
self.cast = P.Cast()
self.log = log_generic
def __init__(self,reduction='mean'):
super().__init__()
self.abs = P.Abs()
self.squeeze = P.Squeeze(-1)
def tensor_abs(inputs):
"""Apply abs function."""
return P.Abs()(inputs)
def _IgammacContinuedFraction(ax, x, a, enabled):
"""Helper function for computing Igammac using a continued fraction."""
abs_x = P.Abs()
logicaland = P.LogicalAnd()
greater = P.Greater()
less = P.Less()
notequal = P.NotEqual()
fill = P.Fill()
shape = P.Shape()
dtype = P.DType()
select = P.Select()
if dtype(ax) == mstype.float16:
epsilon = eps_fp16
else:
epsilon = eps_fp32
def cond(vals):
enabled = vals[0]
c = vals[5]
return logicaland(less(c, 2000), enabled)
def body(vals):
enabled = vals[0]
ans = vals[1]
t = vals[2]
y = vals[3]
z = vals[4]
c = vals[5]
pkm1 = vals[6]
qkm1 = vals[7]
pkm2 = vals[8]
qkm2 = vals[9]
dpkm2_da = vals[10]
dqkm2_da = vals[11]
dpkm1_da = vals[12]
dqkm1_da = vals[13]
dans_da = vals[14]
c = c + 1
y = y + 1
z = z + 2
yc = y * c
pk = pkm1 * z - pkm2 * yc
qk = qkm1 * z - qkm2 * yc
qk_is_nonzero = notequal(qk, 0)
r = pk / qk
t = select(qk_is_nonzero, abs_x((ans - r) / r), fill(dtype(t), shape(t), 1))
ans = select(qk_is_nonzero, r, ans)
dpk_da = dpkm1_da * z - pkm1 - dpkm2_da * yc + pkm2 * c
dqk_da = dqkm1_da * z - qkm1 - dqkm2_da * yc + qkm2 * c
dans_da_new = select(qk_is_nonzero, (dpk_da - ans * dqk_da) / qk, dans_da)
grad_conditional = select(qk_is_nonzero,
abs_x(dans_da_new - dans_da),
fill(dtype(dans_da), shape(dans_da), 1))
pkm2 = pkm1
pkm1 = pk
qkm2 = qkm1
qkm1 = qk
dpkm2_da = dpkm1_da
dqkm2_da = dqkm1_da
dpkm1_da = dpk_da
dqkm1_da = dqk_da
rescale = greater(abs_x(pk), 1 / epsilon)
pkm2 = select(rescale, pkm2 * epsilon, pkm2)
pkm1 = select(rescale, pkm1 * epsilon, pkm1)
qkm2 = select(rescale, qkm2 * epsilon, qkm2)
qkm1 = select(rescale, qkm1 * epsilon, qkm1)
dpkm2_da = select(rescale, dpkm2_da * epsilon, dpkm2_da)
dqkm2_da = select(rescale, dqkm2_da * epsilon, dqkm2_da)
dpkm1_da = select(rescale, dpkm1_da * epsilon, dpkm1_da)
dqkm1_da = select(rescale, dqkm1_da * epsilon, dqkm1_da)
conditional = logicaland(enabled, greater(grad_conditional, epsilon))
return (conditional, select(enabled, ans, vals[1]), select(enabled, t, vals[2]),
select(enabled, y, vals[3]), select(enabled, z, vals[4]),
c, select(enabled, pkm1, vals[6]),
select(enabled, qkm1, vals[7]), select(enabled, pkm2, vals[8]),
select(enabled, qkm2, vals[9]), select(enabled, dpkm2_da, vals[10]),
select(enabled, dqkm2_da, vals[11]), select(enabled, dpkm1_da, vals[12]),
select(enabled, dqkm1_da, vals[13]), select(enabled, dans_da_new, vals[14]))
y = 1 - a
z = x + y + 1
c = fill(dtype(x), shape(x), 0)
pkm2 = fill(dtype(x), shape(x), 1)
qkm2 = x
pkm1 = x + 1
qkm1 = z * x
ans = pkm1 / qkm1
t = fill(dtype(x), shape(x), 1)
dpkm2_da = fill(dtype(x), shape(x), 0)
dqkm2_da = fill(dtype(x), shape(x), 0)
dpkm1_da = fill(dtype(x), shape(x), 0)
dqkm1_da = -x
dans_da = (dpkm1_da - ans * dqkm1_da) / qkm1
vals = (enabled, ans, t, y, z, c, pkm1, qkm1, pkm2, qkm2, dpkm2_da, dqkm2_da, dpkm1_da, dqkm1_da, dans_da)
vals = _while_helper_func(cond, body, vals)
ans = vals[1]
return ans * ax
def __init__(self,
vocab_size,
embedding_size,
field_size,
param_init='normal',
target='CPU',
slice_mode='batch_slice',
feature_num_list=None,
max_norm=None,
sparse=True,
operator='SUM'):
super(MultiFieldEmbeddingLookup,
self).__init__(vocab_size, embedding_size, param_init, target,
slice_mode, feature_num_list, max_norm, sparse)
self.field_size = validator.check_positive_int(field_size,
'field_size')
self.operator = operator
self.mul = P.Mul()
self.inf_mask_mul = P.Mul()
self.bias_add = P.Add()
self.inf_add = P.Add()
self.merge_op = None
self.count_op = P.UnsortedSegmentSum()
self.abs = P.Abs()
self.equal = P.Equal()
self.add = P.Add()
self.cast = P.Cast()
self.div_no_nan = P.DivNoNan()
self.expand = P.ExpandDims()
self.max_mask_mul = P.Mul()
self.max_no_equal = P.NotEqual()
if operator == MultiFieldEmbeddingLookup.OPERATOR_SUM:
self.merge_op = P.UnsortedSegmentSum()
elif operator == MultiFieldEmbeddingLookup.OPERATOR_MAX:
self.merge_op = P.UnsortedSegmentMax()
elif operator == MultiFieldEmbeddingLookup.OPERATOR_MEAN:
self.merge_op = P.UnsortedSegmentSum()
else:
raise ValueError(
"The operator supports ['SUM', 'MAX', 'MEAN'], but found: " +
str(operator))
parallel_mode = _get_parallel_mode()
is_auto_parallel = parallel_mode in (ParallelMode.SEMI_AUTO_PARALLEL,
ParallelMode.AUTO_PARALLEL)
if slice_mode in ["table_row_slice", "batch_slice"
] and is_auto_parallel:
self.merge_op.shard(
((get_group_size(), 1, 1), (get_group_size(), 1)))
self.expand.shard(((get_group_size(), ), ))
self.bias_add.shard(((1, 1), (1, 1)))
self.mul.shard(
((get_group_size(), 1, 1), (get_group_size(), 1, 1)))
self.count_op.shard(((get_group_size(), 1), (get_group_size(), 1)))
self.add.shard(((get_group_size(), ), (get_group_size(), )))
self.div_no_nan.shard(
((get_group_size(), 1), (get_group_size(), 1)))
self.max_mask_mul.shard(
((get_group_size(), 1), (get_group_size(), 1)))
self.max_no_equal.shard(((1, ), ()))
if operator == MultiFieldEmbeddingLookup.OPERATOR_MAX:
self.equal.shard(((get_group_size(), 1, 1), ()))
self.inf_mask_mul.shard(((get_group_size(), 1, 1), ()))
self.merge_op.shard(
((get_group_size(), 1), (get_group_size(), )))
self.count_op.shard(
((get_group_size(), ), (get_group_size(), )))
self.inf_add.shard(
((get_group_size(), 1, 1), (get_group_size(), 1, 1)))
elif slice_mode == "table_column_slice" and is_auto_parallel:
self.merge_op.shard(((1, 1, get_group_size()), (1, 1)))
self.div_no_nan.shard(((1, get_group_size()), (1, 1)))
self.bias_add.shard(((1, 1), (1, 1)))
self.mul.shard(((1, 1, 1), (1, 1, get_group_size())))
self.count_op.shard(((1, 1), (1, 1)))
self.add.shard(((1, ), (1, )))
self.max_mask_mul.shard(((1, get_group_size()), (1, 1)))
self.expand.shard(((1, ), ))
self.max_no_equal.shard(((1, ), ()))
if operator == MultiFieldEmbeddingLookup.OPERATOR_MAX:
self.equal.shard(((1, 1, 1), ()))
self.inf_mask_mul.shard(((1, 1, 1), ()))
self.merge_op.shard(((1, get_group_size()), (1, )))
self.count_op.shard(((1, ), (1, )))
self.inf_add.shard(((1, 1, get_group_size()), (1, 1, 1)))
else:
if is_auto_parallel:
raise ValueError(
"slice_mode should be ['table_row_slice', 'batch_slice' and \
'table_column_slice'], but get " + str(slice_mode))
# Min value for fp32
self.negative_inf_value = -3.402823466E+38
def __init__(self, reduction='mean'):
super(CustomLoss, self).__init__(reduction)
self.abs = P.Abs()
def __init__(self, reduction='mean', weights=1.0):
super(WeightedLoss, self).__init__(reduction)
self.abs = P.Abs()
self.weights = weights
def __init__(self,
in_channels,
out_channels,
weight_init='normal',
bias_init='zeros',
has_bias=True,
activation=None):
super(Dense_Thor, self).__init__()
self.thor = True
self.in_channels = Validator.check_positive_int(in_channels)
self.out_channels = Validator.check_positive_int(out_channels)
self.has_bias = Validator.check_bool(has_bias)
if isinstance(weight_init, Tensor):
if weight_init.dim() != 2 or weight_init.shape[0] != out_channels or \
weight_init.shape[1] != in_channels:
raise ValueError("Weight init shape error.")
self.weight = Parameter(initializer(weight_init,
[out_channels, in_channels]),
name="weight")
self.bias = None
if self.has_bias:
if isinstance(bias_init, Tensor):
if bias_init.dim() != 1 or bias_init.shape[0] != out_channels:
raise ValueError("Bias init shape error.")
self.bias = Parameter(initializer(bias_init, [out_channels]),
name="bias")
self.bias_add = P.BiasAdd()
self.matmul = P.MatMul(transpose_b=True)
self.activation = get_activation(activation)
self.activation_flag = self.activation is not None
self.matrix_A = Parameter(Tensor(
np.zeros([in_channels, in_channels]).astype(np.float32)),
name='matrix_A',
requires_grad=False)
self.shape = P.Shape()
self.reshape = P.Reshape()
self.transpose = P.Transpose()
self.mul = P.Mul()
self.is_Ascend = True
if context.get_context("device_target") == "Ascend":
if out_channels == 1001:
self.matrix_G = Parameter(Tensor(
np.zeros([1024, 1024]).astype(np.float32)),
name='matrix_G',
requires_grad=False)
self.pad = P.Pad(((0, 23), (0, 23)))
self.pad1 = P.Pad(((0, 7), (0, 7)))
self.slice = P.Slice()
self.add = P.TensorAdd()
else:
self.matrix_G = Parameter(Tensor(
np.eye(out_channels).astype(np.float32)),
name="matrix_G",
requires_grad=False)
self.abs = P.Abs()
self.reduce_max = P.ReduceMax(keep_dims=False)
self.neg = P.Neg()
self.reduce_sum = P.ReduceSum()
self.matmul = P.MatMul(transpose_b=True)
self.cube_matmul = P.CusMatMulCube(transpose_a=True)
self.cast = P.Cast()
self.is_nsp_layer = (out_channels == 2)
else:
self.is_Ascend = False
self.matrix_G = Parameter(Tensor(
np.eye(out_channels).astype(np.float32)),
name="matrix_G",
requires_grad=False)
self.cube_matmul = P.MatMul(transpose_a=True)
self.getG = P.InsertGradientOf(self.save_gradient)
def __init__(self,
in_channels,
out_channels,
weight_init='normal',
bias_init='zeros',
damping=0.03,
loss_scale=1,
frequency=100,
has_bias=False,
activation=None,
batch_size=12):
super(Dense_Thor, self).__init__()
self.in_channels = Validator.check_positive_int(in_channels)
self.out_channels = Validator.check_positive_int(out_channels)
self.has_bias = Validator.check_bool(has_bias)
self.thor = True
if isinstance(weight_init, Tensor):
if weight_init.dim() != 2 or weight_init.shape()[0] != out_channels or \
weight_init.shape()[1] != in_channels:
raise ValueError("weight_init shape error")
self.weight = Parameter(initializer(weight_init,
[out_channels, in_channels]),
name="weight")
if self.has_bias:
if isinstance(bias_init, Tensor):
if bias_init.dim() != 1 or bias_init.shape(
)[0] != out_channels:
raise ValueError("bias_init shape error")
self.bias = Parameter(initializer(bias_init, [out_channels]),
name="bias")
self.matmul = P.MatMul(transpose_b=True)
self.bias_add = P.BiasAdd()
self.activation = get_activation(activation)
self.activation_flag = self.activation is not None
self.matrix_A_inv = Parameter(Tensor(
np.zeros([in_channels, in_channels]).astype(np.float16)),
name='matrix_A_inv',
requires_grad=False)
self.matrix_G_inv = Parameter(Tensor(
np.zeros([out_channels, out_channels]).astype(np.float16)),
name="matrix_G_inv",
requires_grad=False)
self.fake_G = Tensor(
np.zeros([out_channels, out_channels]).astype(np.float16))
self.matmul = P.MatMul(transpose_b=True)
self.cube_matmul = P.CusMatMulCube(transpose_a=True)
self.matrix_combine = P.CusMatrixCombine()
self.cholesky = P.CusCholeskyTrsm()
self.shape = P.Shape()
self.reshape = P.Reshape()
self.transpose = P.Transpose()
self.cov_step = Parameter(initializer(0, [1], mstype.int32),
name="cov_step",
requires_grad=False)
self.mul = P.Mul()
self.cast = P.Cast()
self.damping = damping
self.loss_scale = Tensor(1 / loss_scale, mstype.float16)
self.vector_matmul = P.CusBatchMatMul()
self.gather = P.GatherV2()
self.assignadd = P.AssignAdd()
self.freq = Tensor(frequency, mstype.int32)
self.axis = 0
self.abs = P.Abs()
self.reduce_max = P.ReduceMax(keep_dims=False)
self.log = P.Log()
self.exp = P.Exp()
self.dampingA = Tensor(np.identity(in_channels), mstype.float32)
self.dampingG = Tensor(np.identity(out_channels), mstype.float32)
self.sqrt = P.Sqrt()
self.getG = P.InsertGradientOf(self.save_gradient)
self.batch_size = batch_size
'block': P.NPUClearFloatStatus(),
'desc_inputs': [Tensor(np.zeros([8]).astype(np.float32))],
'desc_bprop': [Tensor(np.zeros([8]).astype(np.float32))],
'skip': ['backward']}),
('CheckValid', {
'block': P.CheckValid(),
'desc_inputs': [[20000, 4], [3]],
'desc_bprop': [[20000]],
'skip': ['backward']}),
('NMSWithMask', {
'block': P.NMSWithMask(0.5),
'desc_inputs': [[128, 5]],
'desc_bprop': [[128, 5], [128], [128]],
'skip': ['backward']}),
('Abs', {
'block': P.Abs(),
'desc_inputs': [[4]],
'desc_bprop': [[4]]}),
('CumSum', {
'block': P.CumSum(),
'desc_const': [0],
'desc_inputs': [Tensor(np.array([[3, 4],[1, 6]]).astype(np.float16))],
'desc_bprop': [Tensor(np.array([[3, 4],[4, 10]]).astype(np.float16))]}),
('ReduceSum_3', {
'block': P.ReduceSum(),
'desc_const': [0],
'desc_inputs': [[3, 2]],
'desc_bprop': [[2]]}),
('ReduceSum_4', {
'block': P.ReduceSum(keep_dims=True),
'desc_const': [0],
def __init__(self):
super(Net, self).__init__()
self.ops = P.Abs()
def absolute(inputs: Tensor) -> Tensor:
"""Get the absolute value of a tensor value."""
abs_op = op.Abs()
outputs = abs_op(inputs)
return outputs
'desc_inputs': [Tensor(np.ones([3, 4]).astype(np.int32))],
'skip': ['backward']}),
# dims is not 2
('NMSWithMask2', {
'block': (P.NMSWithMask(), {'exception': ValueError, 'error_keywords': ['NMSWithMask']}),
'desc_inputs': [Tensor(np.ones([3, 4, 2]).astype(np.float32))],
'skip': ['backward']}),
# shape[1] is not 5
('NMSWithMask3', {
'block': (P.NMSWithMask(), {'exception': ValueError, 'error_keywords': ['NMSWithMask']}),
'desc_inputs': [Tensor(np.ones([3, 2]).astype(np.float32))],
'skip': ['backward']}),
# input is not tensor
('Abs0', {
'block': (P.Abs(), {'exception': TypeError, 'error_keywords': ['Abs']}),
'desc_inputs': [5.0],
'skip': ['backward']}),
# input is Tensor(bool)
('Abs1', {
'block': (P.Abs(), {'exception': TypeError, 'error_keywords': ['Abs']}),
'desc_inputs': [Tensor(np.ones([3, 4]).astype(np.bool_))],
'skip': ['backward']}),
# input is not tensor
('Sign0', {
'block': (P.Sign(), {'exception': TypeError, 'error_keywords': ['Sign']}),
'desc_inputs': [5.0],
'skip': ['backward']}),
# input is Tensor(bool)
('Sign1', {
def setup_method(self):
"""Setup the test case."""
self.net = SimpleLinear()
self.relu = P.ReLU()
self.abs_ = P.Abs()
self.reshape = P.Reshape()
def __init__(self, strategy1, strategy2):
super().__init__()
self.matmul = P.MatMul().set_strategy(strategy1)
self.abs = P.Abs().set_strategy(strategy2)
self.matmul2 = P.MatMul().set_strategy(strategy1)
