modify model_zoo resnext and resnet bug

model_zoo/official/cv/resnet/README.md

@@ -380,13 +380,13 @@ Please follow the instructions in the link [GPU-Multi-Host](https://www.mindspor
 - Parameter server training Ascend example
 
 ```bash
-bash run_parameter_server_train.sh [RANK_TABLE_FILE] [DATASET_PATH] [PRETRAINED_CKPT_PATH](optional)
+bash run_parameter_server_train.sh [RANK_TABLE_FILE] [DATASET_PATH] [CONFIG_PATH] [PRETRAINED_CKPT_PATH](optional)
 ```
 
 - Parameter server training GPU example
 
 ```bash
-bash run_parameter_server_train_gpu.sh [DATASET_PATH] [PRETRAINED_CKPT_PATH](optional)
+bash run_parameter_server_train_gpu.sh [DATASET_PATH] [CONFIG_PATH] [PRETRAINED_CKPT_PATH](optional)
 ```
 
 #### Evaluation while training

model_zoo/official/cv/resnet/resnet50_imagenet2012_Thor_config.yaml

@@ -16,7 +16,7 @@ checkpoint_file_path: ""
 
 # ==============================================================================
 # Training options
-optimizer: "Momentum"
+optimizer: "Thor"
 infer_label: ""
 class_num: 1001
 batch_size: 32

model_zoo/official/cv/resnet/train.py

@@ -24,7 +24,7 @@ from mindspore.train.callback import ModelCheckpoint, CheckpointConfig, LossMoni
 from mindspore.nn.loss import SoftmaxCrossEntropyWithLogits
 from mindspore.train.loss_scale_manager import FixedLossScaleManager
 from mindspore.train.serialization import load_checkpoint, load_param_into_net
-from mindspore.communication.management import init
+from mindspore.communication.management import init, get_rank
 from mindspore.common import set_seed
 from mindspore.parallel import set_algo_parameters
 import mindspore.nn as nn
@@ -177,6 +177,18 @@ def run_eval(target, model, ckpt_save_dir, cb):
                                metrics_name="acc")
         cb += [eval_cb]
 
+
+def set_save_ckpt_dir():
+    """set save ckpt dir"""
+    ckpt_save_dir = os.path.join(config.output_path, config.checkpoint_path)
+    if config.enable_modelarts and config.run_distribute:
+        ckpt_save_dir = ckpt_save_dir + "ckpt_" + str(get_rank_id()) + "/"
+    else:
+        if config.run_distribute:
+            ckpt_save_dir = ckpt_save_dir + "ckpt_" + str(get_rank()) + "/"
+    return ckpt_save_dir
+
+
 @moxing_wrapper()
 def train_net():
     """train net"""
@@ -234,8 +246,7 @@ def train_net():
     time_cb = TimeMonitor(data_size=step_size)
     loss_cb = LossMonitor()
     cb = [time_cb, loss_cb]
-    ckpt_save_dir = os.path.join(config.output_path, config.checkpoint_path)
-    ckpt_save_dir = ckpt_save_dir + "ckpt_" + str(get_rank_id()) + "/"
+    ckpt_save_dir = set_save_ckpt_dir()
     if config.save_checkpoint:
         config_ck = CheckpointConfig(save_checkpoint_steps=config.save_checkpoint_epochs * step_size,
                                      keep_checkpoint_max=config.keep_checkpoint_max)

model_zoo/official/cv/resnext/README.md

@@ -221,7 +221,7 @@ or shell script:
 
 ```script
 # Evaluation
-sh run_eval.sh DEVICE_ID DATA_PATH CHECKPOINT_FILE_PATH DEVICE_TARGET
+sh scripts/run_eval.sh DEVICE_ID DATA_PATH CHECKPOINT_FILE_PATH DEVICE_TARGET
 ```
 
 PLATFORM is Ascend or GPU, default is Ascend.

model_zoo/official/cv/resnext/README_CN.md

@@ -229,7 +229,7 @@ python eval.py --data_path ~/imagenet/val/ --device_target Ascend --checkpoint_f
 
 ```shell
 # 评估
-sh run_eval.sh DEVICE_ID DATA_PATH CHECKPOINT_FILE_PATH PLATFORM
+sh scripts/run_eval.sh DEVICE_ID DATA_PATH CHECKPOINT_FILE_PATH PLATFORM
 ```
 
 DEVICE_TARGET is Ascend or GPU, default is Ascend.

model_zoo/official/cv/resnext/eval.py

@@ -127,8 +127,9 @@ def set_graph_kernel_context(device_target):
     if device_target == "GPU":
         context.set_context(enable_graph_kernel=True)
 
+
 @moxing_wrapper()
-def test(cloud_args=None):
+def test():
     """test"""
     set_parameters()
     context.set_context(mode=context.GRAPH_MODE, enable_auto_mixed_precision=True,
@@ -150,7 +151,7 @@ def test(cloud_args=None):
     if os.path.isdir(config.checkpoint_file_path):
         models = list(glob.glob(os.path.join(config.checkpoint_file_path, '*.ckpt')))
         print(models)
-        if config.graph_ckpt:
+        if config.checkpoint_file_path:
             f = lambda x: -1 * int(os.path.splitext(os.path.split(x)[-1])[0].split('-')[-1].split('_')[0])
         else:
             f = lambda x: -1 * int(os.path.splitext(os.path.split(x)[-1])[0].split('_')[-1])

model_zoo/official/cv/resnext/scripts/run_eval.sh

@@ -17,13 +17,11 @@
 export DEVICE_ID=$1
 DATA_DIR=$2
 PATH_CHECKPOINT=$3
-PLATFORM=Ascend
-if [ $# == 4 ]
-then
-  PLATFORM=$4
-fi
+PLATFORM=$4
+
 
 python eval.py  \
     --checkpoint_file_path=$PATH_CHECKPOINT \
     --device_target=$PLATFORM \
-    --data_path=$DATA_DIR > log.txt 2>&1 &
+    --data_path=$DATA_DIR \
+    --device_target=$PLATFORM > log.txt 2>&1 &

tests/st/model_zoo_tests/resnet50/test_resnet50_cifar10.py

@@ -65,12 +65,16 @@ def test_resnet50_cifar10_gpu():
     utils.exec_sed_command(old_list, new_list, os.path.join(cur_model_path, "train.py"))
     dataset_path = os.path.join(utils.data_root, "cifar-10-batches-bin")
     config_path = os.path.join(cur_model_path, "resnet50_cifar10_config.yaml")
+    os.system("nvidia-smi")
     exec_network_shell = "cd resnet/scripts; sh run_distribute_train_gpu.sh {} {}" \
         .format(dataset_path, config_path)
     logger.warning("cmd [{}] is running...".format(exec_network_shell))
     os.system(exec_network_shell)
     cmd = "ps -ef | grep python | grep train.py | grep -v grep"
     ret = utils.process_check(100, cmd)
+    if not ret:
+        cmd = "{} | awk -F' ' '{{print $2}}' | xargs kill -9".format(cmd)
+        os.system(cmd)
     assert ret
     log_file = os.path.join(cur_model_path, "scripts/train_parallel/log")
     pattern = r"per step time: ([\d\.]+) ms"