Merge pull request #1335 from kebinC/metric

support python3

fluid/metric_learning/losses/datareader.py

 import os
 import math
 import random
-import cPickle
 import functools
 import numpy as np
 import paddle
@@ -44,9 +43,9 @@ for i, item in enumerate(test_list):
         test_data[label] = []
     test_data[label].append(path)
 
-print "train_data size:", len(train_data)
-print "test_data size:", len(test_data)
-print "test_data image number:", len(test_image_list)
+print("train_data size:", len(train_data))
+print("test_data size:", len(test_data))
+print("test_data image number:", len(test_image_list))
 random.shuffle(test_image_list)
 
 
@@ -213,11 +212,11 @@ def eml_iterator(data,
                  color_jitter=False,
                  rotate=False):
     def reader():
-        labs = data.keys()
+        labs = list(data.keys())
         lab_num = len(labs)
-        ind = range(0, lab_num)
+        ind = list(range(0, lab_num))
         assert batch_size % samples_each_class == 0, "batch_size % samples_each_class != 0"
-        num_class = batch_size/samples_each_class
+        num_class = batch_size // samples_each_class
         for i in range(iter_size):
             random.shuffle(ind)
             for n in range(num_class):
@@ -244,9 +243,9 @@ def quadruplet_iterator(data,
                         color_jitter=False,
                         rotate=False):
     def reader():
-        labs = data.keys()
+        labs = list(data.keys())
         lab_num = len(labs)
-        ind = range(0, lab_num)
+        ind = list(range(0, lab_num))
         for i in range(iter_size):
             random.shuffle(ind)
             ind_sample = ind[:class_num]
@@ -254,7 +253,7 @@ def quadruplet_iterator(data,
             for ind_i in ind_sample:
                 lab = labs[ind_i]
                 data_list = data[lab]
-                data_ind = range(0, len(data_list))
+                data_ind = list(range(0, len(data_list)))
                 random.shuffle(data_ind)
                 anchor_ind = data_ind[:samples_each_class]
 
@@ -276,15 +275,15 @@ def triplet_iterator(data,
                      color_jitter=False,
                      rotate=False):
     def reader():
-        labs = data.keys()
+        labs = list(data.keys())
         lab_num = len(labs)
-        ind = range(0, lab_num)
+        ind = list(range(0, lab_num))
         for i in range(iter_size):
             random.shuffle(ind)
             ind_pos, ind_neg = ind[:2]
             lab_pos = labs[ind_pos]
             pos_data_list = data[lab_pos]
-            data_ind = range(0, len(pos_data_list))
+            data_ind = list(range(0, len(pos_data_list)))
             random.shuffle(data_ind)
             anchor_ind, pos_ind = data_ind[:2]
 
@@ -345,7 +344,7 @@ def quadruplet_train(class_num, samples_each_class):
             
 def triplet_train(batch_size):
     assert(batch_size % 3 == 0)
-    return triplet_iterator(train_data, 'train', batch_size, iter_size = batch_size/3 * 100, \
+    return triplet_iterator(train_data, 'train', batch_size, iter_size = batch_size//3 * 100, \
                            shuffle=True, color_jitter=False, rotate=False)
 
 def test():

fluid/metric_learning/losses/emlloss.py

-import datareader as reader
 import math
 import numpy as np
 import paddle.fluid as fluid
-from metrics import calculate_order_dist_matrix
-from metrics import get_gpu_num
+from . import datareader as reader
+from .metrics import calculate_order_dist_matrix
+from .metrics import get_gpu_num
 
 class emlloss():
     def __init__(self, train_batch_size = 40, samples_each_class=2):
@@ -11,9 +11,9 @@ class emlloss():
         self.samples_each_class = samples_each_class
         self.train_batch_size = train_batch_size
         assert(train_batch_size % num_gpus == 0)
-        self.cal_loss_batch_size = train_batch_size / num_gpus
+        self.cal_loss_batch_size = train_batch_size // num_gpus
         assert(self.cal_loss_batch_size % samples_each_class == 0)
-        class_num = train_batch_size / samples_each_class
+        class_num = train_batch_size // samples_each_class
         self.train_reader = reader.eml_train(train_batch_size, samples_each_class)
         self.test_reader = reader.test()
 

fluid/metric_learning/losses/metrics.py

@@ -20,12 +20,14 @@ def recall_topk(fea, lab, k = 1):
 
 import subprocess
 import os
+
 def get_gpu_num():
     visibledevice = os.getenv('CUDA_VISIBLE_DEVICES')
     if visibledevice:
         devicenum = len(visibledevice.split(','))
     else:
-        devicenum = subprocess.check_output(['nvidia-smi', '-L']).count('\n')
+        devicenum = subprocess.check_output(
+            [str.encode('nvidia-smi'), str.encode('-L')]).decode('utf-8').count('\n')
     return devicenum
 
 import paddle as paddle

fluid/metric_learning/losses/quadrupletloss.py

 import numpy as np
-import datareader as reader
 import paddle.fluid as fluid
-from metrics import calculate_order_dist_matrix
-from metrics import get_gpu_num
+from . import datareader as reader
+from .metrics import calculate_order_dist_matrix
+from .metrics import get_gpu_num
 
 class quadrupletloss():
     def __init__(self, 
@@ -14,9 +14,9 @@ class quadrupletloss():
         self.samples_each_class = samples_each_class
         self.train_batch_size = train_batch_size
         assert(train_batch_size % num_gpus == 0)
-        self.cal_loss_batch_size = train_batch_size / num_gpus
+        self.cal_loss_batch_size = train_batch_size // num_gpus
         assert(self.cal_loss_batch_size % samples_each_class == 0)
-        class_num = train_batch_size / samples_each_class
+        class_num = train_batch_size // samples_each_class
         self.train_reader = reader.quadruplet_train(class_num, samples_each_class)
         self.test_reader = reader.test()
 

fluid/metric_learning/losses/tripletloss.py

-import datareader as reader
+from . import datareader as reader
 import paddle.fluid as fluid
 
 class tripletloss():

fluid/metric_learning/models/resnet.py

@@ -75,7 +75,7 @@ class ResNet():
             num_filters=num_filters,
             filter_size=filter_size,
             stride=stride,
-            padding=(filter_size - 1) / 2,
+            padding=(filter_size - 1) // 2,
             groups=groups,
             act=None,
             bias_attr=False)

fluid/metric_learning/models/se_resnext.py

@@ -127,7 +127,7 @@ class SE_ResNeXt():
             num_filters=num_filters,
             filter_size=filter_size,
             stride=stride,
-            padding=(filter_size - 1) / 2,
+            padding=(filter_size - 1) // 2,
             groups=groups,
             act=None,
             bias_attr=False)

fluid/metric_learning/train.py

@@ -93,7 +93,7 @@ def train(args):
     elif loss_name == "emlloss":
         metricloss = emlloss(
                 train_batch_size = args.train_batch_size, 
-                samples_each_class=2
+                samples_each_class = args.samples_each_class
         )
         cost_metric = metricloss.loss(out[0])
         avg_cost_metric = fluid.layers.mean(x=cost_metric)

fluid/metric_learning/utility.py

@@ -17,6 +17,7 @@ from __future__ import absolute_import
 from __future__ import division
 from __future__ import print_function
 import distutils.util
+import six
 import numpy as np
 from paddle.fluid import core
 
@@ -37,7 +38,7 @@ def print_arguments(args):
     :type args: argparse.Namespace
     """
     print("-----------  Configuration Arguments -----------")
-    for arg, value in sorted(vars(args).iteritems()):
+    for arg, value in sorted(six.iteritems(vars(args))):
         print("%s: %s" % (arg, value))
     print("------------------------------------------------")