tmp saving

Author: Yibing Liu

fluid/image_classification/inception_v4.py

@@ -2,24 +2,25 @@
 import paddle.fluid as fluid
 
 
-def inception_v4(image, label):
+def inception_v4(img, class_dim):
 
-    tmp = stem(input=image)
+    tmp = stem(input=img)
     for i in range(0, 4):
         tmp = inception_A(input=tmp, depth=i)
     tmp = reduction_A(input=tmp)
-
+    '''
     for i in range(0, 7):
         tmp = inception_B(input=tmp, depth=i)
     reduction_B(input=tmp)
 
     for i in range(0, 3):
         tmp = inception_C(input=tmp, depth=i)
-
+    '''
     pool = fluid.layers.pool2d(
         pool_type='ave', input=tmp, pool_size=7, pool_stride=1)
     dropout = fluid.layers.dropout(input=pool, drop_prob=0.2)
-    out = fluid.layers.softmax(input=dropout)
+    fc = fluid.layers.fc(input=dropout, size=class_dim, act='softmax')
+    out = fluid.layers.softmax(input=fc)
     return out
 
 
@@ -122,15 +123,13 @@ def inception_A(input, depth):
     b2_conv1 = conv_bn_layer(
         name='inceptA{0}_branch2_conv1'.format(depth),
         input=b2_conv0,
-        num_channels=64,
         num_filters=96,
         filter_size=3,
         stride=1,
         padding=1)
     b3_conv0 = conv_bn_layer(
         name='inceptA{0}_branch3_conv0'.format(depth),
         input=input,
-        num_channels=384,
         num_filters=64,
         filter_size=1,
         stride=1,

fluid/image_classification/train.py

@@ -222,7 +222,7 @@ def train_parallel_exe(args,
                        use_nccl=True,
                        lr_strategy=None,
                        layers=50):
-    class_dim = 1000
+    class_dim = 101
     image_shape = [3, 224, 224]
 
     image = fluid.layers.data(name='image', shape=image_shape, dtype='float32')
@@ -284,8 +284,15 @@ def if_exist(var):
 
         fluid.io.load_vars(exe, pretrained_model, predicate=if_exist)
 
-    train_reader = paddle.batch(reader.train(), batch_size=batch_size)
-    test_reader = paddle.batch(reader.test(), batch_size=batch_size)
+    #train_reader = paddle.batch(reader.train(), batch_size=batch_size)
+    #test_reader = paddle.batch(reader.test(), batch_size=batch_size)
+    train_reader = paddle.batch(
+        paddle.reader.shuffle(
+            paddle.dataset.flowers.train(), buf_size=5120),
+        batch_size=args.batch_size)
+    test_reader = paddle.batch(
+        paddle.dataset.flowers.test(), batch_size=args.batch_size)
+
     feeder = fluid.DataFeeder(place=place, feed_list=[image, label])
 
     train_exe = fluid.ParallelExecutor(use_cuda=True, loss_name=avg_cost.name)