Fix typos

Examples/Image/Detection/FastRCNN/LabelMeConverter.py

@@ -33,7 +33,7 @@ for filename in os.listdir(filepath):
                 xlist = []
                 ylist = []
 
-                # loop through to get name and and BBox values from object
+                # loop through to get name and BBox values from object
                 for child in child:
                     if str(child.tag) == 'name':
                         label = child.text

Source/ComputationNetworkLib/ComputationNetwork.cpp

@@ -84,7 +84,7 @@ void ComputationNetwork::ClearNetwork()
 // serialization
 // -----------------------------------------------------------------------
 
-// after after editing--network is possibly not validated/compiled
+// after editing--network is possibly not validated/compiled
 void ComputationNetwork::SaveEdited(const wstring& fileName, const FileOptions fileFormat)
 {
     if (!IsCompiled())

Source/ComputationNetworkLib/ComputationNetworkAnalysis.cpp

@@ -318,7 +318,7 @@ void ComputationNetwork::DetermineSCCsR(ComputationNodeBasePtr cur,
             }
             }
             if (bFound)
-                fprintf(stderr, "\nDetermineSCCsR: %ls %ls operation was discovered multiple times as as loop participant", cur->NodeName().c_str(), cur->OperationName().c_str());
+                fprintf(stderr, "\nDetermineSCCsR: %ls %ls operation was discovered multiple times as loop participant", cur->NodeName().c_str(), cur->OperationName().c_str());
             // TODO: Once we forbid FormRecurrentLoops() from non-NULL, can we ever re-hit a loop here? If not, then turn bFound into a LogicError().
             if (!bFound)
             {

Source/ComputationNetworkLib/TrainingNodes.h

@@ -632,7 +632,7 @@ public:
         }
 
         // Update K (number of url pairs that have smaller or equal gain), rk (rank of 
-        // score in descending order) and and m_pairwiseDifferences (save for gradient 
+        // score in descending order) and m_pairwiseDifferences (save for gradient 
         // computation).
         size_t pairCount = 0;
         for (auto &qu : m_queryUrls)

Source/Readers/BinaryReader/BinaryFile.cpp

@@ -1246,7 +1246,7 @@ void SectionStats::InitCompute(const ConfigArray& compute)
     m_rms = 0.0;  // root mean square
 
     // second pass measures
-    m_varSum = 0.0; // accumulated sum of difference between the mean and and the value squared
+    m_varSum = 0.0; // accumulated sum of difference between the mean and the value squared
 
     // compute after second pass
     m_variance = 0.0;

Source/Readers/BinaryReader/BinaryReader.h

@@ -446,7 +446,7 @@ private:
     double m_pvariance;
 
     // second pass measures
-    double m_varSum; // accumulated sum of difference between the mean and and the value squared
+    double m_varSum; // accumulated sum of difference between the mean and the value squared
 
     // compute after second pass
     double m_variance;

Tutorials/CNTK_104_Finance_Timeseries_Basic_with_Pandas_Numpy.ipynb

@@ -514,7 +514,7 @@
     "training_data = data[\"2001-02-05\":\"2009-01-20\"] \n",
     "\n",
     "# We define our test data as: data[\"2008-01-02\":]\n",
-    "# This example allows to to include data up to current date\n",
+    "# This example allows to include data up to current date\n",
     "\n",
     "test_data= data[\"2009-01-20\":\"2016-12-29\"] \n",
     "training_features = np.asarray(training_data[predictor_names], dtype = \"float32\")\n",

Tutorials/CNTK_599A_Sequence_To_Sequence.ipynb

@@ -593,7 +593,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "Above, we use the function `element_select` which will return one of two options given the condition `is_first_label`. Remember that we're working with sequences so when the decoder LSTM is run its input will be unrolled along with the network. The above allows us to to have a dynamic input that will return a specific element given what time step we're currently processing.\n",
+    "Above, we use the function `element_select` which will return one of two options given the condition `is_first_label`. Remember that we're working with sequences so when the decoder LSTM is run its input will be unrolled along with the network. The above allows us to have a dynamic input that will return a specific element given what time step we're currently processing.\n",
     "\n",
     "Therefore, the `decoder_input` will be `label_sentence_start_scattered` (which is simply `<s>`) when we are at the first time step, and otherwise it will return the `past_value` (i.e. the previous element given what time step we're currently at) of `label_sequence`.\n",
     "\n",

bindings/python/cntk/io/__init__.py

@@ -128,7 +128,7 @@ class MinibatchSource(cntk_py.MinibatchSource):
           **Important:**
           Click :cntkwiki:`here <BrainScript-epochSize-and-Python-epoch_size-in-CNTK>`
           for a description of input and label samples.
-        max_sweeps (`int`, defaults to :const:`cntk.io.INFINITELY_REPEAT`): The maximum number of of sweeps over
+        max_sweeps (`int`, defaults to :const:`cntk.io.INFINITELY_REPEAT`): The maximum number of sweeps over
           the input dataset After this number has been reached, the reader returns empty minibatches on
           subsequent calls to func:`next_minibatch`. `max_samples` and `max_sweeps` are mutually exclusive,
           an exception will be raised if both have non-default values.