!22225 fix midas pynative issue

Merge pull request !22225 from JichenZhao/master

model_zoo/official/cv/faster_rcnn/default_config.yaml

@@ -121,7 +121,7 @@ batch_size: 2
 loss_scale: 256
 momentum: 0.91
 weight_decay: 0.00001
-epoch_size: 20
+epoch_size: 12
 save_checkpoint: True
 save_checkpoint_epochs: 1
 keep_checkpoint_max: 5
@@ -162,7 +162,7 @@ device_num: 1
 rank_id: 0
 image_dir: ''
 anno_path: ''
-backbone: 'resnet_v1_50'
+backbone: 'resnet_v1.5_50'
 
 # eval.py FasterRcnn evaluation
 ann_file: '/cache/data/annotations/instances_val2017.json'

model_zoo/official/cv/faster_rcnn/train.py

@@ -44,6 +44,7 @@ if config.backbone in ("resnet_v1.5_50", "resnet_v1_101", "resnet_v1_152"):
     from src.FasterRcnn.faster_rcnn_resnet import Faster_Rcnn_Resnet
 elif config.backbone == "resnet_v1_50":
     from src.FasterRcnn.faster_rcnn_resnet50v1 import Faster_Rcnn_Resnet
+    config.epoch_size = 20
 
 if config.device_target == "GPU":
     context.set_context(enable_graph_kernel=True)

model_zoo/research/cv/midas/midas_train.py

@@ -21,6 +21,7 @@ from mindspore import nn
 from mindspore import Tensor
 from mindspore.context import ParallelMode
 import mindspore.dataset as ds
+from mindspore.common import set_seed
 from mindspore.train.serialization import load_checkpoint, load_param_into_net
 from mindspore.train.model import Model
 from mindspore.train.callback import LossMonitor, TimeMonitor, ModelCheckpoint, CheckpointConfig
@@ -29,6 +30,8 @@ from src.midas_net import MidasNet, Loss, NetwithCell
 from src.utils import loadImgDepth
 from src.config import config
 
+set_seed(1)
+ds.config.set_seed(1)
 
 def dynamic_lr(num_epoch_per_decay, total_epochs, steps_per_epoch, lr, end_lr):
     """dynamic learning rate generator"""
@@ -78,8 +81,7 @@ def train(mixdata_path):
                             max_call_depth=10000)
         context.set_auto_parallel_context(parallel_mode=ParallelMode.DATA_PARALLEL,
                                           gradients_mean=True,
-                                          device_num=device_num,
-                                          parameter_broadcast=True
+                                          device_num=device_num
                                           )
         init()
         local_data_path = config.train_data_dir + '/data'

model_zoo/research/cv/midas/scripts/run_distribute_train.sh

@@ -24,7 +24,7 @@ set -e
 
 RANK_SIZE=$1
 export RANK_SIZE
-
+export HCCL_CONNECT_TIMEOUT=600
 EXEC_PATH=$(pwd)
 echo "$EXEC_PATH"