changed Trainer() signature from Trainer(model, loss, metric, ...) to…

… Trainer(model, (loss, metric), ...) to prep for Layers lib where (loss, metric) is the criterion function; fixed an existing bug in a doc example
prime9 · Feb 9, 2017 · 99c25c0 · 99c25c0
1 parent 81df1d8
commit 99c25c0
Show file tree

Hide file tree

Showing 44 changed files with 65 additions and 64 deletions.
diff --git a/Examples/Image/Classification/AlexNet/Python/AlexNet_ImageNet_Distributed.py b/Examples/Image/Classification/AlexNet/Python/AlexNet_ImageNet_Distributed.py
@@ -157,7 +157,7 @@ def create_trainer(network, epoch_size, num_quantization_bits):
         distributed_after=0)
 
     # Create trainer
-    return cntk.Trainer(network['output'], network['ce'], network['pe'], parameter_learner)
+    return cntk.Trainer(network['output'], (network['ce'], network['pe']), parameter_learner)
 
 # Train and test
 def train_and_test(network, trainer, train_source, test_source, progress_printer, minibatch_size, epoch_size, restore):

diff --git a/Examples/Image/Classification/ConvNet/Python/ConvNetLRN_CIFAR10_DataAug.py b/Examples/Image/Classification/ConvNet/Python/ConvNetLRN_CIFAR10_DataAug.py
@@ -108,7 +108,7 @@ def convnetlrn_cifar10_dataaug(reader_train, reader_test, epoch_size=50000, max_
     learner = cntk.learner.momentum_sgd(z.parameters, lr_schedule, mm_schedule,
                                         unit_gain = True,
                                         l2_regularization_weight = l2_reg_weight)
-    trainer =  cntk.Trainer(z, ce, pe, learner)
+    trainer =  cntk.Trainer(z, (ce, pe), learner)
 
     # define mapping from reader streams to network inputs
     input_map = {

diff --git a/Examples/Image/Classification/ConvNet/Python/ConvNet_CIFAR10.py b/Examples/Image/Classification/ConvNet/Python/ConvNet_CIFAR10.py
@@ -75,7 +75,7 @@ def convnet_cifar10(debug_output=False):
     # Instantiate the trainer object to drive the model training
     learner = cntk.learner.momentum_sgd(z.parameters, lr_schedule, mm_schedule,
                                         l2_regularization_weight = l2_reg_weight)
-    trainer = cntk.Trainer(z, ce, pe, learner)
+    trainer = cntk.Trainer(z, (ce, pe), learner)
 
     # define mapping from reader streams to network inputs
     input_map = {

diff --git a/Examples/Image/Classification/ConvNet/Python/ConvNet_CIFAR10_DataAug.py b/Examples/Image/Classification/ConvNet/Python/ConvNet_CIFAR10_DataAug.py
@@ -86,7 +86,7 @@ def convnet_cifar10_dataaug(reader_train, reader_test, epoch_size = 50000, max_e
     # trainer object
     learner = cntk.learner.momentum_sgd(z.parameters, lr_schedule, mm_schedule,
                                         l2_regularization_weight = l2_reg_weight)
-    trainer =  cntk.Trainer(z, ce, pe, learner)
+    trainer =  cntk.Trainer(z, (ce, pe), learner)
 
     # define mapping from reader streams to network inputs
     input_map = {

diff --git a/Examples/Image/Classification/ConvNet/Python/ConvNet_CIFAR10_DataAug_Distributed.py b/Examples/Image/Classification/ConvNet/Python/ConvNet_CIFAR10_DataAug_Distributed.py
@@ -111,7 +111,7 @@ def create_trainer(network, epoch_size, num_quantization_bits, block_size, warm_
         parameter_learner = cntk.distributed.data_parallel_distributed_learner(local_learner, num_quantization_bits=num_quantization_bits, distributed_after=warm_up)
 
     # Create trainer
-    return cntk.Trainer(network['output'], network['ce'], network['pe'], parameter_learner)
+    return cntk.Trainer(network['output'], (network['ce'], network['pe']), parameter_learner)
 
 # Train and test
 def train_and_test(network, trainer, train_source, test_source, progress_printer, minibatch_size, epoch_size, restore):

diff --git a/Examples/Image/Classification/ConvNet/Python/ConvNet_MNIST.py b/Examples/Image/Classification/ConvNet/Python/ConvNet_MNIST.py
@@ -65,7 +65,7 @@ def convnet_mnist(debug_output=False):
 
     # Instantiate the trainer object to drive the model training
     learner = cntk.learner.momentum_sgd(z.parameters, lr_schedule, mm_schedule)
-    trainer = cntk.Trainer(z, ce, pe, learner)
+    trainer = cntk.Trainer(z, (ce, pe), learner)
 
     # define mapping from reader streams to network inputs
     input_map = {

diff --git a/Examples/Image/Classification/MLP/Python/SimpleMNIST.py b/Examples/Image/Classification/MLP/Python/SimpleMNIST.py
@@ -65,7 +65,7 @@ def simple_mnist(debug_output=False):
 
     lr_per_minibatch=learning_rate_schedule(0.2, UnitType.minibatch)
     # Instantiate the trainer object to drive the model training
-    trainer = Trainer(z, ce, pe, sgd(z.parameters, lr=lr_per_minibatch))
+    trainer = Trainer(z, (ce, pe), sgd(z.parameters, lr=lr_per_minibatch))
 
     # Get minibatches of images to train with and perform model training
     minibatch_size = 64

diff --git a/Examples/Image/Classification/ResNet/Python/TrainResNet_CIFAR10.py b/Examples/Image/Classification/ResNet/Python/TrainResNet_CIFAR10.py
@@ -88,7 +88,7 @@ def train_and_evaluate(reader_train, reader_test, network_name, epoch_size, max_
     # trainer object
     learner     = momentum_sgd(z.parameters, lr_schedule, mm_schedule,
                                l2_regularization_weight = l2_reg_weight)
-    trainer     = Trainer(z, ce, pe, learner)
+    trainer     = Trainer(z, (ce, pe), learner)
 
     # define mapping from reader streams to network inputs
     input_map = {

diff --git a/Examples/Image/Classification/ResNet/Python/TrainResNet_CIFAR10_Distributed.py b/Examples/Image/Classification/ResNet/Python/TrainResNet_CIFAR10_Distributed.py
@@ -94,7 +94,7 @@ def create_trainer(network, minibatch_size, epoch_size, num_quantization_bits, b
     else:
         learner = data_parallel_distributed_learner(local_learner, num_quantization_bits=num_quantization_bits, distributed_after=warm_up)
 
-    return Trainer(network['output'], network['ce'], network['pe'], learner)
+    return Trainer(network['output'], (network['ce'], network['pe']), learner)
 
 # Train and test
 def train_and_test(network, trainer, train_source, test_source, progress_printer, minibatch_size, epoch_size, profiler_dir=None):

diff --git a/Examples/Image/Classification/VGG/Python/VGG16_ImageNet_Distributed.py b/Examples/Image/Classification/VGG/Python/VGG16_ImageNet_Distributed.py
@@ -151,7 +151,7 @@ def create_trainer(network, epoch_size, num_quantization_bits):
         distributed_after=0)
 
     # Create trainer
-    return cntk.Trainer(network['output'], network['ce'], network['pe'], parameter_learner)
+    return cntk.Trainer(network['output'], (network['ce'], network['pe']), parameter_learner)
 
 # Train and test
 def train_and_test(network, trainer, train_source, test_source, progress_printer, minibatch_size, epoch_size, restore):

diff --git a/Examples/Image/Classification/VGG/Python/VGG19_ImageNet_Distributed.py b/Examples/Image/Classification/VGG/Python/VGG19_ImageNet_Distributed.py
@@ -151,7 +151,7 @@ def create_trainer(network, epoch_size, num_quantization_bits):
         distributed_after=0)
 
     # Create trainer
-    return cntk.Trainer(network['output'], network['ce'], network['pe'], parameter_learner)
+    return cntk.Trainer(network['output'], (network['ce'], network['pe']), parameter_learner)
 
 # Train and test
 def train_and_test(network, trainer, train_source, test_source, progress_printer, minibatch_size, epoch_size, restore):

diff --git a/Examples/Image/Detection/FastRCNN/A2_RunCntk_py3.py b/Examples/Image/Detection/FastRCNN/A2_RunCntk_py3.py
@@ -160,7 +160,7 @@ def train_fast_rcnn(debug_output=False):
 
     # Instantiate the trainer object
     learner = momentum_sgd(frcn_output.parameters, lr_schedule, mm_schedule, l2_regularization_weight=l2_reg_weight)
-    trainer = Trainer(frcn_output, ce, pe, learner)
+    trainer = Trainer(frcn_output, (ce, pe), learner)
 
     # Get minibatches of images and perform model training
     print("Training Fast R-CNN model for %s epochs." % max_epochs)

diff --git a/Examples/Image/GettingStarted/07_Deconvolution_PY.py b/Examples/Image/GettingStarted/07_Deconvolution_PY.py
@@ -62,7 +62,7 @@ def deconv_mnist(max_epochs=3):
 
     # Instantiate the trainer object to drive the model training
     learner = cntk.learner.momentum_sgd(z.parameters, lr_schedule, mm_schedule, unit_gain=True)
-    trainer = cntk.Trainer(z, rmse_loss, rmse_eval, learner)
+    trainer = cntk.Trainer(z, (rmse_loss, rmse_eval), learner)
 
     # define mapping from reader streams to network inputs
     input_map = {

diff --git a/Examples/Image/TransferLearning/TransferLearning.py b/Examples/Image/TransferLearning/TransferLearning.py
@@ -112,7 +112,7 @@ def train_model(base_model_file, feature_node_name, last_hidden_node_name,
     lr_schedule = learning_rate_schedule(lr_per_mb, unit=UnitType.minibatch)
     mm_schedule = momentum_schedule(momentum_per_mb)
     learner = momentum_sgd(tl_model.parameters, lr_schedule, mm_schedule, l2_regularization_weight=l2_reg_weight)
-    trainer = Trainer(tl_model, ce, pe, learner)
+    trainer = Trainer(tl_model, (ce, pe), learner)
 
     # Get minibatches of images and perform model training
     print("Training transfer learning model for {0} epochs (epoch_size = {1}).".format(num_epochs, epoch_size))

diff --git a/Examples/LanguageUnderstanding/ATIS/Python/LanguageUnderstanding.py b/Examples/LanguageUnderstanding/ATIS/Python/LanguageUnderstanding.py
@@ -82,7 +82,7 @@ def train(reader, model, max_epochs):
                                     low_memory=True,
                                     gradient_clipping_threshold_per_sample=15, gradient_clipping_with_truncation=True)
 
-    trainer = cntk.Trainer(z, ce, pe, [learner])
+    trainer = cntk.Trainer(z, (ce, pe), [learner])
 
     # define mapping from reader streams to network inputs
     input_map = {

diff --git a/Examples/SequenceClassification/SimpleExample/Python/SequenceClassification.py b/Examples/SequenceClassification/SimpleExample/Python/SequenceClassification.py
@@ -64,7 +64,7 @@ def train_sequence_classifier(debug_output=False):
 
     lr_per_sample = learning_rate_schedule(0.0005, UnitType.sample)
     # Instantiate the trainer object to drive the model training
-    trainer = Trainer(classifier_output, ce, pe,
+    trainer = Trainer(classifier_output, (ce, pe),
                       sgd(classifier_output.parameters, lr=lr_per_sample))
 
     # Get minibatches of sequences to train with and perform model training

diff --git a/Examples/SequenceToSequence/CMUDict/Python/Sequence2Sequence.py b/Examples/SequenceToSequence/CMUDict/Python/Sequence2Sequence.py
@@ -160,7 +160,7 @@ def sequence_to_sequence_translator(debug_output=False, run_test=False):
                            lr_per_minibatch, momentum_time_constant,
                            gradient_clipping_threshold_per_sample=clipping_threshold_per_sample, 
                            gradient_clipping_with_truncation=gradient_clipping_with_truncation)
-    trainer = Trainer(z, ce, errs, learner)
+    trainer = Trainer(z, (ce, errs), learner)
 
     # setup data
     train_path = os.path.join(os.path.dirname(os.path.abspath(__file__)), "..", "Data", "cmudict-0.7b.train-dev-20-21.ctf")
@@ -240,7 +240,7 @@ def sequence_to_sequence_translator(debug_output=False, run_test=False):
     label_sequence = sequence.slice(find_arg_by_name('raw_labels',z), 1, 0)
     ce = cross_entropy_with_softmax(z, label_sequence)
     errs = classification_error(z, label_sequence)
-    trainer = Trainer(z, ce, errs, [momentum_sgd(
+    trainer = Trainer(z, (ce, errs), [momentum_sgd(
                     z.parameters, lr_per_minibatch, momentum_time_constant, True,
                     clipping_threshold_per_sample, gradient_clipping_with_truncation)])
 

diff --git a/Examples/Tensorboard/LanguageUnderstanding.py b/Examples/Tensorboard/LanguageUnderstanding.py
@@ -86,7 +86,7 @@ def train(reader, model, max_epochs):
                        low_memory=True,
                        gradient_clipping_threshold_per_sample=15, gradient_clipping_with_truncation=True)
 
-    trainer = Trainer(z, ce, pe, [learner])
+    trainer = Trainer(z, (ce, pe), [learner])
 
     # define mapping from reader streams to network inputs
     input_map = {

diff --git a/Examples/Tensorboard/SimpleMNIST.py b/Examples/Tensorboard/SimpleMNIST.py
@@ -70,7 +70,7 @@ def simple_mnist():
 
     lr_per_minibatch = learning_rate_schedule(0.2, UnitType.minibatch)
     # Instantiate the trainer object to drive the model training
-    trainer = Trainer(netout, ce, pe, sgd(netout.parameters, lr=lr_per_minibatch))
+    trainer = Trainer(netout, (ce, pe), sgd(netout.parameters, lr=lr_per_minibatch))
 
     # Instantiate a ProgressPrinter.
     logdir = os.path.join(os.path.dirname(os.path.abspath(__file__)), "mnist_log") 

diff --git a/Examples/Text/CharacterLM/char_rnn.py b/Examples/Text/CharacterLM/char_rnn.py
@@ -167,7 +167,7 @@ def train_lm(training_file, max_num_minibatches):
     learner = momentum_sgd(z.parameters, lr_per_sample, momentum_time_constant,
                            gradient_clipping_threshold_per_sample=clipping_threshold_per_sample,
                            gradient_clipping_with_truncation=gradient_clipping_with_truncation)
-    trainer = Trainer(z, ce, errs, learner)
+    trainer = Trainer(z, (ce, errs), learner)
 
     sample_freq = 1000
     epochs = 50

diff --git a/Examples/Video/GettingStarted/Python/Conv3D_UCF11.py b/Examples/Video/GettingStarted/Python/Conv3D_UCF11.py
@@ -199,7 +199,7 @@ def conv3d_ucf11(train_reader, test_reader, max_epochs=30):
 
     # Instantiate the trainer object to drive the model training
     learner     = momentum_sgd(z.parameters, lr_schedule, mm_schedule, True)
-    trainer     = Trainer(z, ce, pe, learner)
+    trainer     = Trainer(z, (ce, pe), learner)
 
     log_number_of_parameters(z) ; print()
     progress_printer = ProgressPrinter(tag='Training')

diff --git a/Tutorials/CNTK_101_LogisticRegression.ipynb b/Tutorials/CNTK_101_LogisticRegression.ipynb
@@ -514,7 +514,7 @@
     "learning_rate = 0.5\n",
     "lr_schedule = learning_rate_schedule(learning_rate, UnitType.minibatch) \n",
     "learner = sgd(z.parameters, lr_schedule)\n",
-    "trainer = Trainer(z, loss, eval_error, [learner])"
+    "trainer = Trainer(z, (loss, eval_error), [learner])"
    ]
   },
   {

diff --git a/Tutorials/CNTK_102_FeedForward.ipynb b/Tutorials/CNTK_102_FeedForward.ipynb
@@ -597,7 +597,7 @@
     "learning_rate = 0.5\n",
     "lr_schedule = learning_rate_schedule(learning_rate, UnitType.minibatch) \n",
     "learner = C.learner.sgd(z.parameters, lr_schedule)\n",
-    "trainer = Trainer(z, loss, eval_error, [learner])"
+    "trainer = Trainer(z, (loss, eval_error), [learner])"
    ]
   },
   {

diff --git a/Tutorials/CNTK_103B_MNIST_FeedForwardNetwork.ipynb b/Tutorials/CNTK_103B_MNIST_FeedForwardNetwork.ipynb
@@ -428,7 +428,7 @@
     "learning_rate = 0.2\n",
     "lr_schedule = learning_rate_schedule(learning_rate, UnitType.minibatch)\n",
     "learner = sgd(z.parameters, lr_schedule)\n",
-    "trainer = Trainer(z, loss, label_error, [learner])"
+    "trainer = Trainer(z, (loss, label_error), [learner])"
    ]
   },
   {

diff --git a/Tutorials/CNTK_104_Finance_Timeseries_Basic_with_Pandas_Numpy.ipynb b/Tutorials/CNTK_104_Finance_Timeseries_Basic_with_Pandas_Numpy.ipynb
@@ -607,7 +607,7 @@
     "loss = C.cross_entropy_with_softmax(z, label)\n",
     "label_error = C.classification_error(z, label)\n",
     "lr_per_minibatch = learning_rate_schedule(0.125,UnitType.minibatch)\n",
-    "trainer = cntk.Trainer(z, loss, label_error, [sgd(z.parameters, lr=lr_per_minibatch)])"
+    "trainer = cntk.Trainer(z, (loss, label_error), [sgd(z.parameters, lr=lr_per_minibatch)])"
    ]
   },
   {

diff --git a/Tutorials/CNTK_105_Basic_Autoencoder_for_Dimensionality_Reduction.ipynb b/Tutorials/CNTK_105_Basic_Autoencoder_for_Dimensionality_Reduction.ipynb
@@ -351,7 +351,7 @@
     "                       lr=lr_schedule, momentum=momentum_as_time_constant) \n",
     "    \n",
     "    # Instantiate the trainer\n",
-    "    trainer = Trainer(model, loss, label_error, learner)\n",
+    "    trainer = Trainer(model, (loss, label_error), learner)\n",
     "    \n",
     "    # Map the data streams to the input and labels.\n",
     "    # Note: for autoencoders input == label\n",

diff --git a/Tutorials/CNTK_106A_LSTM_Timeseries_with_Simulated_Data.ipynb b/Tutorials/CNTK_106A_LSTM_Timeseries_with_Simulated_Data.ipynb
@@ -409,7 +409,7 @@
     "                             lr = lr_schedule, \n",
     "                             momentum = momentum_time_constant, \n",
     "                             unit_gain = True)\n",
-    "trainer = C.Trainer(z, loss, error, [learner])"
+    "trainer = C.Trainer(z, (loss, error), [learner])"
    ]
   },
   {

diff --git a/Tutorials/CNTK_201B_CIFAR-10_ImageHandsOn.ipynb b/Tutorials/CNTK_201B_CIFAR-10_ImageHandsOn.ipynb
@@ -476,7 +476,7 @@
     "    learner     = momentum_sgd(z.parameters, \n",
     "                               lr = lr_per_minibatch, momentum = momentum_time_constant, \n",
     "                               l2_regularization_weight=l2_reg_weight)\n",
-    "    trainer     = Trainer(z, ce, pe, [learner])\n",
+    "    trainer     = Trainer(z, (ce, pe), [learner])\n",
     "\n",
     "    # define mapping from reader streams to network inputs\n",
     "    input_map = {\n",

diff --git a/Tutorials/CNTK_202_Language_Understanding.ipynb b/Tutorials/CNTK_202_Language_Understanding.ipynb
@@ -435,7 +435,7 @@
     "                       gradient_clipping_threshold_per_sample=15, gradient_clipping_with_truncation=True)\n",
     "\n",
     "    # trainer\n",
-    "    trainer = Trainer(model, criterion.outputs[0], criterion.outputs[1], learner)\n",
+    "    trainer = Trainer(model, criterion, learner)\n",
     "\n",
     "    # process minibatches and perform model training\n",
     "    log_number_of_parameters(model)\n",
@@ -561,7 +561,7 @@
     "    momentum_as_time_constant = momentum_as_time_constant_schedule(0)\n",
     "    dummy_learner = adam_sgd(criterion.parameters, \n",
     "                             lr=lr_schedule, momentum=momentum_as_time_constant, low_memory=True)\n",
-    "    evaluator = Trainer(model, criterion.outputs[0], criterion.outputs[1], dummy_learner)\n",
+    "    evaluator = Trainer(model, criterion, dummy_learner)\n",
     "    progress_printer = ProgressPrinter(tag='Evaluation')\n",
     "\n",
     "    while True:\n",

diff --git a/Tutorials/CNTK_203_Reinforcement_Learning_Basics.ipynb b/Tutorials/CNTK_203_Reinforcement_Learning_Basics.ipynb
@@ -349,7 +349,7 @@
     "        # optimizer=opt\n",
     "        lr_schedule = learning_rate_schedule(lr, UnitType.minibatch)\n",
     "        learner = sgd(model.parameters, lr_schedule, gradient_clipping_threshold_per_sample=10)\n",
-    "        trainer = Trainer(model, loss, meas, learner)\n",
+    "        trainer = Trainer(model, (loss, meas), learner)\n",
     "\n",
     "        # CNTK: return trainer and loss as well\n",
     "        return model, trainer, loss\n",

diff --git a/Tutorials/CNTK_204_Sequence_To_Sequence.ipynb b/Tutorials/CNTK_204_Sequence_To_Sequence.ipynb
@@ -855,7 +855,7 @@
     "                        lr_per_sample, momentum_time_constant,\n",
     "                        gradient_clipping_threshold_per_sample=clipping_threshold_per_sample,\n",
     "                        gradient_clipping_with_truncation=gradient_clipping_with_truncation)\n",
-    "trainer = Trainer(model, ce, errs, learner)"
+    "trainer = Trainer(model, (ce, errs), learner)"
    ]
   },
   {
@@ -1029,7 +1029,7 @@
     "                           lr_per_sample, momentum_time_constant,\n",
     "                           gradient_clipping_threshold_per_sample=clipping_threshold_per_sample,\n",
     "                           gradient_clipping_with_truncation=gradient_clipping_with_truncation)\n",
-    "    trainer = Trainer(model, ce, errs, learner)\n",
+    "    trainer = Trainer(model, (ce, errs), learner)\n",
     "\n",
     "    # Get minibatches of sequences to train with and perform model training\n",
     "    i = 0\n",

diff --git a/Tutorials/CNTK_206_Basic_GAN.ipynb b/Tutorials/CNTK_206_Basic_GAN.ipynb
@@ -417,16 +417,14 @@
     "\n",
     "    # Instantiate the trainers\n",
     "    G_trainer = Trainer(\n",
-    "        model = X_fake,\n",
-    "        loss_function = G_loss,\n",
-    "        eval_function = None,\n",
-    "        parameter_learners = G_learner\n",
+    "        X_fake,\n",
+    "        (G_loss, None),\n",
+    "        G_learner\n",
     "    )\n",
     "    D_trainer = Trainer(\n",
-    "        model = D_real,\n",
-    "        loss_function = D_loss,\n",
-    "        eval_function = None,\n",
-    "        parameter_learners = D_learner\n",
+    "        D_real,\n",
+    "        (D_loss, None,\)n",
+    "        D_learner\n",
     "    )\n",
     "\n",
     "    return X_real, X_fake, Z, G_trainer, D_trainer"

diff --git a/Tutorials/CNTK_207_Training_with_Sampled_Softmax.ipynb b/Tutorials/CNTK_207_Training_with_Sampled_Softmax.ipynb
@@ -325,7 +325,7 @@
     "    learning_rate_schedule = C.learning_rate_schedule(Param.learning_rate, C.UnitType.sample)\n",
     "    momentum_schedule = C.momentum_as_time_constant_schedule(Param.momentum_time_constant)\n",
     "    learner = C.momentum_sgd(z.parameters, learning_rate_schedule, momentum_schedule, True)\n",
-    "    trainer = C.Trainer(z, cross_entropy, errs, learner)\n",
+    "    trainer = C.Trainer(z, (cross_entropy, errs), learner)\n",
     "\n",
     "    minbatch = 0\n",
     "    average_cross_entropy = compute_average_cross_entropy(z)\n",

diff --git a/Tutorials/NumpyInterop/FeedForwardNet.py b/Tutorials/NumpyInterop/FeedForwardNet.py
@@ -54,7 +54,7 @@ def ffnet():
 
     lr_per_minibatch=learning_rate_schedule(0.5, UnitType.minibatch)
     # Instantiate the trainer object to drive the model training
-    trainer = Trainer(netout, ce, pe, sgd(netout.parameters, lr=lr_per_minibatch))
+    trainer = Trainer(netout, (ce, pe), sgd(netout.parameters, lr=lr_per_minibatch))
 
     # Get minibatches of training data and perform model training
     minibatch_size = 25

diff --git a/bindings/python/cntk/ops/__init__.py b/bindings/python/cntk/ops/__init__.py
@@ -996,7 +996,7 @@ def times(left, right, output_rank=1, infer_input_rank_to_map=-1, name=''):
         array([[ 17.],
                [ 39.]], dtype=float32)
 
-        >>> C.times(1.*np.reshape(np.arange(8), (2.,2.,2.)),1.*np.reshape(np.arange(8), (2.,2.,2.)), output_rank=1).eval()
+        >>> C.times(1.*np.reshape(np.arange(8), (2,2,2)),1.*np.reshape(np.arange(8), (2,2,2)), output_rank=1).eval()
         array([[ 28.,  34.],
                [ 76.,  98.]])
 

diff --git a/bindings/python/cntk/ops/tests/userfunction_complex_test.py b/bindings/python/cntk/ops/tests/userfunction_complex_test.py
@@ -92,7 +92,7 @@ def train(nonlinearity, num_hidden_layers, device_id):
     eval_error = classification_error(z, label)
 
     learner = sgd(z.parameters, lr_schedule)
-    trainer = Trainer(z, loss, eval_error, [learner])
+    trainer = Trainer(z, (loss, eval_error), [learner])
 
 
     minibatch_size = 25

diff --git a/bindings/python/cntk/ops/tests/userfunction_test.py b/bindings/python/cntk/ops/tests/userfunction_test.py
@@ -147,7 +147,7 @@ def test_ext_train():
 
     momentum_time_constant = momentum_as_time_constant_schedule(1100)
     lr_per_sample = learning_rate_schedule(0.007, UnitType.sample)
-    trainer = Trainer(z, z+0, z+0, \
+    trainer = Trainer(z, (z+0, z+0), \
             [momentum_sgd(z.parameters, lr_per_sample, momentum_time_constant,
                 True)])
 
@@ -191,7 +191,7 @@ def backward(self, state, root_gradients):
     z = m+p
 
     lr_per_sample = learning_rate_schedule(0.007, UnitType.sample)
-    trainer = Trainer(z, z+0, z+0, \
+    trainer = Trainer(z, (z+0, z+0), \
             [sgd(z.parameters, lr_per_sample)])
 
     for i in range(100):
@@ -243,7 +243,7 @@ def inc(self, arg):
 
     momentum_time_constant = momentum_as_time_constant_schedule(1100)
     lr_per_sample = learning_rate_schedule(0.007, UnitType.sample)
-    trainer = Trainer(z, z+0, z+0, \
+    trainer = Trainer(z, (z+0, z+0), \
             [momentum_sgd(z.parameters, lr_per_sample, momentum_time_constant,
                 True)])