From b1189cd1181d1e020f79bbca4e2b6007b6e27715 Mon Sep 17 00:00:00 2001 From: caozhou Date: Mon, 30 Nov 2020 21:18:16 +0800 Subject: [PATCH] compatible name --- mindspore/common/parameter.py | 35 ++++++++---- mindspore/nn/cell.py | 3 + .../official/cv/centerface/src/mobile_v2.py | 4 +- .../cv/efficientnet/src/efficientnet.py | 4 +- .../cv/faster_rcnn/src/FasterRcnn/rcnn.py | 5 +- .../cv/maskrcnn/src/maskrcnn/rcnn_cls.py | 5 +- model_zoo/official/cv/resnet_thor/src/thor.py | 4 +- .../official/cv/resnet_thor/src/thor_layer.py | 55 ++++++++----------- model_zoo/official/cv/warpctc/src/loss.py | 5 +- model_zoo/official/cv/warpctc/src/warpctc.py | 12 ++-- model_zoo/official/gnn/bgcf/src/bgcf.py | 53 ++++++++---------- model_zoo/official/gnn/gat/src/gat.py | 6 +- model_zoo/official/nlp/bert/pretrain_eval.py | 4 +- model_zoo/official/nlp/bert/src/CRF.py | 2 +- .../nlp/bert/src/bert_for_finetune.py | 13 ++--- .../nlp/bert/src/bert_for_pre_training.py | 19 +++---- model_zoo/official/nlp/bert/src/bert_model.py | 12 ++-- .../official/nlp/bert_thor/pretrain_eval.py | 4 +- .../bert_thor/src/bert_for_pre_training.py | 6 +- .../official/nlp/bert_thor/src/bert_model.py | 9 +-- .../nlp/bert_thor/src/fused_layer_norm.py | 4 +- .../nlp/bert_thor/src/thor_for_bert.py | 4 +- .../nlp/bert_thor/src/thor_for_bert_arg.py | 4 +- .../official/nlp/bert_thor/src/thor_layer.py | 21 +++---- .../nlp/gnmt_v2/src/gnmt_model/attention.py | 7 +-- .../nlp/gnmt_v2/src/gnmt_model/dynamic_rnn.py | 4 +- .../nlp/gnmt_v2/src/gnmt_model/embedding.py | 3 +- .../gnmt_v2/src/gnmt_model/gnmt_for_train.py | 3 +- .../nlp/gnmt_v2/src/utils/optimizer.py | 6 +- model_zoo/official/nlp/gpt/src/gpt.py | 11 ++-- .../official/nlp/gpt/src/gpt_wrapcell.py | 3 +- .../nlp/mass/src/transformer/embedding.py | 3 +- .../src/transformer/transformer_for_train.py | 3 +- .../prophetnet/src/transformer/embedding.py | 3 +- .../nlp/tinybert/src/tinybert_for_gd_td.py | 6 +- .../nlp/tinybert/src/tinybert_model.py | 12 ++-- .../transformer/src/transformer_for_train.py | 3 +- .../nlp/transformer/src/transformer_model.py | 3 +- model_zoo/official/recommend/ncf/src/ncf.py | 4 +- .../research/cv/ghostnet_quant/src/quant.py | 6 +- .../cv/ssd_ghostnet/src/ssd_ghostnet.py | 4 +- model_zoo/research/cv/tinynet/src/tinynet.py | 4 +- model_zoo/research/nlp/dscnn/src/ds_cnn.py | 4 +- 43 files changed, 174 insertions(+), 211 deletions(-) diff --git a/mindspore/common/parameter.py b/mindspore/common/parameter.py index 4456782378..e93ec1d81d 100644 --- a/mindspore/common/parameter.py +++ b/mindspore/common/parameter.py @@ -30,6 +30,7 @@ __all__ = ['Parameter', 'ParameterTuple'] PARAMETER_NAME_DEFAULT = "Parameter" PARAMETER_NAME_PREFIX_MAX_LEN = 1024 + def _is_in_parallel_mode(): """Get parallel mode.""" return auto_parallel_context().get_parallel_mode() in ["semi_auto_parallel", "auto_parallel"] @@ -51,10 +52,12 @@ class Parameter(MetaTensor_): A Parameter has to belong to a Cell. If there is an operator in the network that requires part of the inputs to be Parameter, then the Parameters as this part of the inputs are not allowed to be cast. + It is recommended to use the default value of `name` when initialize a parameter as one attribute of a cell, + otherwise, the parameter name may be different than expected. Args: default_input (Union[Tensor, MetaTensor, Number]): Parameter data, to be set initialized. - name (str): Name of the child parameter. + name (str): Name of the child parameter. Default: None. requires_grad (bool): True if the parameter requires gradient. Default: True. layerwise_parallel (bool): A kind of model parallel mode. When layerwise_parallel is true in parallel mode, broadcast and gradients communication would not be applied to parameters. Default: False. @@ -72,7 +75,7 @@ class Parameter(MetaTensor_): >>> def __init__(self): >>> super(Net, self).__init__() >>> self.matmul = P.MatMul() - >>> self.weight = Parameter(Tensor(np.ones((1,2))), name="w", requires_grad=True) + >>> self.weight = Parameter(Tensor(np.ones((1,2))), requires_grad=True) >>> >>> def construct(self, x): >>> out = self.matmul(self.weight, x) @@ -88,7 +91,7 @@ class Parameter(MetaTensor_): """ __base_type__ = {} - def __new__(cls, default_input, name, *args, **kwargs): + def __new__(cls, default_input, *args, **kwargs): input_class, *class_init_args = Parameter._get_parameter_new_args(default_input) new_type = Parameter._get_base_class(input_class) obj = input_class.__new__(new_type) @@ -112,7 +115,7 @@ class Parameter(MetaTensor_): return ( Parameter, (data, self.name, self.requires_grad, self.layerwise_parallel)) - def __init__(self, default_input, name, requires_grad=True, layerwise_parallel=False): + def __init__(self, default_input, name=None, requires_grad=True, layerwise_parallel=False): self._param_info = ParamInfo() self.name = name self.requires_grad = requires_grad @@ -276,24 +279,20 @@ class Parameter(MetaTensor_): """ self._is_init = is_init_ - def clone(self, prefix, init='same'): + def clone(self, init='same'): """ Clone the parameter. Args: - prefix (str): Namespace of parameter. The cloned Parameter name is - combined of prefix and current name: `f"{perfix}.{self.name}"`. init (Union[Tensor, str, MetaTensor, numbers.Number]): Initialize the shape of the parameter. Default: 'same'. Returns: Parameter, a new parameter. """ - Validator.check_str_by_regular(prefix) x = copy(self) # pylint: disable=protected-access x._param_info = self._param_info.clone() - x._param_info.name = prefix + '.' + self._param_info.name x.is_init = False x.is_param_ps = self.is_param_ps x.init_in_server = self.init_in_server @@ -464,10 +463,25 @@ class ParameterTuple(tuple): def __new__(cls, iterable): """Create instance object of ParameterTuple.""" data = tuple(iterable) + ids = set() + orders = {} for x in data: if not isinstance(x, Parameter): raise TypeError(f"ParameterTuple input should be `Parameter` collection." f"But got a {type(iterable)}, {iterable}") + if id(x) not in ids: + ids.add(id(x)) + if x.name not in orders.keys(): + orders[x.name] = [0, x] + else: + if isinstance(orders[x.name], list): + name = x.name + orders[name][1].name = name + "_" + str(0) + x.name = x.name + "_" + str(1) + orders[name] = 1 + else: + orders[x.name] += 1 + x.name = x.name + "_" + str(orders[x.name]) return tuple.__new__(ParameterTuple, tuple(data)) def clone(self, prefix, init='same'): @@ -484,7 +498,8 @@ class ParameterTuple(tuple): Validator.check_str_by_regular(prefix) new = [] for x in self: - x1 = x.clone(prefix, init) + x1 = x.clone(init) + x1.name = prefix + "." + x1.name new.append(x1) return ParameterTuple(new) diff --git a/mindspore/nn/cell.py b/mindspore/nn/cell.py index 221ceb6b66..7cc9b77cb0 100755 --- a/mindspore/nn/cell.py +++ b/mindspore/nn/cell.py @@ -20,6 +20,7 @@ import os from collections import OrderedDict import numpy from mindspore import log as logger +from mindspore.common.parameter import PARAMETER_NAME_DEFAULT from .. import context from ..common import dtype as mstype from ..common.api import _executor, _pynative_exec @@ -619,6 +620,8 @@ class Cell(Cell_): raise KeyError("Duplicated parameter name '{}'.".format(param_name)) if not isinstance(param, Parameter) and param is not None: raise TypeError("The type of parameter should be 'Parameter' if not None.") + if isinstance(param, Parameter) and param.name == PARAMETER_NAME_DEFAULT: + param.name = param_name self._params[param_name] = param def cast_param(self, param): diff --git a/model_zoo/official/cv/centerface/src/mobile_v2.py b/model_zoo/official/cv/centerface/src/mobile_v2.py index 6c0cd269f8..9345c554e5 100644 --- a/model_zoo/official/cv/centerface/src/mobile_v2.py +++ b/model_zoo/official/cv/centerface/src/mobile_v2.py @@ -55,11 +55,11 @@ class DepthWiseConv(nn.Cell): self.bias_add = P.BiasAdd() weight_shape = [channel_multiplier, in_planes, kernel_size, kernel_size] - self.weight = Parameter(initializer(KaimingNormal(mode='fan_out'), weight_shape), name='weight') + self.weight = Parameter(initializer(KaimingNormal(mode='fan_out'), weight_shape)) if has_bias: bias_shape = [channel_multiplier * in_planes] - self.bias = Parameter(initializer('zeros', bias_shape), name='bias') + self.bias = Parameter(initializer('zeros', bias_shape)) else: self.bias = None diff --git a/model_zoo/official/cv/efficientnet/src/efficientnet.py b/model_zoo/official/cv/efficientnet/src/efficientnet.py index f8985a513a..01bfbf9f4d 100644 --- a/model_zoo/official/cv/efficientnet/src/efficientnet.py +++ b/model_zoo/official/cv/efficientnet/src/efficientnet.py @@ -469,12 +469,12 @@ class DepthWiseConv(nn.Cell): self.depthwise_conv = P.Conv2D(out_channel=in_planes * 1, kernel_size=kernel_size, stride=stride, pad_mode="same", group=in_planes) self.weight = Parameter(initializer( - weight_init, [in_planes * 1, 1, kernel_size, kernel_size]), name='depthwise_weight') + weight_init, [in_planes * 1, 1, kernel_size, kernel_size])) else: self.depthwise_conv = P.DepthwiseConv2dNative( channel_multiplier=1, kernel_size=kernel_size, stride=stride, pad_mode='same',) self.weight = Parameter(initializer( - weight_init, [1, in_planes, kernel_size, kernel_size]), name='depthwise_weight') + weight_init, [1, in_planes, kernel_size, kernel_size])) def construct(self, x): x = self.depthwise_conv(x, self.weight) diff --git a/model_zoo/official/cv/faster_rcnn/src/FasterRcnn/rcnn.py b/model_zoo/official/cv/faster_rcnn/src/FasterRcnn/rcnn.py index 152d37f7b8..4c9e389b95 100644 --- a/model_zoo/official/cv/faster_rcnn/src/FasterRcnn/rcnn.py +++ b/model_zoo/official/cv/faster_rcnn/src/FasterRcnn/rcnn.py @@ -28,9 +28,8 @@ class DenseNoTranpose(nn.Cell): def __init__(self, input_channels, output_channels, weight_init): super(DenseNoTranpose, self).__init__() - self.weight = Parameter(initializer(weight_init, [input_channels, output_channels], mstype.float16), - name="weight") - self.bias = Parameter(initializer("zeros", [output_channels], mstype.float16).to_tensor(), name="bias") + self.weight = Parameter(initializer(weight_init, [input_channels, output_channels], mstype.float16)) + self.bias = Parameter(initializer("zeros", [output_channels], mstype.float16).to_tensor()) self.matmul = P.MatMul(transpose_b=False) self.bias_add = P.BiasAdd() diff --git a/model_zoo/official/cv/maskrcnn/src/maskrcnn/rcnn_cls.py b/model_zoo/official/cv/maskrcnn/src/maskrcnn/rcnn_cls.py index e529f2e22b..d1ef7525dc 100644 --- a/model_zoo/official/cv/maskrcnn/src/maskrcnn/rcnn_cls.py +++ b/model_zoo/official/cv/maskrcnn/src/maskrcnn/rcnn_cls.py @@ -26,9 +26,8 @@ class DenseNoTranpose(nn.Cell): """Dense method""" def __init__(self, input_channels, output_channels, weight_init): super(DenseNoTranpose, self).__init__() - self.weight = Parameter(initializer(weight_init, [input_channels, output_channels], mstype.float16), - name="weight") - self.bias = Parameter(initializer("zeros", [output_channels], mstype.float16).to_tensor(), name="bias") + self.weight = Parameter(initializer(weight_init, [input_channels, output_channels], mstype.float16)) + self.bias = Parameter(initializer("zeros", [output_channels], mstype.float16).to_tensor()) self.matmul = P.MatMul(transpose_b=False) self.bias_add = P.BiasAdd() diff --git a/model_zoo/official/cv/resnet_thor/src/thor.py b/model_zoo/official/cv/resnet_thor/src/thor.py index 6c9c8f914f..3fffd2c709 100644 --- a/model_zoo/official/cv/resnet_thor/src/thor.py +++ b/model_zoo/official/cv/resnet_thor/src/thor.py @@ -55,7 +55,7 @@ class THOR_GPU(Optimizer): Validator.check_value_type("momentum", momentum, [float], self.cls_name) if isinstance(momentum, float) and momentum < 0.0: raise ValueError("momentum should be at least 0.0, but got momentum {}".format(momentum)) - self.momentum = Parameter(Tensor(momentum, mstype.float32), name="momentum") + self.momentum = Parameter(Tensor(momentum, mstype.float32)) self.params = self.parameters self.use_nesterov = Validator.check_bool(use_nesterov) self.moments = self.params.clone(prefix="moments", init='zeros') @@ -160,7 +160,7 @@ class THOR(Optimizer): super(THOR, self).__init__(learning_rate, params, weight_decay, loss_scale) if isinstance(momentum, float) and momentum < 0.0: raise ValueError("momentum should be at least 0.0, but got momentum {}".format(momentum)) - self.momentum = Parameter(Tensor(momentum, mstype.float32), name="momentum") + self.momentum = Parameter(Tensor(momentum, mstype.float32)) self.params = self.parameters self.moments = self.params.clone(prefix="moments", init='zeros') self.hyper_map = C.HyperMap() diff --git a/model_zoo/official/cv/resnet_thor/src/thor_layer.py b/model_zoo/official/cv/resnet_thor/src/thor_layer.py index 1826ba08c5..27e90bfe0e 100644 --- a/model_zoo/official/cv/resnet_thor/src/thor_layer.py +++ b/model_zoo/official/cv/resnet_thor/src/thor_layer.py @@ -109,11 +109,10 @@ class _Conv(Cell): 'attr \'group\' of \'Conv2D\' Op.') self.weight = Parameter(initializer( - weight_init, [out_channels, in_channels // group, *kernel_size]), name='weight') + weight_init, [out_channels, in_channels // group, *kernel_size])) if Validator.check_bool(has_bias): - self.bias = Parameter(_initializer( - bias_init, [out_channels]), name='bias') + self.bias = Parameter(initializer(bias_init, [out_channels])) else: if bias_init != 'zeros': logger.warning("Value of 'has_bias' is False, value of 'bias_init' will be ignored.") @@ -174,12 +173,10 @@ class Conv2d_Thor_GPU(_Conv): split_dim = 128 matrix_A_shape, matrix_G_shape = caculate_matmul_shape(self.matrix_A_dim, self.matrix_G_dim, split_dim) - self.matrix_A_inv = Parameter(np.zeros(matrix_A_shape).astype(np.float32), - name='matrix_A_inv', requires_grad=False) - self.matrix_G_inv = Parameter(np.zeros(matrix_G_shape).astype(np.float32), - name='matrix_A_inv', requires_grad=False) + self.matrix_A_inv = Parameter(np.zeros(matrix_A_shape).astype(np.float32), requires_grad=False) + self.matrix_G_inv = Parameter(np.zeros(matrix_G_shape).astype(np.float32), requires_grad=False) self.broadcast_to = P.BroadcastTo(matrix_A_shape) - self.cov_step = Parameter(initializer(0, [1], mstype.int32), name="cov_step", requires_grad=False) + self.cov_step = Parameter(initializer(0, [1], mstype.int32), requires_grad=False) self.img2col = P.Im2Col(kernel_size=kernel_size, stride=stride, pad_mode="same") self.matmul = P.MatMul(transpose_b=True) self.shape = P.Shape() @@ -195,7 +192,7 @@ class Conv2d_Thor_GPU(_Conv): self.axis = 0 self.sqrt = P.Sqrt() self.reduce_mean = P.ReduceMean(keep_dims=False) - self.damping = Parameter(Tensor(damping), name="damping_value", requires_grad=False) + self.damping = Parameter(Tensor(damping), requires_grad=False) self.dampingA = Tensor(np.identity(self.matrix_A_dim), mstype.float32) self.dampingG = Tensor(np.identity(self.matrix_G_dim), mstype.float32) self.cholesky = P.CholeskyTrsm(split_dim=split_dim) @@ -301,14 +298,14 @@ class Dense_Thor_GPU(Cell): 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.weight = Parameter(initializer(weight_init, [out_channels, in_channels])) 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 = Parameter(initializer(bias_init, [out_channels])) self.matmul = P.MatMul(transpose_b=True) self.bias_add = P.BiasAdd() @@ -317,12 +314,10 @@ class Dense_Thor_GPU(Cell): self.activation_flag = self.activation is not None split_dim = 128 matrix_A_shape, matrix_G_shape = caculate_matmul_shape(self.in_channels, self.out_channels, split_dim) - self.matrix_A_inv = Parameter(Tensor(np.zeros(matrix_A_shape).astype(np.float32)), - name='matrix_A_inv', requires_grad=False) - self.matrix_G_inv = Parameter(Tensor(np.zeros(matrix_G_shape).astype(np.float32)), - name="matrix_G_inv", requires_grad=False) + self.matrix_A_inv = Parameter(Tensor(np.zeros(matrix_A_shape).astype(np.float32)), requires_grad=False) + self.matrix_G_inv = Parameter(Tensor(np.zeros(matrix_G_shape).astype(np.float32)), requires_grad=False) self.broadcast_to = P.BroadcastTo(matrix_A_shape) - self.cov_step = Parameter(initializer(0, [1], mstype.int32), name="cov_step", requires_grad=False) + self.cov_step = Parameter(initializer(0, [1], mstype.int32), requires_grad=False) self.shape = P.Shape() self.reshape = P.Reshape() self.transpose = P.Transpose() @@ -331,7 +326,7 @@ class Dense_Thor_GPU(Cell): self.loss_scale = Tensor(1 / loss_scale, mstype.float16) self.batch_size = Tensor(batch_size, mstype.float16) self.getG = P.InsertGradientOf(self.save_gradient) - self.damping = Parameter(Tensor(damping), name="damping_value", requires_grad=False) + self.damping = Parameter(Tensor(damping), requires_grad=False) self.dampingA = Tensor(np.identity(in_channels), mstype.float32) self.dampingG = Tensor(np.identity(out_channels), mstype.float32) self.cast = P.Cast() @@ -467,20 +462,20 @@ class Conv2d_Thor(_Conv): self.matrix_G_device_shape[3]) self.matrix_A_inv = Parameter( Tensor(np.reshape(np.identity(self.matrix_A_device_dim).astype(np.float16), self.matrix_A_device_shape)), - name='matrix_A_inv', requires_grad=False) - self.A_inv_max = Parameter(initializer(0, [1], mstype.float32), name="A_inv_max", requires_grad=False) + requires_grad=False) + self.A_inv_max = Parameter(initializer(0, [1], mstype.float32), requires_grad=False) self.matrix_G_inv = Parameter( Tensor(np.reshape(np.identity(self.matrix_G_device_dim).astype(np.float16), self.matrix_G_device_shape)), - name="matrix_G_inv", requires_grad=False) + requires_grad=False) - self.G_inv_max = Parameter(initializer(0, [1], mstype.float32), name="G_inv_max", requires_grad=False) + self.G_inv_max = Parameter(initializer(0, [1], mstype.float32), requires_grad=False) self.fake_G = Tensor( np.reshape(np.identity(self.matrix_G_device_dim).astype(np.float16), self.matrix_G_device_shape)) 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.cov_step = Parameter(initializer(0, [1], mstype.int32), requires_grad=False) self.mul = P.Mul() self.cast = P.Cast() self.damping = Tensor(damping) @@ -648,14 +643,14 @@ class Dense_Thor(Cell): 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.weight = Parameter(initializer(weight_init, [out_channels, in_channels])) 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 = Parameter(initializer(bias_init, [out_channels])) self.matmul = P.MatMul(transpose_b=True) self.bias_add = P.BiasAdd() @@ -663,10 +658,8 @@ class Dense_Thor(Cell): self.activation = get_activation(activation) self.activation_flag = self.activation is not None - self.matrix_A_inv = Parameter(Tensor(np.zeros([128, 128, 16, 16]).astype(np.float16)), name='matrix_A_inv', - requires_grad=False) - self.matrix_G_inv = Parameter(Tensor(np.zeros([63, 63, 16, 16]).astype(np.float16)), name="matrix_G_inv", - requires_grad=False) + self.matrix_A_inv = Parameter(Tensor(np.zeros([128, 128, 16, 16]).astype(np.float16)), requires_grad=False) + self.matrix_G_inv = Parameter(Tensor(np.zeros([63, 63, 16, 16]).astype(np.float16)), requires_grad=False) self.fake_G = Tensor(np.zeros([63, 63, 16, 16]).astype(np.float16)) self.matmul = P.MatMul(transpose_b=True) @@ -676,7 +669,7 @@ class Dense_Thor(Cell): 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.cov_step = Parameter(initializer(0, [1], mstype.int32), requires_grad=False) self.mul = P.Mul() self.cast = P.Cast() self.damping = Tensor(damping) @@ -689,8 +682,8 @@ class Dense_Thor(Cell): self.assignadd = P.AssignAdd() self.freq = Tensor(frequency, mstype.int32) self.axis = 0 - self.A_inv_max = Parameter(initializer(0, [1], mstype.float32), name="A_inv_max", requires_grad=False) - self.G_inv_max = Parameter(initializer(0, [1], mstype.float32), name="G_inv_max", requires_grad=False) + self.A_inv_max = Parameter(initializer(0, [1], mstype.float32), requires_grad=False) + self.G_inv_max = Parameter(initializer(0, [1], mstype.float32), requires_grad=False) self.fused_abs_max1 = P.CusFusedAbsMax1([1001, 1001]) self.fused_abs_max2 = P.CusFusedAbsMax1() self.log = P.Log() diff --git a/model_zoo/official/cv/warpctc/src/loss.py b/model_zoo/official/cv/warpctc/src/loss.py index 8ea4c20e94..4624f506a9 100755 --- a/model_zoo/official/cv/warpctc/src/loss.py +++ b/model_zoo/official/cv/warpctc/src/loss.py @@ -33,13 +33,12 @@ class CTCLoss(_Loss): def __init__(self, max_sequence_length, max_label_length, batch_size): super(CTCLoss, self).__init__() - self.sequence_length = Parameter(Tensor(np.array([max_sequence_length] * batch_size), mstype.int32), - name="sequence_length") + self.sequence_length = Parameter(Tensor(np.array([max_sequence_length] * batch_size), mstype.int32)) labels_indices = [] for i in range(batch_size): for j in range(max_label_length): labels_indices.append([i, j]) - self.labels_indices = Parameter(Tensor(np.array(labels_indices), mstype.int64), name="labels_indices") + self.labels_indices = Parameter(Tensor(np.array(labels_indices), mstype.int64)) self.reshape = P.Reshape() self.ctc_loss = P.CTCLoss(ctc_merge_repeated=True) diff --git a/model_zoo/official/cv/warpctc/src/warpctc.py b/model_zoo/official/cv/warpctc/src/warpctc.py index dc8a491784..5bce0012be 100755 --- a/model_zoo/official/cv/warpctc/src/warpctc.py +++ b/model_zoo/official/cv/warpctc/src/warpctc.py @@ -45,12 +45,10 @@ class StackedRNN(nn.Cell): self.rnn1 = P.DynamicRNN(forget_bias=0.0) self.rnn2 = P.DynamicRNN(forget_bias=0.0) - self.w1 = Parameter(np.random.uniform(-k, k, (input_size + hidden_size, 4 * hidden_size)).astype(np.float16), - name="w1") - self.w2 = Parameter(np.random.uniform(-k, k, (hidden_size + hidden_size, 4 * hidden_size)).astype(np.float16), - name="w2") - self.b1 = Parameter(np.random.uniform(-k, k, (4 * hidden_size)).astype(np.float16), name="b1") - self.b2 = Parameter(np.random.uniform(-k, k, (4 * hidden_size)).astype(np.float16), name="b2") + self.w1 = Parameter(np.random.uniform(-k, k, (input_size + hidden_size, 4 * hidden_size)).astype(np.float16)) + self.w2 = Parameter(np.random.uniform(-k, k, (hidden_size + hidden_size, 4 * hidden_size)).astype(np.float16)) + self.b1 = Parameter(np.random.uniform(-k, k, (4 * hidden_size)).astype(np.float16)) + self.b2 = Parameter(np.random.uniform(-k, k, (4 * hidden_size)).astype(np.float16)) self.h1 = Tensor(np.zeros(shape=(1, batch_size, hidden_size)).astype(np.float16)) self.h2 = Tensor(np.zeros(shape=(1, batch_size, hidden_size)).astype(np.float16)) @@ -98,7 +96,7 @@ class StackedRNNForGPU(nn.Cell): self.cast = P.Cast() k = (1 / hidden_size) ** 0.5 weight_shape = 4 * hidden_size * (input_size + 3 * hidden_size + 4) - self.weight = Parameter(np.random.uniform(-k, k, (weight_shape, 1, 1)).astype(np.float32), name='weight') + self.weight = Parameter(np.random.uniform(-k, k, (weight_shape, 1, 1)).astype(np.float32)) self.h = Tensor(np.zeros(shape=(num_layer, batch_size, hidden_size)).astype(np.float32)) self.c = Tensor(np.zeros(shape=(num_layer, batch_size, hidden_size)).astype(np.float32)) diff --git a/model_zoo/official/gnn/bgcf/src/bgcf.py b/model_zoo/official/gnn/bgcf/src/bgcf.py index 82bb442a8e..408f589495 100644 --- a/model_zoo/official/gnn/bgcf/src/bgcf.py +++ b/model_zoo/official/gnn/bgcf/src/bgcf.py @@ -39,7 +39,6 @@ class MeanConv(nn.Cell): """ def __init__(self, - name, feature_in_dim, feature_out_dim, activation, @@ -47,8 +46,7 @@ class MeanConv(nn.Cell): super(MeanConv, self).__init__() self.out_weight = Parameter( - initializer("XavierUniform", [feature_in_dim * 2, feature_out_dim], dtype=mstype.float32), - name=name + 'out_weight') + initializer("XavierUniform", [feature_in_dim * 2, feature_out_dim], dtype=mstype.float32)) if activation == "tanh": self.act = P.Tanh() @@ -90,15 +88,13 @@ class AttenConv(nn.Cell): """ def __init__(self, - name, feature_in_dim, feature_out_dim, dropout=0.2): super(AttenConv, self).__init__() self.out_weight = Parameter( - initializer("XavierUniform", [feature_in_dim * 2, feature_out_dim], dtype=mstype.float32), - name=name + 'out_weight') + initializer("XavierUniform", [feature_in_dim * 2, feature_out_dim], dtype=mstype.float32)) self.cast = P.Cast() self.squeeze = P.Squeeze(1) self.concat = P.Concat(axis=1) @@ -147,10 +143,8 @@ class BGCF(nn.Cell): input_dim): super(BGCF, self).__init__() - self.user_embeddings = Parameter(initializer("XavierUniform", [num_user, input_dim], dtype=mstype.float32), - name='user_embed') - self.item_embeddings = Parameter(initializer("XavierUniform", [num_item, input_dim], dtype=mstype.float32), - name='item_embed') + self.user_embed = Parameter(initializer("XavierUniform", [num_user, input_dim], dtype=mstype.float32)) + self.item_embed = Parameter(initializer("XavierUniform", [num_item, input_dim], dtype=mstype.float32)) self.cast = P.Cast() self.tanh = P.Tanh() self.shape = P.Shape() @@ -163,30 +157,27 @@ class BGCF(nn.Cell): (self.input_dim, self.num_user, self.num_item) = dataset_argv self.layer_dim = architect_argv - self.gnew_agg_mean = MeanConv('gnew_agg_mean', self.input_dim, self.layer_dim, + self.gnew_agg_mean = MeanConv(self.input_dim, self.layer_dim, activation=activation, dropout=neigh_drop_rate[1]) self.gnew_agg_mean.to_float(mstype.float16) - self.gnew_agg_user = AttenConv('gnew_agg_att_user', self.input_dim, - self.layer_dim, dropout=neigh_drop_rate[2]) + self.gnew_agg_user = AttenConv(self.input_dim, self.layer_dim, dropout=neigh_drop_rate[2]) self.gnew_agg_user.to_float(mstype.float16) - self.gnew_agg_item = AttenConv('gnew_agg_att_item', self.input_dim, - self.layer_dim, dropout=neigh_drop_rate[2]) + self.gnew_agg_item = AttenConv(self.input_dim, self.layer_dim, dropout=neigh_drop_rate[2]) self.gnew_agg_item.to_float(mstype.float16) self.user_feature_dim = self.input_dim self.item_feature_dim = self.input_dim self.final_weight = Parameter( - initializer("XavierUniform", [self.input_dim * 3, self.input_dim * 3], dtype=mstype.float32), - name='final_weight') + initializer("XavierUniform", [self.input_dim * 3, self.input_dim * 3], dtype=mstype.float32)) - self.raw_agg_funcs_user = MeanConv('raw_agg_user', self.input_dim, self.layer_dim, + self.raw_agg_funcs_user = MeanConv(self.input_dim, self.layer_dim, activation=activation, dropout=neigh_drop_rate[0]) self.raw_agg_funcs_user.to_float(mstype.float16) - self.raw_agg_funcs_item = MeanConv('raw_agg_item', self.input_dim, self.layer_dim, + self.raw_agg_funcs_item = MeanConv(self.input_dim, self.layer_dim, activation=activation, dropout=neigh_drop_rate[0]) self.raw_agg_funcs_item.to_float(mstype.float16) @@ -207,14 +198,14 @@ class BGCF(nn.Cell): neg_gnew_neighs, neg_item_num): """Aggregate user and item embeddings""" - all_user_embed = self.gather(self.user_embeddings, self.concat_0((u_id, pos_users)), 0) + all_user_embed = self.gather(self.user_embed, self.concat_0((u_id, pos_users)), 0) - u_self_matrix_at_layers = self.gather(self.user_embeddings, u_group_nodes, 0) - u_neigh_matrix_at_layers = self.gather(self.item_embeddings, u_neighs, 0) + u_self_matrix_at_layers = self.gather(self.user_embed, u_group_nodes, 0) + u_neigh_matrix_at_layers = self.gather(self.item_embed, u_neighs, 0) u_output_mean = self.raw_agg_funcs_user(u_self_matrix_at_layers, u_neigh_matrix_at_layers) - u_gnew_neighs_matrix = self.gather(self.item_embeddings, u_gnew_neighs, 0) + u_gnew_neighs_matrix = self.gather(self.item_embed, u_gnew_neighs, 0) u_output_from_gnew_mean = self.gnew_agg_mean(u_self_matrix_at_layers, u_gnew_neighs_matrix) u_output_from_gnew_att = self.gnew_agg_user(u_self_matrix_at_layers, @@ -223,14 +214,14 @@ class BGCF(nn.Cell): u_output = self.concat_1((u_output_mean, u_output_from_gnew_mean, u_output_from_gnew_att)) all_user_rep = self.tanh(u_output) - all_pos_item_embed = self.gather(self.item_embeddings, self.concat_0((pos_item_id, pos_items)), 0) + all_pos_item_embed = self.gather(self.item_embed, self.concat_0((pos_item_id, pos_items)), 0) - i_self_matrix_at_layers = self.gather(self.item_embeddings, i_group_nodes, 0) - i_neigh_matrix_at_layers = self.gather(self.user_embeddings, i_neighs, 0) + i_self_matrix_at_layers = self.gather(self.item_embed, i_group_nodes, 0) + i_neigh_matrix_at_layers = self.gather(self.user_embed, i_neighs, 0) i_output_mean = self.raw_agg_funcs_item(i_self_matrix_at_layers, i_neigh_matrix_at_layers) - i_gnew_neighs_matrix = self.gather(self.user_embeddings, i_gnew_neighs, 0) + i_gnew_neighs_matrix = self.gather(self.user_embed, i_gnew_neighs, 0) i_output_from_gnew_mean = self.gnew_agg_mean(i_self_matrix_at_layers, i_gnew_neighs_matrix) i_output_from_gnew_att = self.gnew_agg_item(i_self_matrix_at_layers, @@ -239,14 +230,14 @@ class BGCF(nn.Cell): i_output = self.concat_1((i_output_mean, i_output_from_gnew_mean, i_output_from_gnew_att)) all_pos_item_rep = self.tanh(i_output) - neg_item_embed = self.gather(self.item_embeddings, neg_item_id, 0) + neg_item_embed = self.gather(self.item_embed, neg_item_id, 0) - neg_self_matrix_at_layers = self.gather(self.item_embeddings, neg_group_nodes, 0) - neg_neigh_matrix_at_layers = self.gather(self.user_embeddings, neg_neighs, 0) + neg_self_matrix_at_layers = self.gather(self.item_embed, neg_group_nodes, 0) + neg_neigh_matrix_at_layers = self.gather(self.user_embed, neg_neighs, 0) neg_output_mean = self.raw_agg_funcs_item(neg_self_matrix_at_layers, neg_neigh_matrix_at_layers) - neg_gnew_neighs_matrix = self.gather(self.user_embeddings, neg_gnew_neighs, 0) + neg_gnew_neighs_matrix = self.gather(self.user_embed, neg_gnew_neighs, 0) neg_output_from_gnew_mean = self.gnew_agg_mean(neg_self_matrix_at_layers, neg_gnew_neighs_matrix) neg_output_from_gnew_att = self.gnew_agg_item(neg_self_matrix_at_layers, diff --git a/model_zoo/official/gnn/gat/src/gat.py b/model_zoo/official/gnn/gat/src/gat.py index d33f765d06..1bb1402576 100644 --- a/model_zoo/official/gnn/gat/src/gat.py +++ b/model_zoo/official/gnn/gat/src/gat.py @@ -80,14 +80,14 @@ class GNNFeatureTransform(nn.Cell): 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.weight = Parameter(initializer(weight_init, [out_channels, in_channels])) 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 = Parameter(initializer(bias_init, [out_channels])) self.matmul = P.MatMul(transpose_b=True) self.bias_add = P.BiasAdd() @@ -280,7 +280,7 @@ class AttentionHead(nn.Cell): self.coef_drop = nn.Dropout(keep_prob=1 - coef_drop_ratio) self.matmul = P.MatMul() self.bias_add = P.BiasAdd() - self.bias = Parameter(initializer('zeros', self.out_channel), name='bias') + self.bias = Parameter(initializer('zeros', self.out_channel)) self.residual = residual if self.residual: if in_channel != out_channel: diff --git a/model_zoo/official/nlp/bert/pretrain_eval.py b/model_zoo/official/nlp/bert/pretrain_eval.py index 4bf503b3fc..fdb36dff54 100644 --- a/model_zoo/official/nlp/bert/pretrain_eval.py +++ b/model_zoo/official/nlp/bert/pretrain_eval.py @@ -80,8 +80,8 @@ class BertPretrainEva(nn.Cell): self.equal = P.Equal() self.mean = P.ReduceMean() self.sum = P.ReduceSum() - self.total = Parameter(Tensor([0], mstype.float32), name='total') - self.acc = Parameter(Tensor([0], mstype.float32), name='acc') + self.total = Parameter(Tensor([0], mstype.float32)) + self.acc = Parameter(Tensor([0], mstype.float32)) self.reshape = P.Reshape() self.shape = P.Shape() self.cast = P.Cast() diff --git a/model_zoo/official/nlp/bert/src/CRF.py b/model_zoo/official/nlp/bert/src/CRF.py index 6c9fd5ea96..ef6cb41297 100644 --- a/model_zoo/official/nlp/bert/src/CRF.py +++ b/model_zoo/official/nlp/bert/src/CRF.py @@ -52,7 +52,7 @@ class CRF(nn.Cell): transitions = np.random.normal(size=(self.target_size, self.target_size)).astype(np.float32) transitions[tag_to_index[self.START_TAG], :] = -10000 transitions[:, tag_to_index[self.STOP_TAG]] = -10000 - self.transitions = Parameter(Tensor(transitions), name="transition_matrix") + self.transitions = Parameter(Tensor(transitions)) self.cat = P.Concat(axis=-1) self.argmax = P.ArgMaxWithValue(axis=-1) self.log = P.Log() diff --git a/model_zoo/official/nlp/bert/src/bert_for_finetune.py b/model_zoo/official/nlp/bert/src/bert_for_finetune.py index fcce6ed346..3ef32cbc17 100644 --- a/model_zoo/official/nlp/bert/src/bert_for_finetune.py +++ b/model_zoo/official/nlp/bert/src/bert_for_finetune.py @@ -90,8 +90,7 @@ class BertFinetuneCell(nn.Cell): self.loss_scale = None self.loss_scaling_manager = scale_update_cell if scale_update_cell: - self.loss_scale = Parameter(Tensor(scale_update_cell.get_loss_scale(), dtype=mstype.float32), - name="loss_scale") + self.loss_scale = Parameter(Tensor(scale_update_cell.get_loss_scale(), dtype=mstype.float32)) def construct(self, input_ids, @@ -185,8 +184,8 @@ class BertSquadCell(nn.Cell): self.loss_scale = None self.loss_scaling_manager = scale_update_cell if scale_update_cell: - self.loss_scale = Parameter(Tensor(scale_update_cell.get_loss_scale(), dtype=mstype.float32), - name="loss_scale") + self.loss_scale = Parameter(Tensor(scale_update_cell.get_loss_scale(), dtype=mstype.float32)) + def construct(self, input_ids, input_mask, @@ -306,9 +305,9 @@ class BertSquad(nn.Cell): self.num_labels = num_labels self.seq_length = config.seq_length self.is_training = is_training - self.total_num = Parameter(Tensor([0], mstype.float32), name='total_num') - self.start_num = Parameter(Tensor([0], mstype.float32), name='start_num') - self.end_num = Parameter(Tensor([0], mstype.float32), name='end_num') + self.total_num = Parameter(Tensor([0], mstype.float32)) + self.start_num = Parameter(Tensor([0], mstype.float32)) + self.end_num = Parameter(Tensor([0], mstype.float32)) self.sum = P.ReduceSum() self.equal = P.Equal() self.argmax = P.ArgMaxWithValue(axis=1) diff --git a/model_zoo/official/nlp/bert/src/bert_for_pre_training.py b/model_zoo/official/nlp/bert/src/bert_for_pre_training.py index 1d3606921d..50b05aecad 100644 --- a/model_zoo/official/nlp/bert/src/bert_for_pre_training.py +++ b/model_zoo/official/nlp/bert/src/bert_for_pre_training.py @@ -84,8 +84,7 @@ class GetMaskedLMOutput(nn.Cell): self.output_bias = Parameter( initializer( 'zero', - config.vocab_size), - name='output_bias') + config.vocab_size)) self.matmul = P.MatMul(transpose_b=True) self.log_softmax = nn.LogSoftmax(axis=-1) self.shape_flat_offsets = (-1, 1) @@ -359,8 +358,7 @@ class BertTrainOneStepWithLossScaleCell(nn.Cell): self.loss_scale = None self.loss_scaling_manager = scale_update_cell if scale_update_cell: - self.loss_scale = Parameter(Tensor(scale_update_cell.get_loss_scale(), dtype=mstype.float32), - name="loss_scale") + self.loss_scale = Parameter(Tensor(scale_update_cell.get_loss_scale(), dtype=mstype.float32)) @C.add_flags(has_effect=True) def construct(self, @@ -465,10 +463,10 @@ class BertTrainAccumulateStepsWithLossScaleCell(nn.Cell): self.enable_global_norm = enable_global_norm self.one = Tensor(np.array([1]).astype(np.int32)) self.zero = Tensor(np.array([0]).astype(np.int32)) - self.local_step = Parameter(initializer(0, [1], mstype.int32), name="local_step") + self.local_step = Parameter(initializer(0, [1], mstype.int32)) self.accu_grads = self.weights.clone(prefix="accu_grads", init='zeros') - self.accu_overflow = Parameter(initializer(0, [1], mstype.int32), name="accu_overflow") - self.loss = Parameter(initializer(0, [1], mstype.float32), name="accu_loss") + self.accu_overflow = Parameter(initializer(0, [1], mstype.int32)) + self.accu_loss = Parameter(initializer(0, [1], mstype.float32)) self.grad = C.GradOperation(get_by_list=True, sens_param=True) self.reducer_flag = False @@ -499,8 +497,7 @@ class BertTrainAccumulateStepsWithLossScaleCell(nn.Cell): self.loss_scale = None self.loss_scaling_manager = scale_update_cell if scale_update_cell: - self.loss_scale = Parameter(Tensor(scale_update_cell.get_loss_scale(), dtype=mstype.float32), - name="loss_scale") + self.loss_scale = Parameter(Tensor(scale_update_cell.get_loss_scale(), dtype=mstype.float32)) @C.add_flags(has_effect=True) def construct(self, @@ -529,8 +526,8 @@ class BertTrainAccumulateStepsWithLossScaleCell(nn.Cell): # update accumulation parameters is_accu_step = self.not_equal(self.local_step, self.accumulation_steps) self.local_step = self.select(is_accu_step, self.local_step + self.one, self.one) - self.loss = self.select(is_accu_step, self.loss + loss, loss) - mean_loss = self.loss / self.local_step + self.accu_loss = self.select(is_accu_step, self.accu_loss + loss, loss) + mean_loss = self.accu_loss / self.local_step is_accu_step = self.not_equal(self.local_step, self.accumulation_steps) # alloc status and clear should be right before gradoperation diff --git a/model_zoo/official/nlp/bert/src/bert_model.py b/model_zoo/official/nlp/bert/src/bert_model.py index 5fb7d90177..77c3ccc7c3 100644 --- a/model_zoo/official/nlp/bert/src/bert_model.py +++ b/model_zoo/official/nlp/bert/src/bert_model.py @@ -110,8 +110,7 @@ class EmbeddingLookup(nn.Cell): self.use_one_hot_embeddings = use_one_hot_embeddings self.embedding_table = Parameter(initializer (TruncatedNormal(initializer_range), - [vocab_size, embedding_size]), - name='embedding_table') + [vocab_size, embedding_size])) self.expand = P.ExpandDims() self.shape_flat = (-1,) self.gather = P.GatherV2() @@ -170,8 +169,7 @@ class EmbeddingPostprocessor(nn.Cell): self.embedding_table = Parameter(initializer (TruncatedNormal(initializer_range), [token_type_vocab_size, - embedding_size]), - name='embedding_table') + embedding_size])) self.shape_flat = (-1,) self.one_hot = P.OneHot() @@ -188,8 +186,7 @@ class EmbeddingPostprocessor(nn.Cell): self.full_position_embeddings = Parameter(initializer (TruncatedNormal(initializer_range), [max_position_embeddings, - embedding_size]), - name='full_position_embeddings') + embedding_size])) def construct(self, token_type_ids, word_embeddings): """Postprocessors apply positional and token type embeddings to word embeddings.""" @@ -314,8 +311,7 @@ class RelaPosEmbeddingsGenerator(nn.Cell): self.embeddings_table = Parameter( initializer(TruncatedNormal(initializer_range), - [self.vocab_size, self.depth]), - name='embeddings_for_position') + [self.vocab_size, self.depth])) self.relative_positions_matrix = RelaPosMatrixGenerator(length=length, max_relative_position=max_relative_position) diff --git a/model_zoo/official/nlp/bert_thor/pretrain_eval.py b/model_zoo/official/nlp/bert_thor/pretrain_eval.py index ea7c563dcc..3f565374f4 100644 --- a/model_zoo/official/nlp/bert_thor/pretrain_eval.py +++ b/model_zoo/official/nlp/bert_thor/pretrain_eval.py @@ -86,8 +86,8 @@ class BertPretrainEva(nn.Cell): self.equal = P.Equal() self.mean = P.ReduceMean() self.sum = P.ReduceSum() - self.total = Parameter(Tensor([0], mstype.float32), name='total') - self.acc = Parameter(Tensor([0], mstype.float32), name='acc') + self.total = Parameter(Tensor([0], mstype.float32)) + self.acc = Parameter(Tensor([0], mstype.float32)) self.reshape = P.Reshape() self.shape = P.Shape() self.cast = P.Cast() diff --git a/model_zoo/official/nlp/bert_thor/src/bert_for_pre_training.py b/model_zoo/official/nlp/bert_thor/src/bert_for_pre_training.py index 507dca9ee0..161273d3cc 100644 --- a/model_zoo/official/nlp/bert_thor/src/bert_for_pre_training.py +++ b/model_zoo/official/nlp/bert_thor/src/bert_for_pre_training.py @@ -98,8 +98,7 @@ class GetMaskedLMOutput(nn.Cell): self.output_bias = Parameter( initializer( 'zero', - config.vocab_size), - name='output_bias') + config.vocab_size)) self.matmul = P.MatMul(transpose_b=True) self.log_softmax = nn.LogSoftmax(axis=-1) self.shape_flat_offsets = (-1, 1) @@ -379,8 +378,7 @@ class BertTrainOneStepWithLossScaleCell(nn.Cell): self.loss_scale = None self.loss_scaling_manager = scale_update_cell if scale_update_cell: - self.loss_scale = Parameter(Tensor(scale_update_cell.get_loss_scale(), dtype=mstype.float32), - name="loss_scale") + self.loss_scale = Parameter(Tensor(scale_update_cell.get_loss_scale(), dtype=mstype.float32)) @C.add_flags(has_effect=True) def construct(self, diff --git a/model_zoo/official/nlp/bert_thor/src/bert_model.py b/model_zoo/official/nlp/bert_thor/src/bert_model.py index d31b822b9e..6e26658b45 100644 --- a/model_zoo/official/nlp/bert_thor/src/bert_model.py +++ b/model_zoo/official/nlp/bert_thor/src/bert_model.py @@ -136,8 +136,7 @@ class EmbeddingLookup(nn.Cell): self.use_one_hot_embeddings = use_one_hot_embeddings self.embedding_table = Parameter(initializer (TruncatedNormal(initializer_range), - [vocab_size, embedding_size]), - name='embedding_table') + [vocab_size, embedding_size])) self.expand = P.ExpandDims() self.shape_flat = (-1,) self.gather = P.GatherV2() @@ -200,7 +199,6 @@ class EmbeddingPostprocessor(nn.Cell): embedding_shape=embedding_shape, use_one_hot_embeddings=use_one_hot_embeddings, initializer_range=initializer_range, - name='embedding_table', batch_size=batch_size, damping=damping, loss_scale=loss_scale, @@ -224,7 +222,6 @@ class EmbeddingPostprocessor(nn.Cell): embedding_shape=position_embedding_shape, use_one_hot_embeddings=use_one_hot_embeddings, initializer_range=initializer_range, - name='full_position_embeddings', batch_size=batch_size, damping=damping, loss_scale=loss_scale, @@ -363,8 +360,7 @@ class RelaPosEmbeddingsGenerator(nn.Cell): self.embeddings_table = Parameter( initializer(TruncatedNormal(initializer_range), - [self.vocab_size, self.depth]), - name='embeddings_for_position') + [self.vocab_size, self.depth])) self.relative_positions_matrix = RelaPosMatrixGenerator(length=length, max_relative_position=max_relative_position) @@ -944,7 +940,6 @@ class BertModel(nn.Cell): embedding_shape=output_embedding_shape, use_one_hot_embeddings=use_one_hot_embeddings, initializer_range=config.initializer_range, - name='embedding_table', batch_size=batch_size, damping=damping, loss_scale=loss_scale, diff --git a/model_zoo/official/nlp/bert_thor/src/fused_layer_norm.py b/model_zoo/official/nlp/bert_thor/src/fused_layer_norm.py index 96930719e3..0932625b6f 100644 --- a/model_zoo/official/nlp/bert_thor/src/fused_layer_norm.py +++ b/model_zoo/official/nlp/bert_thor/src/fused_layer_norm.py @@ -94,9 +94,9 @@ class FusedLayerNorm(Cell): self.begin_norm_axis = begin_norm_axis self.begin_params_axis = begin_params_axis self.gamma = Parameter(initializer( - gamma_init, normalized_shape), name="gamma") + gamma_init, normalized_shape)) self.beta = Parameter(initializer( - beta_init, normalized_shape), name="beta") + beta_init, normalized_shape)) self.layer_norm = P.LayerNorm(begin_norm_axis=self.begin_norm_axis, begin_params_axis=self.begin_params_axis) self.batch_norm = P.BatchNorm(is_training=True, epsilon=1e-5) diff --git a/model_zoo/official/nlp/bert_thor/src/thor_for_bert.py b/model_zoo/official/nlp/bert_thor/src/thor_for_bert.py index 36c20a7992..f9a8839991 100644 --- a/model_zoo/official/nlp/bert_thor/src/thor_for_bert.py +++ b/model_zoo/official/nlp/bert_thor/src/thor_for_bert.py @@ -52,7 +52,7 @@ class THOR(Optimizer): super(THOR, self).__init__(learning_rate, params, weight_decay, loss_scale) if isinstance(momentum, float) and momentum < 0.0: raise ValueError("momentum should be at least 0.0, but got momentum {}".format(momentum)) - self.momentum = Parameter(Tensor(momentum, mstype.float32), name="momentum") + self.momentum = Parameter(Tensor(momentum, mstype.float32)) self.params = self.parameters self.moments = self.params.clone(prefix="moments", init='zeros') self.hyper_map = C.HyperMap() @@ -80,7 +80,7 @@ class THOR(Optimizer): self.batch_size = batch_size self.damping = damping self.one = Tensor(1, mstype.int32) - self.cov_step = Parameter(initializer(0, [1], mstype.int32), name="cov_step", requires_grad=False) + self.cov_step = Parameter(initializer(0, [1], mstype.int32), requires_grad=False) def construct(self, gradients): """construct of THOR""" diff --git a/model_zoo/official/nlp/bert_thor/src/thor_for_bert_arg.py b/model_zoo/official/nlp/bert_thor/src/thor_for_bert_arg.py index c1835b6e3b..4af6bded5f 100644 --- a/model_zoo/official/nlp/bert_thor/src/thor_for_bert_arg.py +++ b/model_zoo/official/nlp/bert_thor/src/thor_for_bert_arg.py @@ -54,7 +54,7 @@ class THOR(Optimizer): super(THOR, self).__init__(learning_rate, params, weight_decay, loss_scale) if isinstance(momentum, float) and momentum < 0.0: raise ValueError("momentum should be at least 0.0, but got momentum {}".format(momentum)) - self.momentum = Parameter(Tensor(momentum, mstype.float32), name="momentum") + self.momentum = Parameter(Tensor(momentum, mstype.float32)) self.params = self.parameters self.moments = self.params.clone(prefix="moments", init='zeros') self.hyper_map = C.HyperMap() @@ -82,7 +82,7 @@ class THOR(Optimizer): self.batch_size = batch_size self.damping = damping self.one = Tensor(1, mstype.int32) - self.cov_step = Parameter(initializer(0, [1], mstype.int32), name="cov_step", requires_grad=False) + self.cov_step = Parameter(initializer(0, [1], mstype.int32), requires_grad=False) mean = _get_gradients_mean() degree = _get_device_num() self.grad_reducer_g = DistributedGradReducerThor(self.parameters, 3, mean, degree) diff --git a/model_zoo/official/nlp/bert_thor/src/thor_layer.py b/model_zoo/official/nlp/bert_thor/src/thor_layer.py index 5814e95604..950e038f28 100644 --- a/model_zoo/official/nlp/bert_thor/src/thor_layer.py +++ b/model_zoo/official/nlp/bert_thor/src/thor_layer.py @@ -41,7 +41,6 @@ class Embedding_Thor(Cell): embedding_shape, use_one_hot_embeddings=False, initializer_range=0.02, - name='embedding_table', batch_size=12, damping=0.03, loss_scale=1, @@ -52,8 +51,7 @@ class Embedding_Thor(Cell): self.use_one_hot_embeddings = use_one_hot_embeddings self.embedding_table = Parameter(initializer (TruncatedNormal(initializer_range), - [vocab_size, embedding_size]), - name=name) + [vocab_size, embedding_size])) self.thor = True self.expand = P.ExpandDims() self.shape_flat = (-1,) @@ -67,14 +65,13 @@ class Embedding_Thor(Cell): self.shape = P.Shape() self.loss_scale = Tensor(1 / loss_scale, mstype.float16) - self.matrix_A_inv = Parameter(Tensor(np.zeros([vocab_size]).astype(np.float16)), - name='matrix_A_inv', requires_grad=False) + self.matrix_A_inv = Parameter(Tensor(np.zeros([vocab_size]).astype(np.float16)), requires_grad=False) self.matrix_G_inv = Parameter(Tensor(np.zeros([embedding_size, embedding_size]).astype(np.float16)), - name="matrix_G_inv", requires_grad=False) + requires_grad=False) self.fake_G = Tensor(np.zeros([embedding_size, embedding_size]).astype(np.float16)) self.dampingA = Tensor(np.ones([vocab_size]).astype(np.float32)) self.dampingG = Tensor(np.identity(embedding_size), mstype.float32) - self.cov_step = Parameter(initializer(0, [1], mstype.int32), name="cov_step", requires_grad=False) + self.cov_step = Parameter(initializer(0, [1], mstype.int32), requires_grad=False) self.freq = Tensor(frequency, mstype.int32) self.axis = 0 self.damping = damping @@ -169,14 +166,14 @@ class Dense_Thor(Cell): 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.weight = Parameter(initializer(weight_init, [out_channels, in_channels])) 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 = Parameter(initializer(bias_init, [out_channels])) self.matmul = P.MatMul(transpose_b=True) self.bias_add = P.BiasAdd() @@ -184,9 +181,9 @@ class Dense_Thor(Cell): 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) + 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) + requires_grad=False) self.fake_G = Tensor(np.zeros([out_channels, out_channels]).astype(np.float16)) self.matmul = P.MatMul(transpose_b=True) @@ -196,7 +193,7 @@ class Dense_Thor(Cell): 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.cov_step = Parameter(initializer(0, [1], mstype.int32), requires_grad=False) self.mul = P.Mul() self.cast = P.Cast() self.damping = damping diff --git a/model_zoo/official/nlp/gnmt_v2/src/gnmt_model/attention.py b/model_zoo/official/nlp/gnmt_v2/src/gnmt_model/attention.py index 36eb1e3ff1..cdc32943ed 100644 --- a/model_zoo/official/nlp/gnmt_v2/src/gnmt_model/attention.py +++ b/model_zoo/official/nlp/gnmt_v2/src/gnmt_model/attention.py @@ -57,11 +57,10 @@ class BahdanauAttention(nn.Cell): self.normalize = normalize self.num_units = num_units self.linear_att = Parameter(Tensor(np.random.uniform(-initializer_range, initializer_range, size=[num_units]), - dtype=mstype.float32), name='linear_att') + dtype=mstype.float32)) if self.normalize: - self.normalize_scalar = Parameter(Tensor(np.array([1.0 / num_units]), dtype=mstype.float32), - name='normalize_scalar') - self.normalize_bias = Parameter(Tensor(np.zeros(num_units), dtype=mstype.float32), name='normalize_bias') + self.normalize_scalar = Parameter(Tensor(np.array([1.0 / num_units]), dtype=mstype.float32)) + self.normalize_bias = Parameter(Tensor(np.zeros(num_units), dtype=mstype.float32)) self.transpose = P.Transpose() self.transpose_orders = (1, 0, 2) self.shape_op = P.Shape() diff --git a/model_zoo/official/nlp/gnmt_v2/src/gnmt_model/dynamic_rnn.py b/model_zoo/official/nlp/gnmt_v2/src/gnmt_model/dynamic_rnn.py index 4eb8399bc8..4f7daaa355 100644 --- a/model_zoo/official/nlp/gnmt_v2/src/gnmt_model/dynamic_rnn.py +++ b/model_zoo/official/nlp/gnmt_v2/src/gnmt_model/dynamic_rnn.py @@ -49,10 +49,10 @@ class DynamicRNNCell(nn.Cell): # w dynamicRNN_w = np.random.uniform(-initializer_range, initializer_range, size=[self.input_size + self.hidden_size, 4 * self.hidden_size]) - self.dynamicRNN_w = Parameter(Tensor(dynamicRNN_w, mstype.float32), name='weight') + self.dynamicRNN_w = Parameter(Tensor(dynamicRNN_w, mstype.float32)) # b dynamicRNN_b = np.random.uniform(-initializer_range, initializer_range, size=[4 * self.hidden_size]) - self.dynamicRNN_b = Parameter(Tensor(dynamicRNN_b, mstype.float32), name='bias') + self.dynamicRNN_b = Parameter(Tensor(dynamicRNN_b, mstype.float32)) self.dynamicRNN_h = Tensor(np.zeros((1, self.batch_size, self.hidden_size)), mstype.float32) self.dynamicRNN_c = Tensor(np.zeros((1, self.batch_size, self.hidden_size)), mstype.float32) diff --git a/model_zoo/official/nlp/gnmt_v2/src/gnmt_model/embedding.py b/model_zoo/official/nlp/gnmt_v2/src/gnmt_model/embedding.py index e9aa0a8e4b..0158d3f8a3 100644 --- a/model_zoo/official/nlp/gnmt_v2/src/gnmt_model/embedding.py +++ b/model_zoo/official/nlp/gnmt_v2/src/gnmt_model/embedding.py @@ -48,8 +48,7 @@ class EmbeddingLookup(nn.Cell): self.use_one_hot_embeddings = use_one_hot_embeddings init_weight = np.random.normal(-initializer_range, initializer_range, size=[vocab_size, embed_dim]) - self.embedding_table = Parameter(Tensor(init_weight, mstype.float32), - name='embedding_table') + self.embedding_table = Parameter(Tensor(init_weight, mstype.float32)) self.expand = P.ExpandDims() self.gather = P.GatherV2() self.one_hot = P.OneHot() diff --git a/model_zoo/official/nlp/gnmt_v2/src/gnmt_model/gnmt_for_train.py b/model_zoo/official/nlp/gnmt_v2/src/gnmt_model/gnmt_for_train.py index fb3a2da4e7..c96d158376 100644 --- a/model_zoo/official/nlp/gnmt_v2/src/gnmt_model/gnmt_for_train.py +++ b/model_zoo/official/nlp/gnmt_v2/src/gnmt_model/gnmt_for_train.py @@ -253,8 +253,7 @@ class GNMTTrainOneStepWithLossScaleCell(nn.Cell): self.loss_scale = None self.loss_scaling_manager = scale_update_cell if scale_update_cell: - self.loss_scale = Parameter(Tensor(scale_update_cell.get_loss_scale(), dtype=mstype.float32), - name="loss_scale") + self.loss_scale = Parameter(Tensor(scale_update_cell.get_loss_scale(), dtype=mstype.float32)) self.add_flags(has_effect=True) self.loss_scalar = P.ScalarSummary() diff --git a/model_zoo/official/nlp/gnmt_v2/src/utils/optimizer.py b/model_zoo/official/nlp/gnmt_v2/src/utils/optimizer.py index f181d9a6e7..b983367213 100644 --- a/model_zoo/official/nlp/gnmt_v2/src/utils/optimizer.py +++ b/model_zoo/official/nlp/gnmt_v2/src/utils/optimizer.py @@ -217,8 +217,8 @@ class Adam(Optimizer): self.beta1 = Tensor(beta1, mstype.float32) self.beta2 = Tensor(beta2, mstype.float32) - self.beta1_power = Parameter(initializer(1, [1], mstype.float32), name="beta1_power") - self.beta2_power = Parameter(initializer(1, [1], mstype.float32), name="beta2_power") + self.beta1_power = Parameter(initializer(1, [1], mstype.float32)) + self.beta2_power = Parameter(initializer(1, [1], mstype.float32)) self.eps = eps self.moment1 = self.parameters.clone(prefix="moment1", init='zeros') @@ -377,7 +377,7 @@ class AdamWeightDecayDynamicLR(Optimizer): _check_param_value(beta1, beta2, eps, weight_decay, self.cls_name) _check_learning_rate_value(learning_rate, end_learning_rate, decay_steps, power, self.cls_name) # turn them to scalar when me support scalar/tensor mix operations - self.global_step = Parameter(initializer(0, [1]), name="global_step") + self.global_step = Parameter(initializer(0, [1])) self.warmup_steps = Tensor(np.array([warmup_steps]).astype(np.float32)) self.warmup_flag = False if warmup_steps > 0: diff --git a/model_zoo/official/nlp/gpt/src/gpt.py b/model_zoo/official/nlp/gpt/src/gpt.py index 06280ef654..5f249236d5 100644 --- a/model_zoo/official/nlp/gpt/src/gpt.py +++ b/model_zoo/official/nlp/gpt/src/gpt.py @@ -41,8 +41,8 @@ class LayerNorm(nn.Cell): """ def __init__(self, normalized_shape, eps=1e-5): super(LayerNorm, self).__init__() - self.gamma = Parameter(initializer('ones', normalized_shape), name="gamma") - self.beta = Parameter(initializer('zeros', normalized_shape), name="beta") + self.gamma = Parameter(initializer('ones', normalized_shape)) + self.beta = Parameter(initializer('zeros', normalized_shape)) self.mean = P.ReduceMean(keep_dims=True) self.eps = eps @@ -100,8 +100,8 @@ class Mapping(nn.Cell): super(Mapping, self).__init__() self.output_size = output_size self.input_size = input_size - self.weight = Parameter(initializer(Normal(sigma=0.02*scale), [input_size, output_size]), name="mapping_weight") - self.bias = Parameter(initializer("zeros", [output_size,]), name="mapping_bias") + self.weight = Parameter(initializer(Normal(sigma=0.02*scale), [input_size, output_size])) + self.bias = Parameter(initializer("zeros", [output_size,])) self.dtype = dtype self.cast = P.Cast() @@ -194,8 +194,7 @@ class EmbeddingLookup(nn.Cell): super(EmbeddingLookup, self).__init__() self.vocab_size = config.vocab_size self.embedding_size = config.embedding_size - self.embedding_table = Parameter(initializer(TruncatedNormal(0.02), [self.vocab_size, self.embedding_size]), - name="embedding_table") + self.embedding_table = Parameter(initializer(TruncatedNormal(0.02), [self.vocab_size, self.embedding_size])) self.gather = P.GatherV2() self.shape = (-1, config.seq_length, config.embedding_size) def construct(self, input_ids): diff --git a/model_zoo/official/nlp/gpt/src/gpt_wrapcell.py b/model_zoo/official/nlp/gpt/src/gpt_wrapcell.py index dbcb9fd436..59858d3bc6 100644 --- a/model_zoo/official/nlp/gpt/src/gpt_wrapcell.py +++ b/model_zoo/official/nlp/gpt/src/gpt_wrapcell.py @@ -106,8 +106,7 @@ class GPTTrainOneStepWithLossScaleCell(nn.Cell): self.loss_scale = None self.loss_scaling_manager = scale_update_cell if scale_update_cell: - self.loss_scale = Parameter(Tensor(scale_update_cell.get_loss_scale(), dtype=mstype.float32), - name="loss_scale") + self.loss_scale = Parameter(Tensor(scale_update_cell.get_loss_scale(), dtype=mstype.float32)) @C.add_flags(has_effect=True) def construct(self, diff --git a/model_zoo/official/nlp/mass/src/transformer/embedding.py b/model_zoo/official/nlp/mass/src/transformer/embedding.py index 22887b0a3e..b44c3391c8 100644 --- a/model_zoo/official/nlp/mass/src/transformer/embedding.py +++ b/model_zoo/official/nlp/mass/src/transformer/embedding.py @@ -44,8 +44,7 @@ class EmbeddingLookup(nn.Cell): init_weight = np.random.normal(0, embed_dim ** -0.5, size=[vocab_size, embed_dim]).astype(np.float32) # 0 is Padding index, thus init it as 0. init_weight[0, :] = 0 - self.embedding_table = Parameter(Tensor(init_weight), - name='embedding_table') + self.embedding_table = Parameter(Tensor(init_weight)) self.expand = P.ExpandDims() self.gather = P.GatherV2() self.one_hot = P.OneHot() diff --git a/model_zoo/official/nlp/mass/src/transformer/transformer_for_train.py b/model_zoo/official/nlp/mass/src/transformer/transformer_for_train.py index 46f0154a33..da683752a5 100644 --- a/model_zoo/official/nlp/mass/src/transformer/transformer_for_train.py +++ b/model_zoo/official/nlp/mass/src/transformer/transformer_for_train.py @@ -277,8 +277,7 @@ class TransformerTrainOneStepWithLossScaleCell(nn.Cell): self.loss_scale = None self.loss_scaling_manager = scale_update_cell if scale_update_cell: - self.loss_scale = Parameter(Tensor(scale_update_cell.get_loss_scale(), dtype=mstype.float32), - name="loss_scale") + self.loss_scale = Parameter(Tensor(scale_update_cell.get_loss_scale(), dtype=mstype.float32)) self.add_flags(has_effect=True) def construct(self, diff --git a/model_zoo/official/nlp/prophetnet/src/transformer/embedding.py b/model_zoo/official/nlp/prophetnet/src/transformer/embedding.py index 22887b0a3e..b44c3391c8 100644 --- a/model_zoo/official/nlp/prophetnet/src/transformer/embedding.py +++ b/model_zoo/official/nlp/prophetnet/src/transformer/embedding.py @@ -44,8 +44,7 @@ class EmbeddingLookup(nn.Cell): init_weight = np.random.normal(0, embed_dim ** -0.5, size=[vocab_size, embed_dim]).astype(np.float32) # 0 is Padding index, thus init it as 0. init_weight[0, :] = 0 - self.embedding_table = Parameter(Tensor(init_weight), - name='embedding_table') + self.embedding_table = Parameter(Tensor(init_weight)) self.expand = P.ExpandDims() self.gather = P.GatherV2() self.one_hot = P.OneHot() diff --git a/model_zoo/official/nlp/tinybert/src/tinybert_for_gd_td.py b/model_zoo/official/nlp/tinybert/src/tinybert_for_gd_td.py index f6e816d55d..f8a1bb2263 100644 --- a/model_zoo/official/nlp/tinybert/src/tinybert_for_gd_td.py +++ b/model_zoo/official/nlp/tinybert/src/tinybert_for_gd_td.py @@ -243,8 +243,7 @@ class BertTrainWithLossScaleCell(nn.Cell): self.loss_scale = None self.loss_scaling_manager = scale_update_cell if scale_update_cell: - self.loss_scale = Parameter(Tensor(scale_update_cell.get_loss_scale(), dtype=mstype.float32), - name="loss_scale") + self.loss_scale = Parameter(Tensor(scale_update_cell.get_loss_scale(), dtype=mstype.float32)) @C.add_flags(has_effect=True) def construct(self, @@ -497,8 +496,7 @@ class BertEvaluationWithLossScaleCell(nn.Cell): self.loss_scale = None self.loss_scaling_manager = scale_update_cell if scale_update_cell: - self.loss_scale = Parameter(Tensor(scale_update_cell.get_loss_scale(), dtype=mstype.float32), - name="loss_scale") + self.loss_scale = Parameter(Tensor(scale_update_cell.get_loss_scale(), dtype=mstype.float32)) @C.add_flags(has_effect=True) def construct(self, diff --git a/model_zoo/official/nlp/tinybert/src/tinybert_model.py b/model_zoo/official/nlp/tinybert/src/tinybert_model.py index d802d4ba8a..5e31ee8e86 100644 --- a/model_zoo/official/nlp/tinybert/src/tinybert_model.py +++ b/model_zoo/official/nlp/tinybert/src/tinybert_model.py @@ -110,8 +110,7 @@ class EmbeddingLookup(nn.Cell): self.use_one_hot_embeddings = use_one_hot_embeddings self.embedding_table = Parameter(initializer (TruncatedNormal(initializer_range), - [vocab_size, embedding_size]), - name='embedding_table') + [vocab_size, embedding_size])) self.expand = P.ExpandDims() self.shape_flat = (-1,) self.gather = P.GatherV2() @@ -170,8 +169,7 @@ class EmbeddingPostprocessor(nn.Cell): self.embedding_table = Parameter(initializer (TruncatedNormal(initializer_range), [token_type_vocab_size, - embedding_size]), - name='embedding_table') + embedding_size])) self.shape_flat = (-1,) self.one_hot = P.OneHot() self.on_value = Tensor(1.0, mstype.float32) @@ -187,8 +185,7 @@ class EmbeddingPostprocessor(nn.Cell): self.full_position_embeddings = Parameter(initializer (TruncatedNormal(initializer_range), [max_position_embeddings, - embedding_size]), - name='full_position_embeddings') + embedding_size])) def construct(self, token_type_ids, word_embeddings): """embedding postprocessor""" @@ -317,8 +314,7 @@ class RelaPosEmbeddingsGenerator(nn.Cell): self.use_one_hot_embeddings = use_one_hot_embeddings self.embeddings_table = Parameter( initializer(TruncatedNormal(initializer_range), - [self.vocab_size, self.depth]), - name='embeddings_for_position') + [self.vocab_size, self.depth])) self.relative_positions_matrix = RelaPosMatrixGenerator(length=length, max_relative_position=max_relative_position) self.reshape = P.Reshape() diff --git a/model_zoo/official/nlp/transformer/src/transformer_for_train.py b/model_zoo/official/nlp/transformer/src/transformer_for_train.py index 3b677d0eee..4a7f083f71 100644 --- a/model_zoo/official/nlp/transformer/src/transformer_for_train.py +++ b/model_zoo/official/nlp/transformer/src/transformer_for_train.py @@ -291,8 +291,7 @@ class TransformerTrainOneStepWithLossScaleCell(nn.Cell): self.loss_scale = None self.loss_scaling_manager = scale_update_cell if scale_update_cell: - self.loss_scale = Parameter(Tensor(scale_update_cell.get_loss_scale(), dtype=mstype.float32), - name="loss_scale") + self.loss_scale = Parameter(Tensor(scale_update_cell.get_loss_scale(), dtype=mstype.float32)) @C.add_flags(has_effect=True) def construct(self, diff --git a/model_zoo/official/nlp/transformer/src/transformer_model.py b/model_zoo/official/nlp/transformer/src/transformer_model.py index 15e945e3db..c045783023 100644 --- a/model_zoo/official/nlp/transformer/src/transformer_model.py +++ b/model_zoo/official/nlp/transformer/src/transformer_model.py @@ -115,8 +115,7 @@ class EmbeddingLookup(nn.Cell): self.vocab_size = vocab_size self.embedding_size = embedding_size self.use_one_hot_embeddings = use_one_hot_embeddings - self.embedding_table = Parameter(normal_weight([vocab_size, embedding_size], embedding_size), - name='embedding_table') + self.embedding_table = Parameter(normal_weight([vocab_size, embedding_size], embedding_size)) self.expand = P.ExpandDims() self.shape_flat = (-1,) self.gather = P.GatherV2() diff --git a/model_zoo/official/recommend/ncf/src/ncf.py b/model_zoo/official/recommend/ncf/src/ncf.py index 5a93918d88..9be2fad4b5 100644 --- a/model_zoo/official/recommend/ncf/src/ncf.py +++ b/model_zoo/official/recommend/ncf/src/ncf.py @@ -47,14 +47,14 @@ class DenseLayer(nn.Cell): 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.weight = Parameter(initializer(weight_init, [out_channels, in_channels])) 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 = Parameter(initializer(bias_init, [out_channels])) self.matmul = P.MatMul(transpose_b=True) self.bias_add = P.BiasAdd() diff --git a/model_zoo/research/cv/ghostnet_quant/src/quant.py b/model_zoo/research/cv/ghostnet_quant/src/quant.py index 1a34d81bc5..b8d50700a8 100644 --- a/model_zoo/research/cv/ghostnet_quant/src/quant.py +++ b/model_zoo/research/cv/ghostnet_quant/src/quant.py @@ -35,10 +35,10 @@ class QuanConv(nn.Conv2d): self.x_upper_bound = Tensor(2 ** 8 - 1, ms.float32) self.w_lower_bound = Tensor(-2 ** 7 - 1, ms.float32) self.w_upper_bound = Tensor(2 ** 7, ms.float32) - self.scale_a = Parameter(initializer('ones', [1]), name='scale_a') + self.scale_a = Parameter(initializer('ones', [1])) self.scale_w = Parameter(initializer( - 'ones', [out_channels]), name='scale_w') - self.zp_a = Parameter(initializer('ones', [1]), name='zp_a') + 'ones', [out_channels])) + self.zp_a = Parameter(initializer('ones', [1])) def construct(self, in_data): r"""construct of QuantConv""" diff --git a/model_zoo/research/cv/ssd_ghostnet/src/ssd_ghostnet.py b/model_zoo/research/cv/ssd_ghostnet/src/ssd_ghostnet.py index f109d115aa..bbbb8f4f24 100644 --- a/model_zoo/research/cv/ssd_ghostnet/src/ssd_ghostnet.py +++ b/model_zoo/research/cv/ssd_ghostnet/src/ssd_ghostnet.py @@ -119,12 +119,12 @@ class DepthwiseConv(nn.Cell): self.bias_add = P.BiasAdd() weight_shape = [channel_multiplier, in_planes, *self.kernel_size] self.weight = Parameter(initializer( - 'ones', weight_shape), name='weight') + 'ones', weight_shape)) if has_bias: bias_shape = [channel_multiplier * in_planes] self.bias = Parameter(initializer( - 'zeros', bias_shape), name='bias') + 'zeros', bias_shape)) else: self.bias = None diff --git a/model_zoo/research/cv/tinynet/src/tinynet.py b/model_zoo/research/cv/tinynet/src/tinynet.py index 2634802e30..70e34edaea 100755 --- a/model_zoo/research/cv/tinynet/src/tinynet.py +++ b/model_zoo/research/cv/tinynet/src/tinynet.py @@ -499,7 +499,7 @@ class DepthWiseConv(nn.Cell): group=in_planes) self.weight = Parameter(initializer(weight_init, - [in_planes*1, 1, kernel_size, kernel_size]), name='depthwise_weight') + [in_planes*1, 1, kernel_size, kernel_size])) else: self.depthwise_conv = P.DepthwiseConv2dNative(channel_multiplier=1, @@ -508,7 +508,7 @@ class DepthWiseConv(nn.Cell): pad=int(kernel_size/2)) self.weight = Parameter(initializer(weight_init, - [1, in_planes, kernel_size, kernel_size]), name='depthwise_weight') + [1, in_planes, kernel_size, kernel_size])) def construct(self, x): x = self.depthwise_conv(x, self.weight) diff --git a/model_zoo/research/nlp/dscnn/src/ds_cnn.py b/model_zoo/research/nlp/dscnn/src/ds_cnn.py index 34a6fe24d6..f7daaa5bfa 100644 --- a/model_zoo/research/nlp/dscnn/src/ds_cnn.py +++ b/model_zoo/research/nlp/dscnn/src/ds_cnn.py @@ -31,11 +31,11 @@ class DepthWiseConv(nn.Cell): self.bias_add = P.BiasAdd() weight_shape = [channel_multiplier, in_planes, kernel_size[0], kernel_size[1]] - self.weight = Parameter(initializer('ones', weight_shape), name='weight') + self.weight = Parameter(initializer('ones', weight_shape)) if has_bias: bias_shape = [channel_multiplier * in_planes] - self.bias = Parameter(initializer('zeros', bias_shape), name='bias') + self.bias = Parameter(initializer('zeros', bias_shape)) else: self.bias = None