!5558 thor optimizer training parameters for GPU

Merge pull request !5558 from wangmin0104/master

model_zoo/official/cv/resnet_thor/scripts/run_distribute_train.sh

@@ -32,13 +32,21 @@ then
 exit 1
 fi
 
-BASE_PATH=$(cd "`dirname $0`" || exit; pwd)
-cd $BASE_PATH/../ || exit
+get_real_path(){
+  if [ "${1:0:1}" == "/" ]; then
+    echo "$1"
+  else
+    echo "$(realpath -m $PWD/$1)"
+  fi
+}
+
+PATH1=$(get_real_path $1)
+PATH2=$(get_real_path $2)
 
 ulimit -u unlimited
 export DEVICE_NUM=$3
 export RANK_SIZE=$3
-export RANK_TABLE_FILE=$1
+export RANK_TABLE_FILE=$PATH1
 
 for((i=0; i<${DEVICE_NUM}; i++))
 do
@@ -46,12 +54,12 @@ do
     export RANK_ID=$i
     rm -rf ./train_parallel$i
     mkdir ./train_parallel$i
-    cp *.py ./train_parallel$i
-    cp -r ./src ./train_parallel$i
+    cp ../*.py ./train_parallel$i
+    cp -r ../src ./train_parallel$i
     cd ./train_parallel$i || exit
     echo "start training for rank $RANK_ID, device $DEVICE_ID"
 
     env > env.log
-    python train.py --run_distribute=True --device_num=$DEVICE_NUM --dataset_path=$2 > log 2>&1 &
+    python train.py --run_distribute=True --device_num=$DEVICE_NUM --dataset_path=$PATH2 > log 2>&1 &
     cd ..
 done

model_zoo/official/cv/resnet_thor/src/config.py

@@ -46,17 +46,17 @@ config_gpu = ed({
     "loss_scale": 128,
     "momentum": 0.9,
     "weight_decay": 5e-4,
-    "epoch_size": 45,
+    "epoch_size": 40,
     "save_checkpoint": True,
     "save_checkpoint_epochs": 1,
     "keep_checkpoint_max": 15,
     "save_checkpoint_path": "./",
     "use_label_smooth": True,
     "label_smooth_factor": 0.1,
-    "lr_init": 0.04,
-    "lr_decay": 5,
-    "lr_end_epoch": 58,
-    "damping_init": 0.02,
-    "damping_decay": 0.87,
+    "lr_init": 0.05672,
+    "lr_decay": 4.9687,
+    "lr_end_epoch": 50,
+    "damping_init": 0.02345,
+    "damping_decay": 0.5467,
     "frequency": 834,
 })

model_zoo/official/cv/resnet_thor/train.py

@@ -109,6 +109,7 @@ if __name__ == '__main__':
             init()
             context.set_auto_parallel_context(device_num=get_group_size(), parallel_mode=ParallelMode.DATA_PARALLEL,
                                               mirror_mean=True)
+            auto_parallel_context().set_all_reduce_fusion_split_indices([107])
             ckpt_save_dir = config.save_checkpoint_path + "ckpt_" + str(get_rank()) + "/"
 
     # create dataset