add ls training for dygraph

ppcls/modeling/architectures/__init__.py

@@ -14,3 +14,5 @@
 
 from .resnet_name import *
 from .dpn import DPN68
+from .densenet import DenseNet121
+from .hrnet import HRNet_W18_C
\ No newline at end of file

ppcls/modeling/architectures/densenet.py

+# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#    http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
 import numpy as np
 import paddle
 import paddle.fluid as fluid
@@ -268,26 +286,26 @@ class DenseNet(fluid.dygraph.Layer):
         return y
 
 
-def DenseNet121():
-    model = DenseNet(layers=121)
+def DenseNet121(**args):
+    model = DenseNet(layers=121, **args)
     return model
 
 
-def DenseNet161():
-    model = DenseNet(layers=161)
+def DenseNet161(**args):
+    model = DenseNet(layers=161, **args)
     return model
 
 
-def DenseNet169():
-    model = DenseNet(layers=169)
+def DenseNet169(**args):
+    model = DenseNet(layers=169, **args)
     return model
 
 
-def DenseNet201():
-    model = DenseNet(layers=201)
+def DenseNet201(**args):
+    model = DenseNet(layers=201, **args)
     return model
 
 
-def DenseNet264():
-    model = DenseNet(layers=264)
+def DenseNet264(**args):
+    model = DenseNet(layers=264, **args)
     return model

ppcls/modeling/architectures/dpn.py

+# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#    http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
 import numpy as np
 import sys
 import paddle
@@ -386,26 +404,26 @@ class DPN(fluid.dygraph.Layer):
         return net_arg
 
 
-def DPN68():
-    model = DPN(layers=68)
+def DPN68(**args):
+    model = DPN(layers=68, **args)
     return model
 
 
-def DPN92():
-    model = DPN(layers=92)
+def DPN92(**args):
+    model = DPN(layers=92, **args)
     return model
 
 
-def DPN98():
-    model = DPN(layers=98)
+def DPN98(**args):
+    model = DPN(layers=98, **args)
     return model
 
 
-def DPN107():
-    model = DPN(layers=107)
+def DPN107(**args):
+    model = DPN(layers=107, **args)
     return model
 
 
-def DPN131():
-    model = DPN(layers=131)
-    return model
\ No newline at end of file
+def DPN131(**args):
+    model = DPN(layers=131, **args)
+    return model

ppcls/modeling/architectures/hrnet.py

+# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#    http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
 import numpy as np
 import paddle
 import paddle.fluid as fluid
@@ -647,81 +665,81 @@ class HRNet(fluid.dygraph.Layer):
         return y
 
 
-def HRNet_W18_C():
-    model = HRNet(width=18)
+def HRNet_W18_C(**args):
+    model = HRNet(width=18, **args)
     return model
 
 
-def HRNet_W30_C():
-    model = HRNet(width=30)
+def HRNet_W30_C(**args):
+    model = HRNet(width=30, **args)
     return model
 
 
-def HRNet_W32_C():
-    model = HRNet(width=32)
+def HRNet_W32_C(**args):
+    model = HRNet(width=32, **args)
     return model
 
 
-def HRNet_W40_C():
-    model = HRNet(width=40)
+def HRNet_W40_C(**args):
+    model = HRNet(width=40, **args)
     return model
 
 
-def HRNet_W44_C():
-    model = HRNet(width=44)
+def HRNet_W44_C(**args):
+    model = HRNet(width=44, **args)
     return model
 
 
-def HRNet_W48_C():
-    model = HRNet(width=48)
+def HRNet_W48_C(**args):
+    model = HRNet(width=48, **args)
     return model
 
 
-def HRNet_W60_C():
-    model = HRNet(width=60)
+def HRNet_W60_C(**args):
+    model = HRNet(width=60, **args)
     return model
 
 
-def HRNet_W64_C():
-    model = HRNet(width=64)
+def HRNet_W64_C(**args):
+    model = HRNet(width=64, **args)
     return model
 
 
-def SE_HRNet_W18_C():
-    model = HRNet(width=18, has_se=True)
+def SE_HRNet_W18_C(**args):
+    model = HRNet(width=18, has_se=True, **args)
     return model
 
 
-def SE_HRNet_W30_C():
-    model = HRNet(width=30, has_se=True)
+def SE_HRNet_W30_C(**args):
+    model = HRNet(width=30, has_se=True, **args)
     return model
 
 
-def SE_HRNet_W32_C():
-    model = HRNet(width=32, has_se=True)
+def SE_HRNet_W32_C(**args):
+    model = HRNet(width=32, has_se=True, **args)
     return model
 
 
-def SE_HRNet_W40_C():
-    model = HRNet(width=40, has_se=True)
+def SE_HRNet_W40_C(**args):
+    model = HRNet(width=40, has_se=True, **args)
     return model
 
 
-def SE_HRNet_W44_C():
-    model = HRNet(width=44, has_se=True)
+def SE_HRNet_W44_C(**args):
+    model = HRNet(width=44, has_se=True, **args)
     return model
 
 
-def SE_HRNet_W48_C():
-    model = HRNet(width=48, has_se=True)
+def SE_HRNet_W48_C(**args):
+    model = HRNet(width=48, has_se=True, **args)
     return model
 
 
-def SE_HRNet_W60_C():
-    model = HRNet(width=60, has_se=True)
+def SE_HRNet_W60_C(**args):
+    model = HRNet(width=60, has_se=True, **args)
     return model
 
 
-def SE_HRNet_W64_C():
-    model = HRNet(width=64, has_se=True)
+def SE_HRNet_W64_C(**args):
+    model = HRNet(width=64, has_se=True, **args)
     return model

tools/program.py

@@ -338,20 +338,29 @@ def run(dataloader, config, net, optimizer=None, epoch=0, mode='train'):
 
     Returns:
     """
-    topk_name = 'top{}'.format(config.topk)
-    metric_list = OrderedDict([
-        ("loss", AverageMeter('loss', '7.4f')),
-        ("top1", AverageMeter('top1', '.4f')),
-        (topk_name, AverageMeter(topk_name, '.4f')),
-        ("lr", AverageMeter(
-            'lr', 'f', need_avg=False)),
-        ("batch_time", AverageMeter('elapse', '.3f')),
-    ])
+    use_mix = config.get("use_mix", False) and mode == "train"
+    if use_mix:
+        metric_list = OrderedDict([
+            ("loss", AverageMeter('loss', '7.4f')),
+            ("lr", AverageMeter(
+                'lr', 'f', need_avg=False)),
+            ("batch_time", AverageMeter('elapse', '.3f')),
+        ])
+    else:
+        topk_name = 'top{}'.format(config.topk)
+        metric_list = OrderedDict([
+            ("loss", AverageMeter('loss', '7.4f')),
+            ("top1", AverageMeter('top1', '.4f')),
+            (topk_name, AverageMeter(topk_name, '.4f')),
+            ("lr", AverageMeter(
+                'lr', 'f', need_avg=False)),
+            ("batch_time", AverageMeter('elapse', '.3f')),
+        ])
 
     tic = time.time()
     for idx, batch in enumerate(dataloader()):
         bs = len(batch[0])
-        feeds = create_feeds(batch, config.get("use_mix", False))
+        feeds = create_feeds(batch, use_mix)
         fetchs = compute(feeds, net, config, mode)
         if mode == 'train':
             avg_loss = net.scale_loss(fetchs['loss'])