fix use of deprecated TF functions.

examples/DisturbLabel/mnist-disturb.py

@@ -49,8 +49,8 @@ class Model(mnist_example.Model):
 
         cost = tf.nn.sparse_softmax_cross_entropy_with_logits(logits, label)
         cost = tf.reduce_mean(cost, name='cross_entropy_loss')
-        wd_cost = tf.mul(1e-5, regularize_cost('fc.*/W', tf.nn.l2_loss),
-                         name='regularize_loss')
+        wd_cost = tf.multiply(1e-5, regularize_cost('fc.*/W', tf.nn.l2_loss),
+                              name='regularize_loss')
 
         self.cost = tf.add_n([wd_cost, cost], name='cost')
         add_moving_summary(cost, wd_cost, self.cost)

examples/DoReFa-Net/resnet-dorefa.py

@@ -107,7 +107,7 @@ class Model(ModelDesc):
                       .BatchNorm('lastbn')
                       .apply(nonlin)
                       .GlobalAvgPooling('gap')
-                      .tf.mul(49)  # this is due to a bug in our model design
+                      .tf.multiply(49)  # this is due to a bug in our model design
                       .FullyConnected('fct', 1000)())
         prob = tf.nn.softmax(logits, name='output')
         wrong = prediction_incorrect(logits, label, 1, name='wrong-top1')

examples/GAN/GAN.py

@@ -59,15 +59,15 @@ def build_GAN_losses(vecpos, vecneg):
     tf.summary.histogram('sigmoid-neg', sigmneg)
 
     d_loss_pos = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(
-        vecpos, tf.ones_like(vecpos)), name='d_CE_loss_pos')
+        logits=vecpos, labels=tf.ones_like(vecpos)), name='d_CE_loss_pos')
     d_loss_neg = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(
-        vecneg, tf.zeros_like(vecneg)), name='d_CE_loss_neg')
+        logits=vecneg, labels=tf.zeros_like(vecneg)), name='d_CE_loss_neg')
 
     d_pos_acc = tf.reduce_mean(tf.cast(sigmpos > 0.5, tf.float32), name='pos_acc')
     d_neg_acc = tf.reduce_mean(tf.cast(sigmneg < 0.5, tf.float32), name='neg_acc')
 
     g_loss = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(
-        vecneg, tf.ones_like(vecneg)), name='g_CE_loss')
+        logits=vecneg, labels=tf.ones_like(vecneg)), name='g_CE_loss')
     d_loss = tf.add(d_loss_pos, d_loss_neg, name='d_CE_loss')
     add_moving_summary(d_loss_pos, d_loss_neg,
                        g_loss, d_loss,

examples/GAN/InfoGAN-mnist.py

@@ -81,7 +81,7 @@ class Model(ModelDesc):
         log_qc = tf.reduce_sum(prior_prob * zc, 1, name='logQc')
         Elog_qc_given_x = tf.reduce_mean(log_qc_given_x, name='ElogQc_x')
         Hc = tf.reduce_mean(-log_qc, name='Hc')
-        MIloss = tf.mul(Hc + Elog_qc_given_x, -1.0, name='neg_MI')
+        MIloss = tf.multiply(Hc + Elog_qc_given_x, -1.0, name='neg_MI')
 
         self.g_loss, self.d_loss = build_GAN_losses(vecpos, vecneg)
         self.g_loss = tf.add(self.g_loss, MIloss, name='total_g_loss')

examples/HED/hed.py

@@ -86,7 +86,7 @@ class Model(ModelDesc):
         if get_current_tower_context().is_training:
             wd_w = tf.train.exponential_decay(2e-4, get_global_step_var(),
                                               80000, 0.7, True)
-            wd_cost = tf.mul(wd_w, regularize_cost('.*/W', tf.nn.l2_loss), name='wd_cost')
+            wd_cost = tf.multiply(wd_w, regularize_cost('.*/W', tf.nn.l2_loss), name='wd_cost')
             costs.append(wd_cost)
 
             add_param_summary(('.*/W', ['histogram']))   # monitor W

examples/Inception/inception-bn.py

@@ -113,7 +113,7 @@ class Model(ModelDesc):
         # weight decay on all W of fc layers
         wd_w = tf.train.exponential_decay(0.0002, get_global_step_var(),
                                           80000, 0.7, True)
-        wd_cost = tf.mul(wd_w, regularize_cost('.*/W', tf.nn.l2_loss), name='l2_regularize_loss')
+        wd_cost = tf.multiply(wd_w, regularize_cost('.*/W', tf.nn.l2_loss), name='l2_regularize_loss')
 
         add_param_summary(('.*/W', ['histogram']))   # monitor W
         self.cost = tf.add_n([cost, wd_cost], name='cost')

examples/Inception/inceptionv3.py

@@ -192,7 +192,7 @@ class Model(ModelDesc):
         # weight decay on all W of fc layers
         wd_w = tf.train.exponential_decay(0.00004, get_global_step_var(),
                                           80000, 0.7, True)
-        wd_cost = tf.mul(wd_w, regularize_cost('.*/W', tf.nn.l2_loss), name='l2_regularize_loss')
+        wd_cost = tf.multiply(wd_w, regularize_cost('.*/W', tf.nn.l2_loss), name='l2_regularize_loss')
 
         self.cost = tf.add_n([0.4 * loss1, loss2, wd_cost], name='cost')
         add_moving_summary(loss1, loss2, wd_cost, self.cost)

examples/OpenAIGym/train-atari.py

@@ -115,7 +115,7 @@ class Model(ModelDesc):
 
         log_pi_a_given_s = tf.reduce_sum(
             log_probs * tf.one_hot(action, NUM_ACTIONS), 1)
-        advantage = tf.sub(tf.stop_gradient(self.value), futurereward, name='advantage')
+        advantage = tf.subtract(tf.stop_gradient(self.value), futurereward, name='advantage')
         policy_loss = tf.reduce_sum(log_pi_a_given_s * advantage, name='policy_loss')
         xentropy_loss = tf.reduce_sum(
             self.logits * log_probs, name='xentropy_loss')

examples/ResNet/cifar10-resnet.py

@@ -100,7 +100,7 @@ class Model(ModelDesc):
         # weight decay on all W of fc layers
         wd_w = tf.train.exponential_decay(0.0002, get_global_step_var(),
                                           480000, 0.2, True)
-        wd_cost = tf.mul(wd_w, regularize_cost('.*/W', tf.nn.l2_loss), name='wd_cost')
+        wd_cost = tf.multiply(wd_w, regularize_cost('.*/W', tf.nn.l2_loss), name='wd_cost')
         add_moving_summary(cost, wd_cost)
 
         add_param_summary(('.*/W', ['histogram']))   # monitor W

examples/ResNet/imagenet-resnet.py

@@ -111,7 +111,7 @@ class Model(ModelDesc):
         wrong = prediction_incorrect(logits, label, 5, name='wrong-top5')
         add_moving_summary(tf.reduce_mean(wrong, name='train-error-top5'))
 
-        wd_cost = tf.mul(1e-4, regularize_cost('.*/W', tf.nn.l2_loss), name='l2_regularize_loss')
+        wd_cost = regularize_cost('.*/W', l2_regularizer(1e-4), name='l2_regularize_loss')
         add_moving_summary(loss, wd_cost)
         self.cost = tf.add_n([loss, wd_cost], name='cost')
 

examples/SpatialTransformer/mnist-addition.py

@@ -81,8 +81,8 @@ class Model(ModelDesc):
         wrong = symbolic_functions.prediction_incorrect(logits, label)
         summary.add_moving_summary(tf.reduce_mean(wrong, name='train_error'))
 
-        wd_cost = tf.mul(1e-5, regularize_cost('fc.*/W', tf.nn.l2_loss),
-                         name='regularize_loss')
+        wd_cost = tf.multiply(1e-5, regularize_cost('fc.*/W', tf.nn.l2_loss),
+                              name='regularize_loss')
         summary.add_moving_summary(cost, wd_cost)
         self.cost = tf.add_n([wd_cost, cost], name='cost')
 

examples/cifar-convnet.py

@@ -66,9 +66,7 @@ class Model(ModelDesc):
         add_moving_summary(tf.reduce_mean(wrong, name='train_error'))
 
         # weight decay on all W of fc layers
-        wd_cost = tf.mul(0.0004,
-                         regularize_cost('fc.*/W', tf.nn.l2_loss),
-                         name='regularize_loss')
+        wd_cost = regularize_cost('fc.*/W', l2_regularizer(4e-4), name='regularize_loss')
         add_moving_summary(cost, wd_cost)
 
         add_param_summary(('.*/W', ['histogram']))   # monitor W

examples/mnist-convnet.py

@@ -96,9 +96,9 @@ class Model(ModelDesc):
         if not USE_SLIM:
             # Use a regex to find parameters to apply weight decay.
             # Here we apply a weight decay on all W (weight matrix) of all fc layers
-            wd_cost = tf.mul(1e-5,
-                             regularize_cost('fc.*/W', tf.nn.l2_loss),
-                             name='regularize_loss')
+            wd_cost = tf.multiply(1e-5,
+                                  regularize_cost('fc.*/W', tf.nn.l2_loss),
+                                  name='regularize_loss')
             self.cost = tf.add_n([wd_cost, cost], name='total_cost')
             summary.add_moving_summary(cost, wd_cost, self.cost)
         else:

tensorpack/dataflow/dataset/__init__.py

@@ -18,7 +18,11 @@ def global_import(name):
         __all__.append(k)
 
 
+_CURR_DIR = os.path.dirname(__file__)
 for _, module_name, _ in walk_packages(
-        [os.path.dirname(__file__)]):
+        [_CURR_DIR]):
+    srcpath = os.path.join(_CURR_DIR, module_name + '.py')
+    if not os.path.isfile(srcpath):
+        continue
     if not module_name.startswith('_'):
         global_import(module_name)

tensorpack/models/nonlin.py

@@ -50,7 +50,7 @@ def PReLU(x, init=0.001, name='output'):
     init = tf.constant_initializer(init)
     alpha = tf.get_variable('alpha', [], initializer=init)
     x = ((1 + alpha) * x + (1 - alpha) * tf.abs(x))
-    return tf.mul(x, 0.5, name=name)
+    return tf.multiply(x, 0.5, name=name)
 
 
 @layer_register(use_scope=False, log_shape=False)

tensorpack/tfutils/symbolic_functions.py

@@ -59,7 +59,7 @@ def class_balanced_cross_entropy(pred, label, name='cross_entropy_loss'):
     eps = 1e-12
     loss_pos = -beta * tf.reduce_mean(y * tf.log(z + eps))
     loss_neg = (1. - beta) * tf.reduce_mean((1. - y) * tf.log(1. - z + eps))
-    cost = tf.sub(loss_pos, loss_neg, name=name)
+    cost = tf.subtract(loss_pos, loss_neg, name=name)
     return cost