!10394 Modify resnet50_adv_pruning net

From: @zhanghuiyao
Reviewed-by: @linqingke,@oacjiewen
Signed-off-by: @linqingke

model_zoo/research/cv/resnet50_adv_pruning/Readme.md

@@ -58,6 +58,7 @@ Dataset used: [Oxford-IIIT Pet](https://www.robots.ox.ac.uk/~vgg/data/pets/)
 ```
 
 ## [Training process](#contents)
+
 To Be Done
 
 ## [Eval process](#contents)
@@ -66,11 +67,10 @@ To Be Done
 
 After installing MindSpore via the official website, you can start evaluation as follows:
 
-
 ### Launch
 
-```
+```bash
-# infer example
+  # infer example
 
   Ascend: python eval.py --dataset_path ~/Pets/test.mindrecord --platform Ascend --checkpoint_path [CHECKPOINT_PATH]
   GPU: python eval.py --dataset_path ~/Pets/test.mindrecord --platform GPU --checkpoint_path [CHECKPOINT_PATH]
@@ -80,18 +80,18 @@ After installing MindSpore via the official website, you can start evaluation as
 
 ### Result
 
-```
+```python
 result: {'acc': 0.8023984736985554} ckpt= ./resnet50-imgnet-0.65x-80.24.ckpt
-
 ```
 
 # [Model Description](#contents)
 
 ## [Performance](#contents)
 
-#### Evaluation Performance
+### Evaluation Performance
+
+#### ResNet50-0.65x on ImageNet2012
 
-###### ResNet50-0.65x on ImageNet2012
 | Parameters                 |                                        |
 | -------------------------- | -------------------------------------- |
 | Model Version              | ResNet50-0.65x                              |
@@ -102,7 +102,8 @@ result: {'acc': 0.8023984736985554} ckpt= ./resnet50-imgnet-0.65x-80.24.ckpt
 | FLOPs (G) | 2.1 |
 | Accuracy (Top1) | 75.80 |
 
-###### ResNet50-0.65x on Oxford-IIIT Pet
+#### ResNet50-0.65x on Oxford-IIIT Pet
+
 | Parameters                 |                                        |
 | -------------------------- | -------------------------------------- |
 | Model Version              | ResNet50-0.65x                               |

model_zoo/research/cv/resnet50_adv_pruning/eval.py

@@ -26,7 +26,7 @@ from mindspore.train.serialization import load_checkpoint, load_param_into_net
 from mindspore.common import dtype as mstype
 
 from src.pet_dataset import create_dataset
-from src.config import config_ascend, config_gpu
+from src.config import cfg
 from src.resnet_imgnet import resnet50
 
 
@@ -40,14 +40,12 @@ args_opt = parser.parse_args()
 
 
 if __name__ == '__main__':
-    config_platform = None
+    config_platform = cfg
     if args_opt.platform == "Ascend":
-        config_platform = config_ascend
         device_id = int(os.getenv('DEVICE_ID'))
         context.set_context(mode=context.GRAPH_MODE, device_target="Ascend",
                             device_id=device_id, save_graphs=False)
     elif args_opt.platform == "GPU":
-        config_platform = config_gpu
         context.set_context(mode=context.GRAPH_MODE,
                             device_target="GPU", save_graphs=False)
     else:
@@ -55,12 +53,6 @@ if __name__ == '__main__':
 
     loss = nn.SoftmaxCrossEntropyWithLogits(sparse=True, reduction='mean')
 
-    if args_opt.platform == "Ascend":
-        net.to_float(mstype.float16)
-        for _, cell in net.cells_and_names():
-            if isinstance(cell, nn.Dense):
-                cell.to_float(mstype.float32)
-
     dataset = create_dataset(dataset_path=args_opt.dataset_path,
                              do_train=False,
                              config=config_platform,
@@ -76,6 +68,13 @@ if __name__ == '__main__':
 
     net = resnet50(
         rate=0.65, class_num=config_platform.num_classes, index=index)
+
+    if args_opt.platform == "Ascend":
+        net.to_float(mstype.float16)
+        for _, cell in net.cells_and_names():
+            if isinstance(cell, nn.Dense):
+                cell.to_float(mstype.float32)
+
     if args_opt.checkpoint_path:
         param_dict = load_checkpoint(args_opt.checkpoint_path)
         load_param_into_net(net, param_dict)

model_zoo/research/cv/resnet50_adv_pruning/src/config.py

@@ -17,25 +17,7 @@ network config setting, will be used in train.py and eval.py
 """
 from easydict import EasyDict as ed
 
-config_ascend = ed({
+cfg = ed({
-    "num_classes": 438,
-    "image_height": 224,
-    "image_width": 224,
-    "batch_size": 256,
-    "epoch_size": 200,
-    "warmup_epochs": 1,
-    "lr": 0.02,
-    "momentum": 0.9,
-    "weight_decay": 4e-5,
-    "label_smooth": 0.1,
-    "loss_scale": 1024,
-    "save_checkpoint": True,
-    "save_checkpoint_epochs": 5,
-    "keep_checkpoint_max": 200,
-    "save_checkpoint_path": "./checkpoint",
-})
-
-config_gpu = ed({
     "num_classes": 37,
     "image_height": 224,
     "image_width": 224,

model_zoo/research/cv/resnet50_adv_pruning/src/pet_dataset.py

@@ -18,10 +18,11 @@ create train or eval dataset.
 import os
 import mindspore.common.dtype as mstype
 import mindspore.dataset.engine as de
-import mindspore.dataset.transforms.vision.c_transforms as C
+import mindspore.dataset.vision.c_transforms as C
-import mindspore.dataset.transforms.vision.py_transforms as P
+import mindspore.dataset.vision.py_transforms as P
 import mindspore.dataset.transforms.c_transforms as C2
-from mindspore.dataset.transforms.vision import Inter
+import mindspore.dataset.transforms.py_transforms as P2
+from mindspore.dataset.vision import Inter
 
 
 def create_dataset(dataset_path, do_train, config, platform, repeat_num=1, batch_size=100):
@@ -78,13 +79,13 @@ def create_dataset(dataset_path, do_train, config, platform, repeat_num=1, batch
     center_crop_p = P.CenterCrop(224)
     totensor = P.ToTensor()
     normalize_p = P.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225))
-    composeop = P.ComposeOp(
+    composeop = P2.Compose(
         [decode_p, resize_p, center_crop_p, totensor, normalize_p])
     if do_train:
         trans = [resize_crop_op, horizontal_flip_op, color_op,
                  rescale_op, normalize_op, change_swap_op]
     else:
-        trans = composeop()
+        trans = composeop
     type_cast_op = C2.TypeCast(mstype.int32)
 
     ds = ds.map(input_columns="image", operations=trans,

model_zoo/research/cv/resnet50_adv_pruning/src/resnet_imgnet.py

@@ -327,21 +327,21 @@ class ResNet(nn.Cell):
         for _, m in self.cells_and_names():
             if isinstance(m, (nn.Conv2d)):
                 n = m.kernel_size[0] * m.kernel_size[1] * m.out_channels
-                m.weight.set_parameter_data(Tensor(np.random.normal(0, np.sqrt(2. / n),
+                m.weight.set_data(Tensor(np.random.normal(0, np.sqrt(2. / n),
                                                           m.weight.data.shape).astype("float32")))
                 if m.bias is not None:
-                    m.bias.set_parameter_data(
+                    m.bias.set_data(
                         Tensor(np.zeros(m.bias.data.shape, dtype="float32")))
             elif isinstance(m, nn.BatchNorm2d):
-                m.gamma.set_parameter_data(
+                m.gamma.set_data(
                     Tensor(np.ones(m.gamma.data.shape, dtype="float32")))
-                m.beta.set_parameter_data(
+                m.beta.set_data(
                     Tensor(np.zeros(m.beta.data.shape, dtype="float32")))
             elif isinstance(m, nn.Dense):
-                m.weight.set_parameter_data(Tensor(np.random.normal(
+                m.weight.set_data(Tensor(np.random.normal(
                     0, 0.01, m.weight.data.shape).astype("float32")))
                 if m.bias is not None:
-                    m.bias.set_parameter_data(
+                    m.bias.set_data(
                         Tensor(np.zeros(m.bias.data.shape, dtype="float32")))