!17618 fix bug: resnet50 thor train failed in master

From: @sl_wang
Reviewed-by: @guoqi1024,@kisnwang
Signed-off-by: @guoqi1024

mindspore/nn/optim/thor.py

@@ -246,7 +246,7 @@ class ThorGpu(Optimizer):
         self.matrix_g = ParameterTuple(self.matrix_g)
         self.weight_decay = weight_decay
         self.decay_flags = tuple(decay_filter(x) for x in self.parameters)
-        self.update_gradient = P.UpdateThorGradient(split_dim=split_dim)
+        self.update_gradient = P.UpdateThorGradient(split_dim=self.split_dim)
         self.enable_clip_grad = enable_clip_grad
         self.frequency = frequency
         self._define_gpu_reducer(split_indices)
@@ -271,9 +271,9 @@ class ThorGpu(Optimizer):
         self.cast = P.Cast()
         self.sqrt = P.Sqrt()
         self.eye = P.Eye()
-        split_dim = 128
+        self.split_dim = 128
         self.embedding_cholesky = P.CholeskyTrsm()
-        self.cholesky = P.CholeskyTrsm(split_dim=split_dim)
+        self.cholesky = P.CholeskyTrsm(split_dim=self.split_dim)
         self.vector_matmul = P.BatchMatMul(transpose_a=True)
         self.reduce_sum = P.ReduceSum(keep_dims=False)
         self.inv = P.Reciprocal()
@@ -312,7 +312,7 @@ class ThorGpu(Optimizer):
                 if layer_type == Conv:
                     matrix_a_dim = in_channels * weight_shape[2] * weight_shape[3]
                 matrix_g_dim = out_channels
-                matrix_a_shape, matrix_g_shape = caculate_matmul_shape(matrix_a_dim, matrix_g_dim, split_dim)
+                matrix_a_shape, matrix_g_shape = caculate_matmul_shape(matrix_a_dim, matrix_g_dim, self.split_dim)
                 matrix_a_inv = Parameter(np.zeros(matrix_a_shape).astype(np.float32),
                                          name='matrix_a_inv_' + str(self.thor_layer_count), requires_grad=False)
                 matrix_g_inv = Parameter(np.zeros(matrix_g_shape).astype(np.float32),

model_zoo/official/cv/resnet_thor/export.py

@@ -17,7 +17,7 @@ import argparse
 import numpy as np
 
 from mindspore import context, Tensor, load_checkpoint, load_param_into_net, export
-from src.resnet_thor import resnet50 as resnet
+from src.resnet import resnet50 as resnet
 from src.config import config
 
 parser = argparse.ArgumentParser(description='checkpoint export')

model_zoo/official/cv/resnet_thor/mindspore_hub_conf.py

@@ -13,7 +13,7 @@
 # limitations under the License.
 # ============================================================================
 """hub config."""
-from src.resnet_thor import resnet50
+from src.resnet import resnet50
 
 def create_network(name, *args, **kwargs):
     if name == 'resnet50_thor':

model_zoo/official/nlp/bert/run_pretrain.py

@@ -272,7 +272,11 @@ def run_pretrain():
                                                    accumulation_steps=accumulation_steps,
                                                    enable_global_norm=enable_global_norm)
     else:
-        net_with_grads = BertTrainOneStepCell(net_with_loss, optimizer=optimizer, enable_clip_grad=False)
+        net_with_grads = BertTrainOneStepCell(net_with_loss, optimizer=optimizer, enable_clip_grad=True)
+        if cfg.optimizer == "Thor":
+            net_with_grads = BertTrainOneStepCell(net_with_loss, optimizer=optimizer, sens=cfg.Thor.loss_scale,
+                                                  enable_clip_grad=False)
+
 
     model = Model(net_with_grads)
     model = ConvertModelUtils().convert_to_thor_model(model, network=net_with_grads, optimizer=optimizer)

model_zoo/official/nlp/bert/src/bert_for_pre_training.py

@@ -269,10 +269,10 @@ class BertTrainOneStepCell(nn.TrainOneStepCell):
         network (Cell): The training network. Note that loss function should have been added.
         optimizer (Optimizer): Optimizer for updating the weights.
         sens (Number): The adjust parameter. Default: 1.0.
-        enable_clip_grad (boolean): If True, clip gradients in BertTrainOneStepCell. Default: False.
+        enable_clip_grad (boolean): If True, clip gradients in BertTrainOneStepCell. Default: True.
     """
 
-    def __init__(self, network, optimizer, sens=1.0, enable_clip_grad=False):
+    def __init__(self, network, optimizer, sens=1.0, enable_clip_grad=True):
         super(BertTrainOneStepCell, self).__init__(network, optimizer, sens)
         self.cast = P.Cast()
         self.hyper_map = C.HyperMap()

model_zoo/official/nlp/bert_thor/mindspore_hub_conf.py

@@ -19,7 +19,6 @@ from src.bert_model import BertModel
 from src.bert_model import BertConfig
 import mindspore.common.dtype as mstype
 bert_net_cfg = BertConfig(
-    batch_size=12,
     seq_length=512,
     vocab_size=30522,
     hidden_size=1024,
@@ -33,11 +32,8 @@ bert_net_cfg = BertConfig(
     type_vocab_size=2,
     initializer_range=0.02,
     use_relative_positions=False,
-    input_mask_from_dataset=True,
-    token_type_ids_from_dataset=True,
     dtype=mstype.float32,
-    compute_type=mstype.float16,
-    enable_fused_layernorm=True
+    compute_type=mstype.float16
 )
 def create_network(name, *args, **kwargs):
     '''