diff --git a/configure b/configure
index 23603a969d5..e455893ffc8 100755
--- a/configure
+++ b/configure
@@ -230,7 +230,7 @@ if [ "$TF_NEED_MKL" == "1" ]; then # TF_NEED_MKL
     if [ -z "$MKL_INSTALL_PATH" ]; then
       MKL_INSTALL_PATH=$default_mkl_path
     fi
-    # Result returned from "read" will be used unexpanded. That make "~" unuseable.
+    # Result returned from "read" will be used unexpanded. That make "~" unusable.
     # Going through one more level of expansion to handle that.
     MKL_INSTALL_PATH=`${PYTHON_BIN_PATH} -c "import os; print(os.path.realpath(os.path.expanduser('${MKL_INSTALL_PATH}')))"`
   fi
@@ -565,7 +565,7 @@ while true; do
     if [ -z "$CUDNN_INSTALL_PATH" ]; then
       CUDNN_INSTALL_PATH=$default_cudnn_path
     fi
-    # Result returned from "read" will be used unexpanded. That make "~" unuseable.
+    # Result returned from "read" will be used unexpanded. That make "~" unusable.
     # Going through one more level of expansion to handle that.
     CUDNN_INSTALL_PATH=`"${PYTHON_BIN_PATH}" -c "import os; print(os.path.realpath(os.path.expanduser('${CUDNN_INSTALL_PATH}')))"`
   fi
diff --git a/tensorflow/compiler/xla/array4d.h b/tensorflow/compiler/xla/array4d.h
index 1c6ba1f5195..56b638d9782 100644
--- a/tensorflow/compiler/xla/array4d.h
+++ b/tensorflow/compiler/xla/array4d.h
@@ -65,7 +65,7 @@ class Array4D {
     Fill(T());
   }
 
-  // Creates a 4D array, initalized to value.
+  // Creates a 4D array, initialized to value.
   Array4D(int64 planes, int64 depth, int64 height, int64 width, T value)
       : Array4D(planes, depth, height, width) {
     Fill(value);
diff --git a/tensorflow/compiler/xla/client/local_client.h b/tensorflow/compiler/xla/client/local_client.h
index 49ffed4dde6..c903cd27112 100644
--- a/tensorflow/compiler/xla/client/local_client.h
+++ b/tensorflow/compiler/xla/client/local_client.h
@@ -56,7 +56,7 @@ class ExecutableBuildOptions {
 
   // If set, this specifies the layout of the result of the computation. If not
   // set, the service will chose the layout of the result. A Shape is used to
-  // store the layout to accomodate tuple result shapes. A value of nullptr
+  // store the layout to accommodate tuple result shapes. A value of nullptr
   // indicates the option has not been set.
   ExecutableBuildOptions& set_result_layout(const Shape& shape_with_layout);
   const Shape* result_layout() const;
diff --git a/tensorflow/compiler/xla/literal_util.h b/tensorflow/compiler/xla/literal_util.h
index a05dc968ee5..8e06f35b33d 100644
--- a/tensorflow/compiler/xla/literal_util.h
+++ b/tensorflow/compiler/xla/literal_util.h
@@ -128,7 +128,7 @@ class LiteralUtil {
 
   // Creates a new value that has the equivalent value as literal, but conforms
   // to new_layout; e.g. a literal matrix that was in {0, 1} minor-to-major
-  // dimension layout can be re-layed-out as {1, 0} minor-to-major dimension
+  // dimension layout can be re-laid-out as {1, 0} minor-to-major dimension
   // layout and the value in the cell at any given logical index (i0, i1) will
   // be the same.
   //
diff --git a/tensorflow/compiler/xla/service/buffer_liveness_test.cc b/tensorflow/compiler/xla/service/buffer_liveness_test.cc
index c2184cc680b..bee9a351f5d 100644
--- a/tensorflow/compiler/xla/service/buffer_liveness_test.cc
+++ b/tensorflow/compiler/xla/service/buffer_liveness_test.cc
@@ -628,7 +628,7 @@ class FusedDynamicUpdateSliceLivenessTest : public BufferLivenessTest {
         BufferLiveness::Run(module.get(),
                             MakeUnique<DependencyHloOrdering>(module.get()))
             .ConsumeValueOrDie();
-    // Return whether or not buffers interfernce is detected between
+    // Return whether or not buffers interference is detected between
     // 'tuple_param0' and 'tuple_root' at shape index '{1}'.
     return TupleElementsMayInterfere(*liveness, tuple_param0, tuple_root, {1});
   }
@@ -740,7 +740,7 @@ class DynamicUpdateSliceLivenessTest : public BufferLivenessTest {
         BufferLiveness::Run(module.get(),
                             MakeUnique<DependencyHloOrdering>(module.get()))
             .ConsumeValueOrDie();
-    // Return whether or not buffers interfernce is detected between
+    // Return whether or not buffers interference is detected between
     // 'tuple_param0' and 'tuple_root' at shape index '{1}'.
     return TupleElementsMayInterfere(*liveness, tuple_param0, tuple_root, {1});
   }
diff --git a/tensorflow/compiler/xla/service/copy_insertion.cc b/tensorflow/compiler/xla/service/copy_insertion.cc
index 907b0307d4b..3a1a9fe8709 100644
--- a/tensorflow/compiler/xla/service/copy_insertion.cc
+++ b/tensorflow/compiler/xla/service/copy_insertion.cc
@@ -141,7 +141,7 @@ class InstructionCopier {
   Status RecordAmbiguousOrNonDistinctIndices(
       const TuplePointsToAnalysis& points_to_analysis);
 
-  // Records instruction buffer indices which have interferring live ranges
+  // Records instruction buffer indices which have interfering live ranges
   // with 'other_instruction' buffers at same index.
   Status RecordIndicesWhichInterfereWithOtherInstruction(
       const BufferLiveness& liveness, const HloInstruction* other_instruction,
@@ -431,7 +431,7 @@ HloInstruction* InstructionCopier::Copy() {
   return copy;
 }
 
-// The 'read_only_indices' are initalized based on points-to analysis on the
+// The 'read_only_indices' are initialized based on points-to analysis on the
 // while body corresponding to 'while_hlo'. If the init buffer corresponding to
 // a read-only index aliases with an entry parameter (or constant), it cannot be
 // considered read-only, and must be copied. This is necessary because some
diff --git a/tensorflow/compiler/xla/service/copy_insertion_test.cc b/tensorflow/compiler/xla/service/copy_insertion_test.cc
index 4a14fc5397b..661f682e38a 100644
--- a/tensorflow/compiler/xla/service/copy_insertion_test.cc
+++ b/tensorflow/compiler/xla/service/copy_insertion_test.cc
@@ -972,7 +972,7 @@ TEST_F(WhileCopyInsertionTest, InitPointsToNonDistinct) {
                                   op::Copy(old_init->operand(1)->operand(0)))));
 }
 
-// Tests while init instruction buffer which interfers with while result buffer.
+// Tests while init instruction buffer which interferes with while result buffer.
 //
 //     init_data = Broadcast(...)
 //     add_unrelated = Add(init_data) // takes a reference to cause interference
diff --git a/tensorflow/compiler/xla/service/gpu/convolution_thunk.cc b/tensorflow/compiler/xla/service/gpu/convolution_thunk.cc
index f6b7fe1e8ef..94acf5a3594 100644
--- a/tensorflow/compiler/xla/service/gpu/convolution_thunk.cc
+++ b/tensorflow/compiler/xla/service/gpu/convolution_thunk.cc
@@ -125,7 +125,7 @@ tensorflow::Status ConvolutionThunk::ExecuteOnStream(
   CHECK_LE(num_dimensions, 3);
   // cuDNN does not support 1D convolutions. We therefore express 1D
   // convolutions as 2D convolutions where the first spatial dimension is 1.
-  // This matches the behaviour of TF (see definition of conv1d in
+  // This matches the behavior of TF (see definition of conv1d in
   // tensorflow/python/ops/nn_ops.py).
   const int effective_num_dimensions = std::max(2, num_dimensions);
 
diff --git a/tensorflow/compiler/xla/service/gpu/llvm_gpu_backend/gpu_backend_lib.cc b/tensorflow/compiler/xla/service/gpu/llvm_gpu_backend/gpu_backend_lib.cc
index 383729185df..4f34cb77b03 100644
--- a/tensorflow/compiler/xla/service/gpu/llvm_gpu_backend/gpu_backend_lib.cc
+++ b/tensorflow/compiler/xla/service/gpu/llvm_gpu_backend/gpu_backend_lib.cc
@@ -405,9 +405,9 @@ StatusOr<string> CompileModuleToPtx(llvm::Module* module,
 
   AddOptimizationPasses(flags->opt_level, /*size_level=*/0,
                         target_machine.get(), &module_passes, &function_passes);
-  // Loop unrolling exposes more opportunites for SROA. Therefore, we run SROA
+  // Loop unrolling exposes more opportunities for SROA. Therefore, we run SROA
   // again after the standard optimization passes [http://b/13329423].
-  // TODO(jingyue): SROA may further expose more optimization opportunites, such
+  // TODO(jingyue): SROA may further expose more optimization opportunities, such
   // as more precise alias analysis and more function inlining (SROA may change
   // the inlining cost of a function). For now, running SROA already emits good
   // enough code for the evaluated benchmarks. We may want to run more
diff --git a/tensorflow/compiler/xla/service/gpu/partition_assignment.h b/tensorflow/compiler/xla/service/gpu/partition_assignment.h
index 8ac4c599663..8f7fce884ac 100644
--- a/tensorflow/compiler/xla/service/gpu/partition_assignment.h
+++ b/tensorflow/compiler/xla/service/gpu/partition_assignment.h
@@ -33,7 +33,7 @@ namespace gpu {
 enum class PartitionStrategy {
   // Optimized for latency by allowing maximum number of registers per thread.
   kLatency,
-  // Optimized for throughtput. This may limit registers per thread and cause
+  // Optimized for throughput. This may limit registers per thread and cause
   // longer latency.
   kThroughput
 };
diff --git a/tensorflow/compiler/xla/service/gpu/while_transformer.cc b/tensorflow/compiler/xla/service/gpu/while_transformer.cc
index 61bc6f60557..61a9e7e9e1b 100644
--- a/tensorflow/compiler/xla/service/gpu/while_transformer.cc
+++ b/tensorflow/compiler/xla/service/gpu/while_transformer.cc
@@ -37,7 +37,7 @@ namespace {
 // patterns to match.
 //
 // Each ExprTree node is comprised of an HloOpcode, and a set of operands (each
-// of type ExprTree). Operands can be added by specifing the index and HloOpcode
+// of type ExprTree). Operands can be added by specifying the index and HloOpcode
 // of the operand.
 //
 // For example, the following computation:
diff --git a/tensorflow/compiler/xla/service/hlo_constant_folding.h b/tensorflow/compiler/xla/service/hlo_constant_folding.h
index f45eccf8253..331480bd029 100644
--- a/tensorflow/compiler/xla/service/hlo_constant_folding.h
+++ b/tensorflow/compiler/xla/service/hlo_constant_folding.h
@@ -21,7 +21,7 @@ limitations under the License.
 
 namespace xla {
 
-// A pass which performs constant folding in order to avoid unecessary
+// A pass which performs constant folding in order to avoid unnecessary
 // computation on constants.
 class HloConstantFolding : public HloPassInterface {
  public:
diff --git a/tensorflow/compiler/xla/service/hlo_cost_analysis.h b/tensorflow/compiler/xla/service/hlo_cost_analysis.h
index 7d22548234f..b2c40f75ca4 100644
--- a/tensorflow/compiler/xla/service/hlo_cost_analysis.h
+++ b/tensorflow/compiler/xla/service/hlo_cost_analysis.h
@@ -133,7 +133,7 @@ class HloCostAnalysis : public DfsHloVisitor {
   int64 bytes_accessed() const { return bytes_accessed_; }
 
  private:
-  // An FMA counts as two floating point operations in these analyses.
+  // An FMA counts as two floating point operations in these analyzes.
   static constexpr int64 kFmaFlops = 2;
 
   // Utility function to handle all element-wise operations.
diff --git a/tensorflow/compiler/xla/service/hlo_evaluator.h b/tensorflow/compiler/xla/service/hlo_evaluator.h
index 50cb32eb85c..040fd3d73c8 100644
--- a/tensorflow/compiler/xla/service/hlo_evaluator.h
+++ b/tensorflow/compiler/xla/service/hlo_evaluator.h
@@ -104,7 +104,7 @@ class HloEvaluator : public DfsHloVisitorWithDefault {
                            std::hash<int>>
       typed_visitors_;
 
-  // Tracks the HLO instruciton and its evaluated literal result.
+  // Tracks the HLO instruction and its evaluated literal result.
   // TODO(b/35950897): have better memory management here to free instructions
   // that are no longer a parent for any other subsequent instruction in
   // post-orderring.
diff --git a/tensorflow/compiler/xla/service/hlo_ordering.cc b/tensorflow/compiler/xla/service/hlo_ordering.cc
index 725ce17d664..d1ef8cb6918 100644
--- a/tensorflow/compiler/xla/service/hlo_ordering.cc
+++ b/tensorflow/compiler/xla/service/hlo_ordering.cc
@@ -343,7 +343,7 @@ class ListScheduler {
     return freed_bytes;
   }
 
-  // Construct the scheduling priority of the given instruciton.
+  // Construct the scheduling priority of the given instruction.
   Priority GetPriority(const HloInstruction* instruction) {
     return {BytesFreedIfScheduled(instruction), instruction->user_count()};
   }
diff --git a/tensorflow/compiler/xla/service/layout_assignment.h b/tensorflow/compiler/xla/service/layout_assignment.h
index ae53702a28a..689e4510ed2 100644
--- a/tensorflow/compiler/xla/service/layout_assignment.h
+++ b/tensorflow/compiler/xla/service/layout_assignment.h
@@ -273,7 +273,7 @@ class LayoutAssignment : public HloPassInterface {
     return Status::OK();
   }
 
-  // This method can be overriden to mark instructions as requiring the operands
+  // This method can be overridden to mark instructions as requiring the operands
   // to have the same layout as the result, for performance or correctness. This
   // will propagate constraints through the instruction from the result into the
   // operands.
diff --git a/tensorflow/compiler/xla/service/llvm_ir/llvm_util.h b/tensorflow/compiler/xla/service/llvm_ir/llvm_util.h
index 28488ca9991..964b359bb09 100644
--- a/tensorflow/compiler/xla/service/llvm_ir/llvm_util.h
+++ b/tensorflow/compiler/xla/service/llvm_ir/llvm_util.h
@@ -130,7 +130,7 @@ llvm::AllocaInst* EmitAllocaAtFunctionEntryWithCount(
     llvm::Type* type, llvm::Value* element_count, tensorflow::StringPiece name,
     llvm::IRBuilder<>* ir_builder, int alignment = 0);
 
-// Creates a basic block with the same context and funtion as for the
+// Creates a basic block with the same context and function as for the
 // builder. Inserts at the end of the function if insert_before is
 // null.
 llvm::BasicBlock* CreateBasicBlock(llvm::BasicBlock* insert_before,
diff --git a/tensorflow/compiler/xla/tests/hlo_test_base.h b/tensorflow/compiler/xla/tests/hlo_test_base.h
index 91fc9b87cd5..d94602ffda2 100644
--- a/tensorflow/compiler/xla/tests/hlo_test_base.h
+++ b/tensorflow/compiler/xla/tests/hlo_test_base.h
@@ -65,7 +65,7 @@ class HloTestBase : public ::testing::Test {
   perftools::gputools::DeviceMemoryBase TransferToDevice(
       const Literal& literal);
 
-  // Transfers the array refered to by the given handle from the device and
+  // Transfers the array referred to by the given handle from the device and
   // returns as a Literal.
   std::unique_ptr<Literal> TransferFromDevice(
       const Shape& shape, perftools::gputools::DeviceMemoryBase device_base);
diff --git a/tensorflow/compiler/xla/tests/prng_test.cc b/tensorflow/compiler/xla/tests/prng_test.cc
index 5a6aa467e54..a0f98fcfef3 100644
--- a/tensorflow/compiler/xla/tests/prng_test.cc
+++ b/tensorflow/compiler/xla/tests/prng_test.cc
@@ -194,7 +194,7 @@ XLA_TEST_F(PrngTest, MapUsingRng) {
   }
 }
 
-// This tests demonstrates the global seeding behaviour.
+// This tests demonstrates the global seeding behavior.
 // * If a seed is passed in via Execute (ExecuteAndTransfer) then the output is
 //   fixed (i.e., there is a single output for a given seed);
 // * If no seed is passed in then the output of every call can be different;
diff --git a/tensorflow/contrib/bayesflow/python/ops/monte_carlo_impl.py b/tensorflow/contrib/bayesflow/python/ops/monte_carlo_impl.py
index 55e0e6d57b3..3590f940acf 100644
--- a/tensorflow/contrib/bayesflow/python/ops/monte_carlo_impl.py
+++ b/tensorflow/contrib/bayesflow/python/ops/monte_carlo_impl.py
@@ -177,7 +177,7 @@ def _logspace_mean(log_values):
       `Log[Mean[values]]`.
   """
   # center = Max[Log[values]],  with stop-gradient
-  # The center hopefully keep the exponentiated term small.  It is cancelled
+  # The center hopefully keep the exponentiated term small.  It is canceled
   # from the final result, so putting stop gradient on it will not change the
   # final result.  We put stop gradient on to eliminate unnecessary computation.
   center = array_ops.stop_gradient(_sample_max(log_values))
diff --git a/tensorflow/contrib/boosted_trees/lib/testutil/random_tree_gen.h b/tensorflow/contrib/boosted_trees/lib/testutil/random_tree_gen.h
index dc584bbd3cf..5e12429ba77 100644
--- a/tensorflow/contrib/boosted_trees/lib/testutil/random_tree_gen.h
+++ b/tensorflow/contrib/boosted_trees/lib/testutil/random_tree_gen.h
@@ -42,7 +42,7 @@ class RandomTreeGen {
   boosted_trees::trees::DecisionTreeConfig Generate(
       const boosted_trees::trees::DecisionTreeConfig& tree);
 
-  // Requried: depth >= 1; tree_count >= 1.
+  // Required: depth >= 1; tree_count >= 1.
   boosted_trees::trees::DecisionTreeEnsembleConfig GenerateEnsemble(
       int dept, int tree_count);
 
diff --git a/tensorflow/contrib/cloud/kernels/bigquery_reader_ops.cc b/tensorflow/contrib/cloud/kernels/bigquery_reader_ops.cc
index 02a759eefd1..093000559b7 100644
--- a/tensorflow/contrib/cloud/kernels/bigquery_reader_ops.cc
+++ b/tensorflow/contrib/cloud/kernels/bigquery_reader_ops.cc
@@ -46,7 +46,7 @@ Status GetTableAttrs(OpKernelConstruction* context, string* project_id,
 
 }  // namespace
 
-// Note that overriden methods with names ending in "Locked" are called by
+// Note that overridden methods with names ending in "Locked" are called by
 // ReaderBase while a mutex is held.
 // See comments for ReaderBase.
 class BigQueryReader : public ReaderBase {
diff --git a/tensorflow/contrib/cloud/python/ops/bigquery_reader_ops_test.py b/tensorflow/contrib/cloud/python/ops/bigquery_reader_ops_test.py
index 9acdb4b102b..493b3c6f1b5 100644
--- a/tensorflow/contrib/cloud/python/ops/bigquery_reader_ops_test.py
+++ b/tensorflow/contrib/cloud/python/ops/bigquery_reader_ops_test.py
@@ -46,7 +46,7 @@ _TABLE = "test-table"
 # The values for rows are generated such that some columns have null values. The
 # general formula here is:
 #   - The int64 column is present in every row.
-#   - The string column is only avaiable in even rows.
+#   - The string column is only available in even rows.
 #   - The float column is only available in every third row.
 _ROWS = [[0, "s_0", 0.1], [1, None, None], [2, "s_2", None], [3, None, 3.1],
          [4, "s_4", None], [5, None, None], [6, "s_6", 6.1], [7, None, None],
diff --git a/tensorflow/contrib/cudnn_rnn/python/ops/cudnn_rnn_ops.py b/tensorflow/contrib/cudnn_rnn/python/ops/cudnn_rnn_ops.py
index 4f70b275e8f..cc0c7b08296 100644
--- a/tensorflow/contrib/cudnn_rnn/python/ops/cudnn_rnn_ops.py
+++ b/tensorflow/contrib/cudnn_rnn/python/ops/cudnn_rnn_ops.py
@@ -141,7 +141,7 @@ _cudnn_rnn_common_doc_string = """
     * Once a while, the user saves the parameter buffer into model checkpoints
         with Saver.save().
     * When restoring, the user creates a RNNParamsSaveable object and uses
-      Saver.restore() to restore the paramter buffer from the canonical format
+      Saver.restore() to restore the parameter buffer from the canonical format
       to a user-defined format, as well as to restore other savable objects
       in the checkpoint file.
 """
diff --git a/tensorflow/contrib/data/python/framework/function.py b/tensorflow/contrib/data/python/framework/function.py
index a446c935aba..6aa44a736ad 100644
--- a/tensorflow/contrib/data/python/framework/function.py
+++ b/tensorflow/contrib/data/python/framework/function.py
@@ -39,7 +39,7 @@ class _ExperimentalFuncGraph(function._FuncGraph):
   _ExperimentalFuncGraph overrides ops.Graph's create_op() so that we can keep
   track of every inputs into every op created inside the function.  If
   any input is from other graphs, we keep track of it in self.capture
-  and substitue the input with a place holder.
+  and substitute the input with a place holder.
 
   Each captured input's corresponding place holder is converted into a
   function argument and the caller passes in the captured tensor.
diff --git a/tensorflow/contrib/distributions/python/kernel_tests/vector_student_t_test.py b/tensorflow/contrib/distributions/python/kernel_tests/vector_student_t_test.py
index 9d0ffd63763..b8a3a262ce0 100644
--- a/tensorflow/contrib/distributions/python/kernel_tests/vector_student_t_test.py
+++ b/tensorflow/contrib/distributions/python/kernel_tests/vector_student_t_test.py
@@ -38,7 +38,7 @@ class _FakeVectorStudentT(object):
 
   Other `Vector*` implementations need only test new code. That we don't need
   to test every Vector* distribution is good because there aren't SciPy
-  analogues and reimplementing everything in NumPy sort of defeats the point of
+  analogs and reimplementing everything in NumPy sort of defeats the point of
   having the `TransformedDistribution + Affine` API.
   """
 
diff --git a/tensorflow/contrib/distributions/python/ops/binomial.py b/tensorflow/contrib/distributions/python/ops/binomial.py
index ecf6a611565..9304a56491e 100644
--- a/tensorflow/contrib/distributions/python/ops/binomial.py
+++ b/tensorflow/contrib/distributions/python/ops/binomial.py
@@ -269,7 +269,7 @@ class Binomial(distribution.Distribution):
             message="total_count must be non-negative."),
         distribution_util.assert_integer_form(
             total_count,
-            message="total_count cannot contain fractional componentes."),
+            message="total_count cannot contain fractional components."),
     ], total_count)
 
   def _maybe_assert_valid_sample(self, counts, check_integer=True):
diff --git a/tensorflow/contrib/graph_editor/util.py b/tensorflow/contrib/graph_editor/util.py
index ec32beda5a5..959905e9826 100644
--- a/tensorflow/contrib/graph_editor/util.py
+++ b/tensorflow/contrib/graph_editor/util.py
@@ -130,7 +130,7 @@ def transform_tree(tree, fn, iterable_type=tuple):
     tree: iterable or not. If iterable, its elements (child) can also be
       iterable or not.
     fn: function to apply to each leaves.
-    iterable_type: type use to construct the resulting tree for unknwon
+    iterable_type: type use to construct the resulting tree for unknown
       iterable, typically `list` or `tuple`.
   Returns:
     A tree whose leaves has been transformed by `fn`.
diff --git a/tensorflow/contrib/hooks/README.md b/tensorflow/contrib/hooks/README.md
index c7f88bb1113..84dd6ac8792 100644
--- a/tensorflow/contrib/hooks/README.md
+++ b/tensorflow/contrib/hooks/README.md
@@ -5,7 +5,7 @@ of `SessionRunHook` and are to be used with helpers like `MonitoredSession`
 and `learn.Estimator` that wrap `tensorflow.Session`.
 
 The hooks are called between invocations of `Session.run()` to perform custom
-behaviour.
+behavior.
 
 For example the `ProfilerHook` periodically collects `RunMetadata` after
 `Session.run()` and saves profiling information that can be viewed in a
diff --git a/tensorflow/contrib/image/ops/image_ops.cc b/tensorflow/contrib/image/ops/image_ops.cc
index 6b24eaf2a5e..740854930c4 100644
--- a/tensorflow/contrib/image/ops/image_ops.cc
+++ b/tensorflow/contrib/image/ops/image_ops.cc
@@ -77,7 +77,7 @@ REGISTER_OP("BipartiteMatch")
     .Doc(R"doc(
 Find bipartite matching based on a given distance matrix.
 
-A greedy bi-partite matching alogrithm is used to obtain the matching with the
+A greedy bi-partite matching algorithm is used to obtain the matching with the
 (greedy) minimum distance.
 
 distance_mat: A 2-D float tensor of shape `[num_rows, num_columns]`. It is a
diff --git a/tensorflow/contrib/image/python/ops/image_ops.py b/tensorflow/contrib/image/python/ops/image_ops.py
index da374f8cef5..b396dcea211 100644
--- a/tensorflow/contrib/image/python/ops/image_ops.py
+++ b/tensorflow/contrib/image/python/ops/image_ops.py
@@ -266,7 +266,7 @@ def bipartite_match(
     top_k=-1):
   """Find bipartite matching based on a given distance matrix.
 
-  A greedy bi-partite matching alogrithm is used to obtain the matching with
+  A greedy bi-partite matching algorithm is used to obtain the matching with
   the (greedy) minimum distance.
 
   Args:
diff --git a/tensorflow/contrib/keras/python/keras/applications/resnet50.py b/tensorflow/contrib/keras/python/keras/applications/resnet50.py
index 12f7ca424ed..640cc9a3868 100644
--- a/tensorflow/contrib/keras/python/keras/applications/resnet50.py
+++ b/tensorflow/contrib/keras/python/keras/applications/resnet50.py
@@ -59,7 +59,7 @@ def identity_block(input_tensor, kernel_size, filters, stage, block):
 
   Arguments:
       input_tensor: input tensor
-      kernel_size: defualt 3, the kernel size of middle conv layer at main path
+      kernel_size: default 3, the kernel size of middle conv layer at main path
       filters: list of integers, the filterss of 3 conv layer at main path
       stage: integer, current stage label, used for generating layer names
       block: 'a','b'..., current block label, used for generating layer names
@@ -98,7 +98,7 @@ def conv_block(input_tensor, kernel_size, filters, stage, block, strides=(2,
 
   Arguments:
       input_tensor: input tensor
-      kernel_size: defualt 3, the kernel size of middle conv layer at main path
+      kernel_size: default 3, the kernel size of middle conv layer at main path
       filters: list of integers, the filterss of 3 conv layer at main path
       stage: integer, current stage label, used for generating layer names
       block: 'a','b'..., current block label, used for generating layer names
diff --git a/tensorflow/contrib/keras/python/keras/backend.py b/tensorflow/contrib/keras/python/keras/backend.py
index 905ef13e143..8bc3327552b 100644
--- a/tensorflow/contrib/keras/python/keras/backend.py
+++ b/tensorflow/contrib/keras/python/keras/backend.py
@@ -91,7 +91,7 @@ _IMAGE_DATA_FORMAT = 'channels_last'
 def backend():
   """Publicly accessible method for determining the current backend.
 
-  Only exists for API compatibily with multi-backend Keras.
+  Only exists for API compatibility with multi-backend Keras.
 
   Returns:
       The string "tensorflow".
@@ -2617,7 +2617,7 @@ def in_train_phase(x, alt, training=None):
           (tensor or callable that returns a tensor).
       training: Optional scalar tensor
           (or Python boolean, or Python integer)
-          specifing the learning phase.
+          specifying the learning phase.
 
   Returns:
       Either `x` or `alt` based on the `training` flag.
@@ -2660,7 +2660,7 @@ def in_test_phase(x, alt, training=None):
           (tensor or callable that returns a tensor).
       training: Optional scalar tensor
           (or Python boolean, or Python integer)
-          specifing the learning phase.
+          specifying the learning phase.
 
   Returns:
       Either `x` or `alt` based on `K.learning_phase`.
diff --git a/tensorflow/contrib/keras/python/keras/engine/topology.py b/tensorflow/contrib/keras/python/keras/engine/topology.py
index 8bc0c412b50..6f45bf74336 100644
--- a/tensorflow/contrib/keras/python/keras/engine/topology.py
+++ b/tensorflow/contrib/keras/python/keras/engine/topology.py
@@ -1546,7 +1546,7 @@ class Container(Layer):
     """Retrieve the model's updates.
 
     Will only include updates that are either
-    inconditional, or conditional on inputs to this model
+    unconditional, or conditional on inputs to this model
     (e.g. will not include updates that depend on tensors
     that aren't inputs to this model).
 
@@ -1573,7 +1573,7 @@ class Container(Layer):
     """Retrieve the model's losses.
 
     Will only include losses that are either
-    inconditional, or conditional on inputs to this model
+    unconditional, or conditional on inputs to this model
     (e.g. will not include losses that depend on tensors
     that aren't inputs to this model).
 
diff --git a/tensorflow/contrib/keras/python/keras/wrappers/scikit_learn.py b/tensorflow/contrib/keras/python/keras/wrappers/scikit_learn.py
index 9f8cea375b7..0d04fc120f1 100644
--- a/tensorflow/contrib/keras/python/keras/wrappers/scikit_learn.py
+++ b/tensorflow/contrib/keras/python/keras/wrappers/scikit_learn.py
@@ -109,7 +109,7 @@ class BaseWrapper(object):
     """Gets parameters for this estimator.
 
     Arguments:
-        **params: ignored (exists for API compatiblity).
+        **params: ignored (exists for API compatibility).
 
     Returns:
         Dictionary of parameter names mapped to their values.
diff --git a/tensorflow/contrib/kernel_methods/python/mappers/random_fourier_features_test.py b/tensorflow/contrib/kernel_methods/python/mappers/random_fourier_features_test.py
index 200d00b6637..6f4a2644859 100644
--- a/tensorflow/contrib/kernel_methods/python/mappers/random_fourier_features_test.py
+++ b/tensorflow/contrib/kernel_methods/python/mappers/random_fourier_features_test.py
@@ -85,7 +85,7 @@ class RandomFourierFeatureMapperTest(TensorFlowTestCase):
       mapped_x = rffm.map(x)
       mapped_x_copy = rffm.map(x)
       # Two different evaluations of tensors output by map on the same input
-      # are identical because the same paramaters are used for the mappings.
+      # are identical because the same parameters are used for the mappings.
       self.assertAllClose(mapped_x.eval(), mapped_x_copy.eval(), atol=0.001)
 
   def testTwoMapperObjects(self):
diff --git a/tensorflow/contrib/labeled_tensor/python/ops/core.py b/tensorflow/contrib/labeled_tensor/python/ops/core.py
index e6aded92ca5..d886a17c498 100644
--- a/tensorflow/contrib/labeled_tensor/python/ops/core.py
+++ b/tensorflow/contrib/labeled_tensor/python/ops/core.py
@@ -618,7 +618,7 @@ def identity(labeled_tensor, name=None):
 def slice_function(labeled_tensor, selection, name=None):
   """Slice out a subset of the tensor.
 
-  This is an analogue of tf.slice.
+  This is an analog of tf.slice.
   For example:
   >>> tensor = tf.reshape(tf.range(0, 6), [3, 2])
   >>> labeled_tensor = lt.LabeledTensor(tensor, ['a', ('b', ['foo', 'bar'])])
diff --git a/tensorflow/contrib/layers/python/layers/target_column_test.py b/tensorflow/contrib/layers/python/layers/target_column_test.py
index 1baa663151a..d5d03fb1ebc 100644
--- a/tensorflow/contrib/layers/python/layers/target_column_test.py
+++ b/tensorflow/contrib/layers/python/layers/target_column_test.py
@@ -28,7 +28,7 @@ from tensorflow.python.platform import test
 
 class RegressionTargetColumnTest(test.TestCase):
 
-  # TODO(zakaria): test multilabel regresssion.
+  # TODO(zakaria): test multilabel regression.
   def testRegression(self):
     target_column = target_column_lib.regression_target()
     with ops.Graph().as_default(), session.Session() as sess:
diff --git a/tensorflow/contrib/learn/python/learn/dataframe/tensorflow_dataframe.py b/tensorflow/contrib/learn/python/learn/dataframe/tensorflow_dataframe.py
index b17a4b8d05b..f316c5c9804 100644
--- a/tensorflow/contrib/learn/python/learn/dataframe/tensorflow_dataframe.py
+++ b/tensorflow/contrib/learn/python/learn/dataframe/tensorflow_dataframe.py
@@ -97,7 +97,7 @@ class TensorFlowDataFrame(df.DataFrame):
       graph: the `Graph` in which the `DataFrame` should be built.
       session: the `Session` in which to run the columns of the `DataFrame`.
       start_queues: if true, queues will be started before running and halted
-        after producting `n` batches.
+        after producing `n` batches.
       initialize_variables: if true, variables will be initialized.
       **kwargs: Additional keyword arguments e.g. `num_epochs`.
 
diff --git a/tensorflow/contrib/learn/python/learn/estimators/estimator.py b/tensorflow/contrib/learn/python/learn/estimators/estimator.py
index d7ba2209ada..ddd3d087e7b 100644
--- a/tensorflow/contrib/learn/python/learn/estimators/estimator.py
+++ b/tensorflow/contrib/learn/python/learn/estimators/estimator.py
@@ -89,7 +89,7 @@ SCIKIT_DECOUPLE_INSTRUCTIONS = (
 
 
 def _verify_input_args(x, y, input_fn, feed_fn, batch_size):
-  """Verifies validity of co-existance of input arguments."""
+  """Verifies validity of co-existence of input arguments."""
   if input_fn is None:
     if x is None:
       raise ValueError('Either x or input_fn must be provided.')
@@ -358,7 +358,7 @@ class BaseEstimator(
   """
   __metaclass__ = abc.ABCMeta
 
-  # Note that for Google users, this is overriden with
+  # Note that for Google users, this is overridden with
   # learn_runner.EstimatorConfig.
   # TODO(wicke): Remove this once launcher takes over config functionality
   _Config = run_config.RunConfig  # pylint: disable=invalid-name
@@ -703,7 +703,7 @@ class BaseEstimator(
   def _get_eval_ops(self, features, labels, metrics):
     """Method that builds model graph and returns evaluation ops.
 
-    Expected to be overriden by sub-classes that require custom support.
+    Expected to be overridden by sub-classes that require custom support.
 
     Args:
       features: `Tensor` or `dict` of `Tensor` objects.
@@ -1149,7 +1149,7 @@ class Estimator(BaseEstimator):
   def _get_train_ops(self, features, labels):
     """Method that builds model graph and returns trainer ops.
 
-    Expected to be overriden by sub-classes that require custom support.
+    Expected to be overridden by sub-classes that require custom support.
     This implementation uses `model_fn` passed as parameter to constructor to
     build model.
 
@@ -1165,7 +1165,7 @@ class Estimator(BaseEstimator):
   def _get_eval_ops(self, features, labels, metrics):
     """Method that builds model graph and returns evaluation ops.
 
-    Expected to be overriden by sub-classes that require custom support.
+    Expected to be overridden by sub-classes that require custom support.
     This implementation uses `model_fn` passed as parameter to constructor to
     build model.
 
@@ -1204,7 +1204,7 @@ class Estimator(BaseEstimator):
   def _get_predict_ops(self, features):
     """Method that builds model graph and returns prediction ops.
 
-    Expected to be overriden by sub-classes that require custom support.
+    Expected to be overridden by sub-classes that require custom support.
     This implementation uses `model_fn` passed as parameter to constructor to
     build model.
 
diff --git a/tensorflow/contrib/learn/python/learn/estimators/head.py b/tensorflow/contrib/learn/python/learn/estimators/head.py
index e4ef6996d8d..d270d89c12b 100644
--- a/tensorflow/contrib/learn/python/learn/estimators/head.py
+++ b/tensorflow/contrib/learn/python/learn/estimators/head.py
@@ -620,7 +620,7 @@ def _create_model_fn_ops(features,
     weight_tensor = _weight_tensor(features, weight_column_name)
     loss, weighted_average_loss = loss_fn(labels, logits, weight_tensor)
     # Uses the deprecated API to set the tag explicitly.
-    # Without it, trianing and eval losses will show up in different graphs.
+    # Without it, training and eval losses will show up in different graphs.
     logging_ops.scalar_summary(
         _summary_key(head_name, mkey.LOSS), weighted_average_loss)
 
@@ -1141,7 +1141,7 @@ def _to_labels_tensor(labels, label_name):
   """Returns label as a tensor.
 
   Args:
-    labels: Label `Tensor` or `SparseTensor` or a dict containig labels.
+    labels: Label `Tensor` or `SparseTensor` or a dict containing labels.
     label_name: Label name if labels is a dict.
 
   Returns:
@@ -1575,7 +1575,7 @@ class _MultiHead(Head):
     Args:
       all_model_fn_ops: list of ModelFnOps for the individual heads.
       train_op_fn: Function to create train op. See `create_model_fn_ops`
-          documentaion for more details.
+          documentation for more details.
 
     Returns:
       ModelFnOps that merges all heads for TRAIN.
diff --git a/tensorflow/contrib/learn/python/learn/estimators/rnn_common.py b/tensorflow/contrib/learn/python/learn/estimators/rnn_common.py
index 6bb2b8b2aad..0f09b111bd8 100644
--- a/tensorflow/contrib/learn/python/learn/estimators/rnn_common.py
+++ b/tensorflow/contrib/learn/python/learn/estimators/rnn_common.py
@@ -119,7 +119,7 @@ def apply_dropout(cells, dropout_keep_probabilities, random_seed=None):
   """
   if len(dropout_keep_probabilities) != len(cells) + 1:
     raise ValueError(
-        'The number of dropout probabilites must be one greater than the '
+        'The number of dropout probabilities must be one greater than the '
         'number of cells. Got {} cells and {} dropout probabilities.'.format(
             len(cells), len(dropout_keep_probabilities)))
   wrapped_cells = [
diff --git a/tensorflow/contrib/learn/python/learn/estimators/run_config.py b/tensorflow/contrib/learn/python/learn/estimators/run_config.py
index 7af1c541c6c..3aaee5862df 100644
--- a/tensorflow/contrib/learn/python/learn/estimators/run_config.py
+++ b/tensorflow/contrib/learn/python/learn/estimators/run_config.py
@@ -309,7 +309,7 @@ class RunConfig(ClusterConfig, core_run_config.RunConfig):
     Args:
       whitelist: A list of the string names of the properties uid should not
         include. If `None`, defaults to `_DEFAULT_UID_WHITE_LIST`, which
-        includes most properites user allowes to change.
+        includes most properties user allowes to change.
 
     Returns:
       A uid string.
diff --git a/tensorflow/contrib/learn/python/learn/experiment.py b/tensorflow/contrib/learn/python/learn/experiment.py
index d82bc321e76..c60ecac5df4 100644
--- a/tensorflow/contrib/learn/python/learn/experiment.py
+++ b/tensorflow/contrib/learn/python/learn/experiment.py
@@ -53,7 +53,7 @@ class Experiment(object):
   """
 
   # TODO(ispir): remove delay_workers_by_global_step and make global step based
-  # waiting as only behaviour.
+  # waiting as only behavior.
   @deprecated_args(
       "2016-10-23",
       "local_eval_frequency is deprecated as local_run will be renamed to "
@@ -550,7 +550,7 @@ class Experiment(object):
     eval_result = None
 
     # Set the default value for train_steps_per_iteration, which will be
-    # overriden by other settings.
+    # overridden by other settings.
     train_steps_per_iteration = 1000
     if self._train_steps_per_iteration is not None:
       train_steps_per_iteration = self._train_steps_per_iteration
diff --git a/tensorflow/contrib/learn/python/learn/learn_runner.py b/tensorflow/contrib/learn/python/learn/learn_runner.py
index a3398a87e1e..943c5553140 100644
--- a/tensorflow/contrib/learn/python/learn/learn_runner.py
+++ b/tensorflow/contrib/learn/python/learn/learn_runner.py
@@ -155,7 +155,7 @@ def run(experiment_fn, output_dir=None, schedule=None, run_config=None,
       to create the `Estimator` (passed as `model_dir` to its constructor). It
       must return an `Experiment`. For this case, `run_config` and `hparams`
       must be None.
-      2) It accpets two arguments `run_config` and `hparams`, which should be
+      2) It accepts two arguments `run_config` and `hparams`, which should be
       used to create the `Estimator` (`run_config` passed as `config` to its
       constructor; `hparams` used as the hyper-paremeters of the model).
       It must return an `Experiment`. For this case, `output_dir` must be None.
diff --git a/tensorflow/contrib/learn/python/learn/ops/seq2seq_ops.py b/tensorflow/contrib/learn/python/learn/ops/seq2seq_ops.py
index 0faba7cee5e..45727faab43 100644
--- a/tensorflow/contrib/learn/python/learn/ops/seq2seq_ops.py
+++ b/tensorflow/contrib/learn/python/learn/ops/seq2seq_ops.py
@@ -140,7 +140,7 @@ def rnn_seq2seq(encoder_inputs,
     scope: Scope to use, if None new will be produced.
 
   Returns:
-    List of tensors for outputs and states for trianing and sampling sub-graphs.
+    List of tensors for outputs and states for training and sampling sub-graphs.
   """
   with vs.variable_scope(scope or "rnn_seq2seq"):
     _, last_enc_state = rnn.static_rnn(
diff --git a/tensorflow/contrib/learn/python/learn/preprocessing/categorical_vocabulary.py b/tensorflow/contrib/learn/python/learn/preprocessing/categorical_vocabulary.py
index 9d4fed99987..5709955c49f 100644
--- a/tensorflow/contrib/learn/python/learn/preprocessing/categorical_vocabulary.py
+++ b/tensorflow/contrib/learn/python/learn/preprocessing/categorical_vocabulary.py
@@ -128,9 +128,9 @@ class CategoricalVocabulary(object):
       Class name.
 
     Raises:
-      ValueError: if this vocabulary wasn't initalized with support_reverse.
+      ValueError: if this vocabulary wasn't initialized with support_reverse.
     """
     if not self._support_reverse:
-      raise ValueError("This vocabulary wasn't initalized with "
+      raise ValueError("This vocabulary wasn't initialized with "
                        "support_reverse to support reverse() function.")
     return self._reverse_mapping[class_id]
diff --git a/tensorflow/contrib/learn/python/learn/trainable.py b/tensorflow/contrib/learn/python/learn/trainable.py
index 2d1d4604251..972fec026f2 100644
--- a/tensorflow/contrib/learn/python/learn/trainable.py
+++ b/tensorflow/contrib/learn/python/learn/trainable.py
@@ -49,7 +49,7 @@ class Trainable(object):
       steps: Number of steps for which to train model. If `None`, train forever.
         'steps' works incrementally. If you call two times fit(steps=10) then
         training occurs in total 20 steps. If you don't want to have incremental
-        behaviour please set `max_steps` instead. If set, `max_steps` must be
+        behavior please set `max_steps` instead. If set, `max_steps` must be
         `None`.
       batch_size: minibatch size to use on the input, defaults to first
         dimension of `x`. Must be `None` if `input_fn` is provided.
diff --git a/tensorflow/contrib/learn/python/learn/utils/export.py b/tensorflow/contrib/learn/python/learn/utils/export.py
index 36a1f5f60cd..6af22877612 100644
--- a/tensorflow/contrib/learn/python/learn/utils/export.py
+++ b/tensorflow/contrib/learn/python/learn/utils/export.py
@@ -89,7 +89,7 @@ def _export_graph(graph, saver, checkpoint_path, export_dir,
 def generic_signature_fn(examples, unused_features, predictions):
   """Creates generic signature from given examples and predictions.
 
-  This is needed for backward compatibility with default behaviour of
+  This is needed for backward compatibility with default behavior of
   export_estimator.
 
   Args:
diff --git a/tensorflow/contrib/learn/python/learn/utils/saved_model_export_utils.py b/tensorflow/contrib/learn/python/learn/utils/saved_model_export_utils.py
index fa314e69c7a..3f0f3092534 100644
--- a/tensorflow/contrib/learn/python/learn/utils/saved_model_export_utils.py
+++ b/tensorflow/contrib/learn/python/learn/utils/saved_model_export_utils.py
@@ -309,7 +309,7 @@ def get_most_recent_export(export_dir_base):
                      directories.
 
   Returns:
-    A gc.Path, whith is just a namedtuple of (path, export_version).
+    A gc.Path, with is just a namedtuple of (path, export_version).
   """
   select_filter = gc.largest_export_versions(1)
   results = select_filter(gc.get_paths(export_dir_base,
diff --git a/tensorflow/contrib/learn/python/learn/utils/saved_model_export_utils_test.py b/tensorflow/contrib/learn/python/learn/utils/saved_model_export_utils_test.py
index 157744bed1d..a15eadd018f 100644
--- a/tensorflow/contrib/learn/python/learn/utils/saved_model_export_utils_test.py
+++ b/tensorflow/contrib/learn/python/learn/utils/saved_model_export_utils_test.py
@@ -109,7 +109,7 @@ class SavedModelExportUtilsTest(test.TestCase):
     self.assertEqual(actual_signature_def, expected_signature_def)
 
   def test_build_standardized_signature_def_classification2(self):
-    """Tests multiple output tensors that include classes and probabilites."""
+    """Tests multiple output tensors that include classes and probabilities."""
     input_tensors = {
         "input-1":
             array_ops.placeholder(
diff --git a/tensorflow/contrib/legacy_seq2seq/python/kernel_tests/seq2seq_test.py b/tensorflow/contrib/legacy_seq2seq/python/kernel_tests/seq2seq_test.py
index 2898935a478..824a9648693 100644
--- a/tensorflow/contrib/legacy_seq2seq/python/kernel_tests/seq2seq_test.py
+++ b/tensorflow/contrib/legacy_seq2seq/python/kernel_tests/seq2seq_test.py
@@ -837,7 +837,7 @@ class Seq2SeqTest(test.TestCase):
   #     with variable_scope.variable_scope("new"):
   #       _, losses2 = SampleGRUSeq2Seq
   #           inp, out, weights, per_example_loss=True)
-  #       # First loss is scalar, the second one is a 1-dimensinal tensor.
+  #       # First loss is scalar, the second one is a 1-dimensional tensor.
   #       self.assertEqual([], losses1[0].get_shape().as_list())
   #       self.assertEqual([None], losses2[0].get_shape().as_list())
 
diff --git a/tensorflow/contrib/memory_stats/python/kernel_tests/memory_stats_ops_test.py b/tensorflow/contrib/memory_stats/python/kernel_tests/memory_stats_ops_test.py
index 0f3a5f13136..ec25c032f05 100644
--- a/tensorflow/contrib/memory_stats/python/kernel_tests/memory_stats_ops_test.py
+++ b/tensorflow/contrib/memory_stats/python/kernel_tests/memory_stats_ops_test.py
@@ -49,7 +49,7 @@ class MemoryStatsOpsTest(test_util.TensorFlowTestCase):
   # The memory for matrix "a" can be reused for matrix "d". Therefore, this
   # computation needs space for only three matrix plus some small overhead.
   def testChainOfMatmul(self):
-    # MaxBytesInUse is registerd on GPU only. See kernels/memory_stats_ops.cc.
+    # MaxBytesInUse is registered on GPU only. See kernels/memory_stats_ops.cc.
     if not test.is_gpu_available():
       return
 
diff --git a/tensorflow/contrib/metrics/python/ops/metric_ops_test.py b/tensorflow/contrib/metrics/python/ops/metric_ops_test.py
index f42e974e238..f97f03e30e1 100644
--- a/tensorflow/contrib/metrics/python/ops/metric_ops_test.py
+++ b/tensorflow/contrib/metrics/python/ops/metric_ops_test.py
@@ -1507,7 +1507,7 @@ class StreamingAUCTest(test.TestCase):
       self.assertAlmostEqual(1, auc.eval(), 6)
 
   def np_auc(self, predictions, labels, weights):
-    """Computes the AUC explicitely using Numpy.
+    """Computes the AUC explicitly using Numpy.
 
     Args:
       predictions: an ndarray with shape [N].
diff --git a/tensorflow/contrib/rnn/python/ops/rnn_cell.py b/tensorflow/contrib/rnn/python/ops/rnn_cell.py
index 0898e78837b..566c84443d1 100644
--- a/tensorflow/contrib/rnn/python/ops/rnn_cell.py
+++ b/tensorflow/contrib/rnn/python/ops/rnn_cell.py
@@ -466,7 +466,7 @@ class GridLSTMCell(core_rnn_cell.RNNCell):
         state is clipped by this value prior to the cell output activation.
       initializer: (optional) The initializer to use for the weight and
         projection matrices, default None.
-      num_unit_shards: (optional) int, defualt 1, How to split the weight
+      num_unit_shards: (optional) int, default 1, How to split the weight
         matrix. If > 1,the weight matrix is stored across num_unit_shards.
       forget_bias: (optional) float, default 1.0, The initial bias of the
         forget gates, used to reduce the scale of forgetting at the beginning
@@ -1809,12 +1809,12 @@ class PhasedLSTMCell(core_rnn_cell.RNNCell):
           period during which the gates are open.
       trainable_ratio_on: bool, weather ratio_on is trainable.
       period_init_min: float or scalar float Tensor. With value > 0.
-          Minimum value of the initalized period.
+          Minimum value of the initialized period.
           The period values are initialized by drawing from the distribution:
           e^U(log(period_init_min), log(period_init_max))
           Where U(.,.) is the uniform distribution.
       period_init_max: float or scalar float Tensor.
-          With value > period_init_min. Maximum value of the initalized period.
+          With value > period_init_min. Maximum value of the initialized period.
       reuse: (optional) Python boolean describing whether to reuse variables
         in an existing scope. If not `True`, and the existing scope already has
         the given variables, an error is raised.
diff --git a/tensorflow/contrib/seq2seq/python/ops/attention_wrapper.py b/tensorflow/contrib/seq2seq/python/ops/attention_wrapper.py
index fd76882d846..33ee19605d2 100644
--- a/tensorflow/contrib/seq2seq/python/ops/attention_wrapper.py
+++ b/tensorflow/contrib/seq2seq/python/ops/attention_wrapper.py
@@ -474,7 +474,7 @@ class AttentionWrapperState(
 
     Returns:
       A new `AttentionWrapperState` whose properties are the same as
-      this one, except any overriden properties as provided in `kwargs`.
+      this one, except any overridden properties as provided in `kwargs`.
     """
     return super(AttentionWrapperState, self)._replace(**kwargs)
 
diff --git a/tensorflow/contrib/slim/README.md b/tensorflow/contrib/slim/README.md
index 61148c0b269..d37c632be7f 100644
--- a/tensorflow/contrib/slim/README.md
+++ b/tensorflow/contrib/slim/README.md
@@ -352,7 +352,7 @@ we can both ensure that each layer uses the same values and simplify the code:
 ```
 
 As the example illustrates, the use of arg_scope makes the code cleaner,
-simpler and easier to maintain. Notice that while argument values are specifed
+simpler and easier to maintain. Notice that while argument values are specified
 in the arg_scope, they can be overwritten locally. In particular, while
 the padding argument has been set to 'SAME', the second convolution overrides
 it with the value of 'VALID'.
diff --git a/tensorflow/contrib/slim/python/slim/data/dataset_data_provider.py b/tensorflow/contrib/slim/python/slim/data/dataset_data_provider.py
index 3a78c0471d3..82c6b5a6196 100644
--- a/tensorflow/contrib/slim/python/slim/data/dataset_data_provider.py
+++ b/tensorflow/contrib/slim/python/slim/data/dataset_data_provider.py
@@ -33,7 +33,7 @@ To read data using multiple readers simultaneous with shuffling:
       shuffle=True)
   images, labels = pascal_voc_data_provider.get(['images', 'labels'])
 
-Equivalently, one may request different fields of the same sample seperately:
+Equivalently, one may request different fields of the same sample separately:
 
   [images] = pascal_voc_data_provider.get(['images'])
   [labels] = pascal_voc_data_provider.get(['labels'])
diff --git a/tensorflow/contrib/tensorboard/plugins/projector/__init__.py b/tensorflow/contrib/tensorboard/plugins/projector/__init__.py
index 635c569d734..0c630de7f53 100644
--- a/tensorflow/contrib/tensorboard/plugins/projector/__init__.py
+++ b/tensorflow/contrib/tensorboard/plugins/projector/__init__.py
@@ -40,7 +40,7 @@ def visualize_embeddings(summary_writer, config):
   """Stores a config file used by the embedding projector.
 
   Args:
-    summary_writer: The summary writer used for writting events.
+    summary_writer: The summary writer used for writing events.
     config: `tf.contrib.tensorboard.plugins.projector.ProjectorConfig`
       proto that holds the configuration for the projector such as paths to
       checkpoint files and metadata files for the embeddings. If
diff --git a/tensorflow/contrib/tensorboard/plugins/projector/projector_api_test.py b/tensorflow/contrib/tensorboard/plugins/projector/projector_api_test.py
index 91ea6bc7531..5f86f57a1c6 100644
--- a/tensorflow/contrib/tensorboard/plugins/projector/projector_api_test.py
+++ b/tensorflow/contrib/tensorboard/plugins/projector/projector_api_test.py
@@ -46,7 +46,7 @@ class ProjectorApiTest(test.TestCase):
     writer = writer_lib.FileWriter(temp_dir)
     projector.visualize_embeddings(writer, config)
 
-    # Read the configuratin from disk and make sure it matches the original.
+    # Read the configurations from disk and make sure it matches the original.
     with gfile.GFile(os.path.join(temp_dir, 'projector_config.pbtxt')) as f:
       config2 = projector_config_pb2.ProjectorConfig()
       text_format.Parse(f.read(), config2)
diff --git a/tensorflow/contrib/training/python/training/evaluation.py b/tensorflow/contrib/training/python/training/evaluation.py
index bc0c60c85ce..24b733dd29c 100644
--- a/tensorflow/contrib/training/python/training/evaluation.py
+++ b/tensorflow/contrib/training/python/training/evaluation.py
@@ -370,7 +370,7 @@ def evaluate_repeatedly(checkpoint_dir,
 
   One may also consider using a `tf.contrib.training.SummaryAtEndHook` to record
   summaries after the `eval_ops` have run. If `eval_ops` is `None`, the
-  summaries run immedietly after the model checkpoint has been restored.
+  summaries run immediately after the model checkpoint has been restored.
 
   Note that `evaluate_once` creates a local variable used to track the number of
   evaluations run via `tf.contrib.training.get_or_create_eval_step`.
diff --git a/tensorflow/contrib/training/python/training/hparam.py b/tensorflow/contrib/training/python/training/hparam.py
index 2e085936997..c19a36eabcf 100644
--- a/tensorflow/contrib/training/python/training/hparam.py
+++ b/tensorflow/contrib/training/python/training/hparam.py
@@ -422,7 +422,7 @@ class HParams(object):
     elif issubclass(param_type, float):
       typename = 'float'
     else:
-      raise ValueError('Unsupported paramter type: %s' % str(param_type))
+      raise ValueError('Unsupported parameter type: %s' % str(param_type))
 
     suffix = 'list' if is_list else 'value'
     return '_'.join([typename, suffix])
diff --git a/tensorflow/contrib/training/python/training/sequence_queueing_state_saver.py b/tensorflow/contrib/training/python/training/sequence_queueing_state_saver.py
index 2c7c30911c7..9312070e52b 100644
--- a/tensorflow/contrib/training/python/training/sequence_queueing_state_saver.py
+++ b/tensorflow/contrib/training/python/training/sequence_queueing_state_saver.py
@@ -344,7 +344,7 @@ def _prepare_sequence_inputs(inputs, states):
   key = _check_rank(inputs.key, 0)
 
   if length.dtype != dtypes.int32:
-    raise TypeError("length dtype must be int32, but recieved: %s" %
+    raise TypeError("length dtype must be int32, but received: %s" %
                     length.dtype)
   if key.dtype != dtypes.string:
     raise TypeError("key dtype must be string, but received: %s" % key.dtype)
@@ -1673,7 +1673,7 @@ def _move_sparse_tensor_out_context(input_context, input_sequences, num_unroll):
     shape = array_ops.concat(
         [array_ops.expand_dims(value_length, 0), sp_tensor.dense_shape], 0)
 
-    # Construct new indices by mutliplying old ones and prepending [0, n).
+    # Construct new indices by multiplying old ones and prepending [0, n).
     # First multiply indices n times along a newly created 0-dimension.
     multiplied_indices = array_ops.tile(
         array_ops.expand_dims(sp_tensor.indices, 0),
diff --git a/tensorflow/core/common_runtime/constant_folding.cc b/tensorflow/core/common_runtime/constant_folding.cc
index 8fa61d098eb..914683d9fa3 100644
--- a/tensorflow/core/common_runtime/constant_folding.cc
+++ b/tensorflow/core/common_runtime/constant_folding.cc
@@ -83,7 +83,7 @@ bool IsConstantFoldable(const Node* n,
 }
 
 // Returns the constant foldable nodes in `nodes` in topological order.
-// Populates `constant_control_deps` with the non-constant control depedencies
+// Populates `constant_control_deps` with the non-constant control dependencies
 // of each constant node.
 void FindConstantFoldableNodes(
     const Graph* graph, ConstantFoldingOptions opts, std::vector<Node*>* nodes,
diff --git a/tensorflow/core/common_runtime/executor.h b/tensorflow/core/common_runtime/executor.h
index 93b58906dda..e09dc4e3463 100644
--- a/tensorflow/core/common_runtime/executor.h
+++ b/tensorflow/core/common_runtime/executor.h
@@ -74,8 +74,8 @@ class Executor {
   //
   // RunAsync() uses "cancellation_manager", if not nullptr, to
   // register callbacks that should be called if the graph computation
-  // is cancelled. Note that the callbacks merely unblock any
-  // long-running computation, and a cancelled step will terminate by
+  // is canceled. Note that the callbacks merely unblock any
+  // long-running computation, and a canceled step will terminate by
   // returning/calling the DoneCallback as usual.
   //
   // RunAsync() dispatches closures to "runner". Typically, "runner"
diff --git a/tensorflow/core/common_runtime/session_factory.h b/tensorflow/core/common_runtime/session_factory.h
index 2a1632e0359..df3198a70dd 100644
--- a/tensorflow/core/common_runtime/session_factory.h
+++ b/tensorflow/core/common_runtime/session_factory.h
@@ -47,7 +47,7 @@ class SessionFactory {
   // Old sessions may continue to have side-effects on resources not in
   // containers listed in "containers", and thus may affect future
   // sessions' results in ways that are hard to predict.  Thus, if well-defined
-  // behaviour is desired, is it recommended that all containers be listed in
+  // behavior is desired, is it recommended that all containers be listed in
   // "containers".
   //
   // If the "containers" vector is empty, the default container is assumed.
diff --git a/tensorflow/core/common_runtime/simple_graph_execution_state.cc b/tensorflow/core/common_runtime/simple_graph_execution_state.cc
index 3806f9f47f5..b291a6f9948 100644
--- a/tensorflow/core/common_runtime/simple_graph_execution_state.cc
+++ b/tensorflow/core/common_runtime/simple_graph_execution_state.cc
@@ -243,7 +243,7 @@ Status SimpleGraphExecutionState::InitBaseGraph(
       session_options_->config.graph_options().rewrite_options();
 
   if (grappler::MetaOptimizerEnabled(rewrite_options)) {
-    // Adding this functionalty in steps. The first step is to make sure
+    // Adding this functionality in steps. The first step is to make sure
     // we don't break dependencies. The second step will be to turn the
     // functionality on by default.
     grappler::GrapplerItem item;
diff --git a/tensorflow/core/common_runtime/simple_placer.cc b/tensorflow/core/common_runtime/simple_placer.cc
index ae225e8b35d..73f49706b44 100644
--- a/tensorflow/core/common_runtime/simple_placer.cc
+++ b/tensorflow/core/common_runtime/simple_placer.cc
@@ -660,7 +660,7 @@ Status SimplePlacer::Run() {
       if (!edge->IsControlEdge() &&
           (IsRefType(node->input_type(edge->dst_input())) ||
            node->input_type(edge->dst_input()) == DT_RESOURCE)) {
-        // If both the source node and this node have paritally
+        // If both the source node and this node have partially
         // specified a device, then 'node's device should be
         // cleared: the reference edge forces 'node' to be on the
         // same device as the source node.
diff --git a/tensorflow/core/debug/debug_service.proto b/tensorflow/core/debug/debug_service.proto
index 1adba5d653f..63d6668292a 100644
--- a/tensorflow/core/debug/debug_service.proto
+++ b/tensorflow/core/debug/debug_service.proto
@@ -20,7 +20,7 @@ package tensorflow;
 import "tensorflow/core/util/event.proto";
 
 // Reply message from EventListener to the client, i.e., to the source of the
-// Event protocal buffers, e.g., debug ops inserted by a debugged runtime to a
+// Event protocol buffers, e.g., debug ops inserted by a debugged runtime to a
 // TensorFlow graph being executed.
 message EventReply {
   message DebugOpStateChange {
diff --git a/tensorflow/core/distributed_runtime/graph_mgr.h b/tensorflow/core/distributed_runtime/graph_mgr.h
index 50391f47e4d..4ee3711d028 100644
--- a/tensorflow/core/distributed_runtime/graph_mgr.h
+++ b/tensorflow/core/distributed_runtime/graph_mgr.h
@@ -108,9 +108,9 @@ class GraphMgr {
   };
 
   struct Item : public core::RefCounted {
-    // TOOD(zhifengc): Keeps a copy of the original graph if the need arises.
-    // TOOD(zhifengc): Stats, updated by multiple runs potentially.
-    // TOOD(zhifengc): Dup-detection. Ensure step_id only run once.
+    // TODO(zhifengc): Keeps a copy of the original graph if the need arises.
+    // TODO(zhifengc): Stats, updated by multiple runs potentially.
+    // TODO(zhifengc): Dup-detection. Ensure step_id only run once.
     ~Item() override;
 
     // Session handle.
@@ -126,7 +126,7 @@ class GraphMgr {
     // has a root executor which may call into the runtime library.
     std::vector<ExecutionUnit> units;
 
-    // Used to deresgister a cost model when cost model is requried in graph
+    // Used to deresgister a cost model when cost model is required in graph
     // manager.
     GraphMgr* graph_mgr;
   };
@@ -157,7 +157,7 @@ class GraphMgr {
                               CancellationManager* cancellation_manager,
                               StatusCallback done);
 
-  // Don't attempt to process cost models unless explicitely requested for at
+  // Don't attempt to process cost models unless explicitly requested for at
   // least one of the items.
   bool skip_cost_models_ = true;
 
diff --git a/tensorflow/core/distributed_runtime/master.cc b/tensorflow/core/distributed_runtime/master.cc
index e860c99d953..81d938bd4f0 100644
--- a/tensorflow/core/distributed_runtime/master.cc
+++ b/tensorflow/core/distributed_runtime/master.cc
@@ -25,7 +25,7 @@ limitations under the License.
 // A Master discovers remote devices on-demand and keeps track of
 // statistics of those remote devices.
 //
-// Each session analyses the graph, places nodes across available
+// Each session analyzes the graph, places nodes across available
 // devices, and ultimately drives the graph computation by initiating
 // RunGraph on the workers.
 
diff --git a/tensorflow/core/distributed_runtime/rpc/grpc_call.h b/tensorflow/core/distributed_runtime/rpc/grpc_call.h
index 3b45e7e8a70..e85b8ccbd39 100644
--- a/tensorflow/core/distributed_runtime/rpc/grpc_call.h
+++ b/tensorflow/core/distributed_runtime/rpc/grpc_call.h
@@ -89,7 +89,7 @@ class UntypedCall : public core::RefCounted {
   virtual void RequestReceived(Service* service, bool ok) = 0;
 
   // This method will be called either (i) when the server is notified
-  // that the request has been cancelled, or (ii) when the request completes
+  // that the request has been canceled, or (ii) when the request completes
   // normally. The implementation should distinguish these cases by querying
   // the `grpc::ServerContext` associated with the request.
   virtual void RequestCancelled(Service* service, bool ok) = 0;
@@ -175,7 +175,7 @@ class Call : public UntypedCall<Service> {
   }
 
   // Registers `callback` as the function that should be called if and when this
-  // call is cancelled by the client.
+  // call is canceled by the client.
   void SetCancelCallback(std::function<void()> callback) {
     mutex_lock l(mu_);
     cancel_callback_ = std::move(callback);
diff --git a/tensorflow/core/distributed_runtime/rpc/grpc_master_service.cc b/tensorflow/core/distributed_runtime/rpc/grpc_master_service.cc
index f0d311b853e..fae1c5227b2 100644
--- a/tensorflow/core/distributed_runtime/rpc/grpc_master_service.cc
+++ b/tensorflow/core/distributed_runtime/rpc/grpc_master_service.cc
@@ -25,7 +25,7 @@ limitations under the License.
 // A GrpcMasterService discovers remote devices in the background and
 // keeps track of statistics of those remote devices.
 //
-// Each session analyses the graph, places nodes across available
+// Each session analyzes the graph, places nodes across available
 // devices, and ultimately drives the graph computation by initiating
 // RunGraph on workers.
 #include "tensorflow/core/distributed_runtime/rpc/grpc_master_service.h"
diff --git a/tensorflow/core/distributed_runtime/rpc/grpc_session_test.cc b/tensorflow/core/distributed_runtime/rpc/grpc_session_test.cc
index 368468093e9..eeb3e02966a 100644
--- a/tensorflow/core/distributed_runtime/rpc/grpc_session_test.cc
+++ b/tensorflow/core/distributed_runtime/rpc/grpc_session_test.cc
@@ -517,7 +517,7 @@ TEST(GrpcSessionTest, Error) {
     //
     // Subgraph for "b" sleeps at the node "b_delay". When the sleep
     // finishes, the subgraph "b" will continue execution till it
-    // notices that it is cancelled. Meanwhile, subgraph's executor
+    // notices that it is canceled. Meanwhile, subgraph's executor
     // and its related state (registered ops) should still be alive.
     auto b = test::graph::Constant(&g, Tensor());
     b->set_assigned_device_name(dev_b);
diff --git a/tensorflow/core/distributed_runtime/worker_cache_logger.cc b/tensorflow/core/distributed_runtime/worker_cache_logger.cc
index 6d68c82fd19..9e16ffa9574 100644
--- a/tensorflow/core/distributed_runtime/worker_cache_logger.cc
+++ b/tensorflow/core/distributed_runtime/worker_cache_logger.cc
@@ -35,7 +35,7 @@ void WorkerCacheLogger::SetLogging(bool v) {
     ++want_logging_count_;
   } else {
     --want_logging_count_;
-    // If RPCs get cancelled, it may be possible for the count
+    // If RPCs get canceled, it may be possible for the count
     // to go negative.  This should not be a fatal error, since
     // logging is non-critical.
     if (want_logging_count_ < 0) want_logging_count_ = 0;
diff --git a/tensorflow/core/framework/cancellation.h b/tensorflow/core/framework/cancellation.h
index 4cc3f923531..651c054fe8b 100644
--- a/tensorflow/core/framework/cancellation.h
+++ b/tensorflow/core/framework/cancellation.h
@@ -36,7 +36,7 @@ namespace tensorflow {
 // CancellationManager::get_cancellation_token.
 typedef int64 CancellationToken;
 
-// A callback that is invoked when a step is cancelled.
+// A callback that is invoked when a step is canceled.
 //
 // NOTE(mrry): See caveats about CancelCallback implementations in the
 // comment for CancellationManager::RegisterCallback.
diff --git a/tensorflow/core/framework/function_test.cc b/tensorflow/core/framework/function_test.cc
index c83ecf4e5e8..f3ad935c787 100644
--- a/tensorflow/core/framework/function_test.cc
+++ b/tensorflow/core/framework/function_test.cc
@@ -163,7 +163,7 @@ REGISTER_OP("HasDefaultType")
 
 // This verifies that a function using an op before a type attr (with
 // a default) is added, still works.  This is important for backwards
-// compatibilty.
+// compatibility.
 TEST(TFunc, MissingTypeAttr) {
   auto fdef = FDH::Create(
       // Name
@@ -1021,7 +1021,7 @@ TEST(FunctionLibraryDefinitionTest, AddLibrary) {
   EXPECT_EQ(s.error_message(),
             "Gradient for function 'XTimesTwo' already exists.");
 
-  // No conflicing functions or gradients OK
+  // No conflicting functions or gradients OK
   proto.Clear();
   *proto.add_function() = test::function::XTimesFour();
   grad.set_function_name(test::function::XTimes16().signature().name());
diff --git a/tensorflow/core/graph/graph_constructor.h b/tensorflow/core/graph/graph_constructor.h
index 9b80f211fc6..54d38cac65c 100644
--- a/tensorflow/core/graph/graph_constructor.h
+++ b/tensorflow/core/graph/graph_constructor.h
@@ -51,7 +51,7 @@ extern Status ConvertGraphDefToGraph(const GraphConstructorOptions& opts,
 // On error, returns non-OK and leaves *g unmodified.
 //
 // "shape_refiner" can be null. It should be non-null if the caller
-// intends to add additonal nodes to the graph after the import. This
+// intends to add additional nodes to the graph after the import. This
 // allows the caller to validate shapes of those nodes (since
 // ShapeRefiner::AddNode must be called in topological order).
 //
diff --git a/tensorflow/core/grappler/costs/analytical_cost_estimator.h b/tensorflow/core/grappler/costs/analytical_cost_estimator.h
index 03e7faa4ff5..f267fac73ff 100644
--- a/tensorflow/core/grappler/costs/analytical_cost_estimator.h
+++ b/tensorflow/core/grappler/costs/analytical_cost_estimator.h
@@ -40,7 +40,7 @@ class AnalyticalCostEstimator : public CostEstimator {
   explicit AnalyticalCostEstimator(Cluster* cluster, bool use_static_shapes);
   ~AnalyticalCostEstimator() override {}
 
-  // Initalizes the estimator for the specified grappler item.
+  // Initializes the estimator for the specified grappler item.
   // This implementation always returns OK.
   Status Initialize(const GrapplerItem& item) override;
 
diff --git a/tensorflow/core/grappler/costs/cost_estimator.h b/tensorflow/core/grappler/costs/cost_estimator.h
index b3fb3522a39..786840384ad 100644
--- a/tensorflow/core/grappler/costs/cost_estimator.h
+++ b/tensorflow/core/grappler/costs/cost_estimator.h
@@ -130,7 +130,7 @@ class CostEstimator {
  public:
   virtual ~CostEstimator() {}
 
-  // Initalizes the estimator for the specified grappler item.
+  // Initializes the estimator for the specified grappler item.
   // The estimator shouldn't be used if this function returns any status other
   // that OK.
   virtual Status Initialize(const GrapplerItem& item) = 0;
diff --git a/tensorflow/core/grappler/costs/measuring_cost_estimator.h b/tensorflow/core/grappler/costs/measuring_cost_estimator.h
index a84853f6c71..1b3edb4c27b 100644
--- a/tensorflow/core/grappler/costs/measuring_cost_estimator.h
+++ b/tensorflow/core/grappler/costs/measuring_cost_estimator.h
@@ -50,7 +50,7 @@ class MeasuringCostEstimator : public CostEstimator {
                                   int measurement_threads);
   ~MeasuringCostEstimator() override {}
 
-  // Initalizes the estimator for the specified grappler item.
+  // Initializes the estimator for the specified grappler item.
   // This implementation always returns OK.
   Status Initialize(const GrapplerItem& item) override;
 
diff --git a/tensorflow/core/grappler/costs/op_level_cost_estimator.h b/tensorflow/core/grappler/costs/op_level_cost_estimator.h
index 266b6339225..ea7d3d3f69b 100644
--- a/tensorflow/core/grappler/costs/op_level_cost_estimator.h
+++ b/tensorflow/core/grappler/costs/op_level_cost_estimator.h
@@ -36,7 +36,7 @@ class OpLevelCostEstimator {
 
  protected:
   // Returns an estimate of device performance (in billions of operations
-  // executed per second) and memory bandwith (in GigaBytes/second) for the
+  // executed per second) and memory bandwidth (in GigaBytes/second) for the
   // specified device.
   virtual std::pair<double, double> GetDeviceInfo(
       const DeviceProperties& device) const;
diff --git a/tensorflow/core/grappler/optimizers/model_pruner.cc b/tensorflow/core/grappler/optimizers/model_pruner.cc
index 47072665728..efa21638369 100644
--- a/tensorflow/core/grappler/optimizers/model_pruner.cc
+++ b/tensorflow/core/grappler/optimizers/model_pruner.cc
@@ -46,7 +46,7 @@ Status ModelPruner::Optimize(Cluster* cluster, const GrapplerItem& item,
     if (nodes_to_preserve.find(node.name()) != nodes_to_preserve.end()) {
       continue;
     }
-    // Don't remove nodes that are explicitely placed.
+    // Don't remove nodes that are explicitly placed.
     if (!node.device().empty()) {
       continue;
     }
diff --git a/tensorflow/core/grappler/optimizers/model_pruner.h b/tensorflow/core/grappler/optimizers/model_pruner.h
index 3956d339613..3d76aebef43 100644
--- a/tensorflow/core/grappler/optimizers/model_pruner.h
+++ b/tensorflow/core/grappler/optimizers/model_pruner.h
@@ -22,7 +22,7 @@ namespace tensorflow {
 namespace grappler {
 
 // Prune a model to make it more efficient:
-// * Remove unecessary operations.
+// * Remove unnecessary operations.
 // * Optimize gradient computations.
 class ModelPruner : public GraphOptimizer {
  public:
diff --git a/tensorflow/core/grappler/utils.h b/tensorflow/core/grappler/utils.h
index 17b980c5b8c..0fb531ef1bd 100644
--- a/tensorflow/core/grappler/utils.h
+++ b/tensorflow/core/grappler/utils.h
@@ -34,7 +34,7 @@ class NodeMap {
   NodeDef* GetNode(const string& name);
   std::set<NodeDef*> GetOutputs(const string& node_name);
   // This method doesn't record the outputs of the added node; the outputs need
-  // to be explictly added by the AddOutput method.
+  // to be explicitly added by the AddOutput method.
   void AddNode(const string& name, NodeDef* node);
   void AddOutput(const string& node, const string& output);
   void UpdateOutput(const string& node, const string& old_output,
diff --git a/tensorflow/core/kernels/cast_op.h b/tensorflow/core/kernels/cast_op.h
index 0def600ac0c..5c24f164a41 100644
--- a/tensorflow/core/kernels/cast_op.h
+++ b/tensorflow/core/kernels/cast_op.h
@@ -50,7 +50,7 @@ template <typename From, typename To>
 struct scalar_cast_op<std::complex<From>, To> {
   EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE To
   operator()(const std::complex<From>& a) const {
-    // Replicate numpy behaviour of returning just the real part
+    // Replicate numpy behavior of returning just the real part
     return static_cast<To>(a.real());
   }
 };
@@ -59,7 +59,7 @@ template <typename From, typename To>
 struct scalar_cast_op<From, std::complex<To>> {
   EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE std::complex<To> operator()(
       const From& a) const {
-    // Replicate numpy behaviour of setting the imaginary part to 0
+    // Replicate numpy behavior of setting the imaginary part to 0
     return std::complex<To>(static_cast<To>(a), To(0));
   }
 };
diff --git a/tensorflow/core/kernels/conv_ops_fused.cc b/tensorflow/core/kernels/conv_ops_fused.cc
index f7348f10772..291ebf22987 100644
--- a/tensorflow/core/kernels/conv_ops_fused.cc
+++ b/tensorflow/core/kernels/conv_ops_fused.cc
@@ -713,7 +713,7 @@ class FusedResizeConv2DUsingGemmOp : public OpKernel {
       const int32 before =
           paddings_matrix(d, 0);  // Pad before existing elements.
       const int32 after =
-          paddings_matrix(d, 1);  // Pad after exisitng elements.
+          paddings_matrix(d, 1);  // Pad after existing elements.
       OP_REQUIRES(context, before >= 0 && after >= 0,
                   errors::InvalidArgument("paddings must be non-negative: ",
                                           before, " ", after));
diff --git a/tensorflow/core/kernels/cuda_solvers.h b/tensorflow/core/kernels/cuda_solvers.h
index 70ccbb90ccb..5d1c807e66e 100644
--- a/tensorflow/core/kernels/cuda_solvers.h
+++ b/tensorflow/core/kernels/cuda_solvers.h
@@ -116,7 +116,7 @@ class CudaSolver {
   // Launches a memcpy of solver status data specified by dev_lapack_info from
   // device to the host, and asynchronously invokes the given callback when the
   // copy is complete. The first Status argument to the callback will be
-  // Status::OK if all lapack infos retrived are zero, otherwise an error status
+  // Status::OK if all lapack infos retrieved are zero, otherwise an error status
   // is given. The second argument contains a host-side copy of the entire set
   // of infos retrieved, and can be used for generating detailed error messages.
   Status CopyLapackInfoToHostAsync(
diff --git a/tensorflow/core/kernels/deep_conv2d.cc b/tensorflow/core/kernels/deep_conv2d.cc
index a4814014798..8e9b8a7e2e7 100644
--- a/tensorflow/core/kernels/deep_conv2d.cc
+++ b/tensorflow/core/kernels/deep_conv2d.cc
@@ -26,7 +26,7 @@ limitations under the License.
 
 namespace tensorflow {
 
-// DeepConv2D is a Conv2D implementation specialzied for deep convolutions (i.e
+// DeepConv2D is a Conv2D implementation specialized for deep convolutions (i.e
 // large 'in_depth' and 'out_depth' product. See cost models below for details).
 //
 // DeepConv2D is implemented by computing the following equation:
diff --git a/tensorflow/core/kernels/deep_conv2d.h b/tensorflow/core/kernels/deep_conv2d.h
index a9de20e7ae7..c3f6f66dc9b 100644
--- a/tensorflow/core/kernels/deep_conv2d.h
+++ b/tensorflow/core/kernels/deep_conv2d.h
@@ -22,7 +22,7 @@ namespace tensorflow {
 
 class OpKernelContext;
 
-// DeepConv2D is a Conv2D implementation specialzied for deep (i.e. large
+// DeepConv2D is a Conv2D implementation specialized for deep (i.e. large
 // in_depth * out_depth product) convolutions (see deep_conv2d.cc for details).
 
 // DeepConv2DTransform is an interface for implementing transforms for
diff --git a/tensorflow/core/kernels/hinge-loss.h b/tensorflow/core/kernels/hinge-loss.h
index 36b02fcc5d6..789a7ce7a3d 100644
--- a/tensorflow/core/kernels/hinge-loss.h
+++ b/tensorflow/core/kernels/hinge-loss.h
@@ -44,7 +44,7 @@ class HingeLossUpdater : public DualLossUpdater {
                             const double current_dual, const double wx,
                             const double weighted_example_norm) const final {
     // Intutitvely there are 3 cases:
-    // a. new optimal value of the dual variable falls withing the admissible
+    // a. new optimal value of the dual variable falls within the admissible
     // range [0, 1]. In this case we set new dual to this value.
     // b. new optimal value is < 0. Then, because of convexity, the optimal
     // valid value for new dual = 0
diff --git a/tensorflow/core/kernels/image_resizer_state.h b/tensorflow/core/kernels/image_resizer_state.h
index 9ef44a57827..f088315ff53 100644
--- a/tensorflow/core/kernels/image_resizer_state.h
+++ b/tensorflow/core/kernels/image_resizer_state.h
@@ -13,7 +13,7 @@ See the License for the specific language governing permissions and
 limitations under the License.
 ==============================================================================*/
 
-// This is a helper struct to package up the input and ouput
+// This is a helper struct to package up the input and output
 // parameters of an image resizer (the height, widths, etc.).  To
 // reduce code duplication and ensure consistency across the different
 // resizers, it performs the input validation.
diff --git a/tensorflow/core/kernels/map_stage_op.cc b/tensorflow/core/kernels/map_stage_op.cc
index 6ec7cce59c8..f69298e9524 100644
--- a/tensorflow/core/kernels/map_stage_op.cc
+++ b/tensorflow/core/kernels/map_stage_op.cc
@@ -238,7 +238,7 @@ private:
     {
       IncompleteTuple empty(dtypes_.size());
 
-      // Initialise empty tuple with given dta
+      // Initialize empty tuple with given dta
       for(std::size_t i = 0; i < findices.dimension(0); ++i)
       {
         std::size_t index = findices(i);
diff --git a/tensorflow/core/kernels/meta_support.h b/tensorflow/core/kernels/meta_support.h
index 0d87baf0344..53aece78e87 100644
--- a/tensorflow/core/kernels/meta_support.h
+++ b/tensorflow/core/kernels/meta_support.h
@@ -64,7 +64,7 @@ bool IsSupportedAndEnabled();
 //     sum((a_data[i, l] + offset_a) * (b_data[l, j] + offset_b)) : l in [0, k)
 //
 // If transpose_a is false the lhs operand has row major layout, otherwise
-// column major. Similarily transpose_b describes the layout of the rhs operand.
+// column major. Similarly transpose_b describes the layout of the rhs operand.
 // lda, ldb, and ldc are the strides of the lhs operand, rhs operand and the
 // result arrays.
 void QuantizedGemm(OpKernelContext* context, bool transpose_a, bool transpose_b,
diff --git a/tensorflow/core/kernels/mkl_conv_grad_input_ops.cc b/tensorflow/core/kernels/mkl_conv_grad_input_ops.cc
index c97f1dd7b73..23827ceea50 100644
--- a/tensorflow/core/kernels/mkl_conv_grad_input_ops.cc
+++ b/tensorflow/core/kernels/mkl_conv_grad_input_ops.cc
@@ -295,7 +295,7 @@ class MklConv2DCustomBackpropInputOp : public OpKernel {
         dnnDelete_F32(mkl_convert_filter);
       } else {
         // If we do not need any layout conversion for filter, then
-        // we direclty assign input filter to resources[].
+        // we directly assign input filter to resources[].
         conv_res[dnnResourceFilter] =
             static_cast<void*>(const_cast<T*>(filter.flat<T>().data()));
       }
diff --git a/tensorflow/core/kernels/padded_batch_dataset_op.cc b/tensorflow/core/kernels/padded_batch_dataset_op.cc
index df78e1f0641..b0c000dd25f 100644
--- a/tensorflow/core/kernels/padded_batch_dataset_op.cc
+++ b/tensorflow/core/kernels/padded_batch_dataset_op.cc
@@ -306,7 +306,7 @@ class PaddedBatchDatasetOp : public OpKernel {
             const TensorShape& element_shape =
                 batch_elements[i][component_index].shape();
             // TODO(mrry): Perform this check in the shape function if
-            // enough static information is avaiable to do so.
+            // enough static information is available to do so.
             if (element_shape.dims() != padded_shape.dims()) {
               return errors::InvalidArgument(
                   "All elements in a batch must have the same rank as the "
diff --git a/tensorflow/core/kernels/quantization_utils.h b/tensorflow/core/kernels/quantization_utils.h
index be67dfd112b..dae40c4678b 100644
--- a/tensorflow/core/kernels/quantization_utils.h
+++ b/tensorflow/core/kernels/quantization_utils.h
@@ -80,7 +80,7 @@ float QuantizedToFloat(T input, float range_min, float range_max) {
       static_cast<int64>(Eigen::NumTraits<T>::lowest());
   const double offset_input = static_cast<double>(input) - lowest_quantized;
   // For compatibility with DEQUANTIZE_WITH_EIGEN, we should convert
-  // range_scale to a float, otherwise range_min_rounded might be slighly
+  // range_scale to a float, otherwise range_min_rounded might be slightly
   // different.
   const double range_min_rounded =
       round(range_min / static_cast<float>(range_scale)) *
diff --git a/tensorflow/core/kernels/smooth-hinge-loss.h b/tensorflow/core/kernels/smooth-hinge-loss.h
index 45da0fb1172..5074ad0795d 100644
--- a/tensorflow/core/kernels/smooth-hinge-loss.h
+++ b/tensorflow/core/kernels/smooth-hinge-loss.h
@@ -35,7 +35,7 @@ class SmoothHingeLossUpdater : public DualLossUpdater {
                             const double current_dual, const double wx,
                             const double weighted_example_norm) const final {
     // Intutitvely there are 3 cases:
-    // a. new optimal value of the dual variable falls withing the admissible
+    // a. new optimal value of the dual variable falls within the admissible
     // range [0, 1]. In this case we set new dual to this value.
     // b. new optimal value is < 0. Then, because of convexity, the optimal
     // valid value for new dual = 0
diff --git a/tensorflow/core/kernels/sparse_tensor_dense_add_op.h b/tensorflow/core/kernels/sparse_tensor_dense_add_op.h
index b06dcf143ec..353cf0e5190 100644
--- a/tensorflow/core/kernels/sparse_tensor_dense_add_op.h
+++ b/tensorflow/core/kernels/sparse_tensor_dense_add_op.h
@@ -24,7 +24,7 @@ limitations under the License.
 namespace tensorflow {
 namespace functor {
 
-// TOOD(zongheng): this should be a general functor that powers SparseAdd and
+// TODO(zongheng): this should be a general functor that powers SparseAdd and
 // ScatterNd ops.  It should be moved to its own head file, once the other ops
 // are implemented.
 template <typename Device, typename T, typename Index, int NDIMS,
diff --git a/tensorflow/core/lib/gtl/optional.h b/tensorflow/core/lib/gtl/optional.h
index f80b5c113d0..8ba4b091434 100644
--- a/tensorflow/core/lib/gtl/optional.h
+++ b/tensorflow/core/lib/gtl/optional.h
@@ -541,7 +541,7 @@ class optional : private internal_optional::optional_data<T>,
   // opt.emplace(arg1,arg2,arg3);  (Constructs Foo(arg1,arg2,arg3))
   //
   // If the optional is non-empty, and the `args` refer to subobjects of the
-  // current object, then behaviour is undefined.  This is because the current
+  // current object, then behavior is undefined.  This is because the current
   // object will be destructed before the new object is constructed with `args`.
   //
   template <typename... Args,
@@ -586,7 +586,7 @@ class optional : private internal_optional::optional_data<T>,
 
   // [optional.observe], observers
   // You may use `*opt`, and `opt->m`, to access the underlying T value and T's
-  // member `m`, respectively.  If the optional is empty, behaviour is
+  // member `m`, respectively.  If the optional is empty, behavior is
   // undefined.
   constexpr const T* operator->() const { return this->pointer(); }
   T* operator->() {
diff --git a/tensorflow/core/lib/jpeg/jpeg_mem.cc b/tensorflow/core/lib/jpeg/jpeg_mem.cc
index e27904ea12a..5dce3673fc0 100644
--- a/tensorflow/core/lib/jpeg/jpeg_mem.cc
+++ b/tensorflow/core/lib/jpeg/jpeg_mem.cc
@@ -45,7 +45,7 @@ enum JPEGErrors {
   JPEGERRORS_BAD_PARAM
 };
 
-// Prevent bad compiler behaviour in ASAN mode by wrapping most of the
+// Prevent bad compiler behavior in ASAN mode by wrapping most of the
 // arguments in a struct struct.
 class FewerArgsForCompiler {
  public:
diff --git a/tensorflow/core/ops/data_flow_ops.cc b/tensorflow/core/ops/data_flow_ops.cc
index 2be804acbf6..282778e495d 100644
--- a/tensorflow/core/ops/data_flow_ops.cc
+++ b/tensorflow/core/ops/data_flow_ops.cc
@@ -827,7 +827,7 @@ operations that would block will fail immediately.
 
 handle: The handle to a queue.
 cancel_pending_enqueues: If true, all pending enqueue requests that are
-  blocked on the given queue will be cancelled.
+  blocked on the given queue will be canceled.
 )doc");
 
 REGISTER_OP("QueueCloseV2")
@@ -845,7 +845,7 @@ operations that would block will fail immediately.
 
 handle: The handle to a queue.
 cancel_pending_enqueues: If true, all pending enqueue requests that are
-  blocked on the given queue will be cancelled.
+  blocked on the given queue will be canceled.
 )doc");
 
 REGISTER_OP("QueueSize")
@@ -1879,7 +1879,7 @@ Subsequent TakeMany operations that would block will fail immediately.
 
 handle: The handle to a barrier.
 cancel_pending_enqueues: If true, all pending enqueue requests that are
-  blocked on the barrier's queue will be cancelled. InsertMany will fail, even
+  blocked on the barrier's queue will be canceled. InsertMany will fail, even
   if no new key is introduced.
 )doc");
 
diff --git a/tensorflow/core/ops/ops.pbtxt b/tensorflow/core/ops/ops.pbtxt
index d3ee4d74fc9..c7a30d5ae26 100644
--- a/tensorflow/core/ops/ops.pbtxt
+++ b/tensorflow/core/ops/ops.pbtxt
@@ -2380,7 +2380,7 @@ op {
     default_value {
       b: false
     }
-    description: "If true, all pending enqueue requests that are\nblocked on the barrier\'s queue will be cancelled. InsertMany will fail, even\nif no new key is introduced."
+    description: "If true, all pending enqueue requests that are\nblocked on the barrier\'s queue will be canceled. InsertMany will fail, even\nif no new key is introduced."
   }
   summary: "Closes the given barrier."
   description: "This operation signals that no more new elements will be inserted in the\ngiven barrier. Subsequent InsertMany that try to introduce a new key will fail.\nSubsequent InsertMany operations that just add missing components to already\nexisting elements will continue to succeed. Subsequent TakeMany operations will\ncontinue to succeed if sufficient completed elements remain in the barrier.\nSubsequent TakeMany operations that would block will fail immediately."
@@ -15613,7 +15613,7 @@ op {
     default_value {
       b: false
     }
-    description: "If true, all pending enqueue requests that are\nblocked on the given queue will be cancelled."
+    description: "If true, all pending enqueue requests that are\nblocked on the given queue will be canceled."
   }
   summary: "Closes the given queue."
   description: "This operation signals that no more elements will be enqueued in the\ngiven queue. Subsequent Enqueue(Many) operations will fail.\nSubsequent Dequeue(Many) operations will continue to succeed if\nsufficient elements remain in the queue. Subsequent Dequeue(Many)\noperations that would block will fail immediately."
@@ -15631,7 +15631,7 @@ op {
     default_value {
       b: false
     }
-    description: "If true, all pending enqueue requests that are\nblocked on the given queue will be cancelled."
+    description: "If true, all pending enqueue requests that are\nblocked on the given queue will be canceled."
   }
   summary: "Closes the given queue."
   description: "This operation signals that no more elements will be enqueued in the\ngiven queue. Subsequent Enqueue(Many) operations will fail.\nSubsequent Dequeue(Many) operations will continue to succeed if\nsufficient elements remain in the queue. Subsequent Dequeue(Many)\noperations that would block will fail immediately."
diff --git a/tensorflow/core/platform/posix/subprocess.cc b/tensorflow/core/platform/posix/subprocess.cc
index fc511fdf727..cefc66831a9 100644
--- a/tensorflow/core/platform/posix/subprocess.cc
+++ b/tensorflow/core/platform/posix/subprocess.cc
@@ -28,7 +28,7 @@ limitations under the License.
 // A danger of calling fork() (as opposed to clone() or vfork()) is that if
 // many people have used pthread_atfork() to acquire locks, fork() can deadlock,
 // because it's unlikely that the locking order will be correct in a large
-// programme where different layers are unaware of one another and using
+// program where different layers are unaware of one another and using
 // pthread_atfork() independently.
 //
 // The danger of not calling fork() is that if libc managed to use
diff --git a/tensorflow/core/protobuf/master.proto b/tensorflow/core/protobuf/master.proto
index e607b1c42a5..0a825bbb928 100644
--- a/tensorflow/core/protobuf/master.proto
+++ b/tensorflow/core/protobuf/master.proto
@@ -202,7 +202,7 @@ message CloseSessionResponse {
 // Old sessions may continue to have side-effects on resources not in
 // containers listed in "containers", and thus may affect future
 // sessions' results in ways that are hard to predict.  Thus, if well-defined
-// behaviour is desired, is it recommended that all containers be listed in
+// behavior is desired, is it recommended that all containers be listed in
 // "containers".  Similarly, if a device_filter is specified, results may be
 // hard to predict.
 message ResetRequest {
diff --git a/tensorflow/core/public/session.h b/tensorflow/core/public/session.h
index eaa076ffb91..4792b32a529 100644
--- a/tensorflow/core/public/session.h
+++ b/tensorflow/core/public/session.h
@@ -199,7 +199,7 @@ Status NewSession(const SessionOptions& options, Session** out_session);
 /// Old sessions may continue to have side-effects on resources not in
 /// containers listed in "containers", and thus may affect future
 /// sessions' results in ways that are hard to predict.  Thus, if well-defined
-/// behaviour is desired, it is recommended that all containers be listed in
+/// behavior is desired, it is recommended that all containers be listed in
 /// "containers".
 ///
 /// `containers` is a vector of string representation of resource container
diff --git a/tensorflow/core/util/ctc/ctc_loss_calculator.h b/tensorflow/core/util/ctc/ctc_loss_calculator.h
index 567bad38c33..f181ab93e77 100644
--- a/tensorflow/core/util/ctc/ctc_loss_calculator.h
+++ b/tensorflow/core/util/ctc/ctc_loss_calculator.h
@@ -48,7 +48,7 @@ class CTCLossCalculator {
   // these examples.
   //
   // Reference materials:
-  //  GravesTh: Alex Graves, "Supervised Sequence Labelling with Recurrent
+  //  GravesTh: Alex Graves, "Supervised Sequence Labeling with Recurrent
   //    Neural Networks" (PhD Thesis), Technische Universit¨at M¨unchen.
  public:
   typedef std::vector<std::vector<int>> LabelSequences;
diff --git a/tensorflow/core/util/example_proto_fast_parsing.h b/tensorflow/core/util/example_proto_fast_parsing.h
index 5f8b4af5fe2..20536cee163 100644
--- a/tensorflow/core/util/example_proto_fast_parsing.h
+++ b/tensorflow/core/util/example_proto_fast_parsing.h
@@ -45,7 +45,7 @@ struct FastParseExampleConfig {
     DataType dtype;
     // These 2 fields correspond exactly to dense_shapes and dense_defaults in
     // ParseExample op.
-    // Documentation is avaliable in: tensorflow/core/ops/parsing_ops.cc
+    // Documentation is available in: tensorflow/core/ops/parsing_ops.cc
     PartialTensorShape shape;
     Tensor default_value;
     bool variable_length;
@@ -62,7 +62,7 @@ struct FastParseExampleConfig {
 };
 
 // This is exactly the output of TF's ParseExample Op.
-// Documentation is avaliable in: tensorflow/core/ops/parsing_ops.cc
+// Documentation is available in: tensorflow/core/ops/parsing_ops.cc
 struct Result {
   std::vector<Tensor> sparse_indices;
   std::vector<Tensor> sparse_values;
diff --git a/tensorflow/core/util/tensor_bundle/tensor_bundle.h b/tensorflow/core/util/tensor_bundle/tensor_bundle.h
index b0bddf7e423..2c40388250c 100644
--- a/tensorflow/core/util/tensor_bundle/tensor_bundle.h
+++ b/tensorflow/core/util/tensor_bundle/tensor_bundle.h
@@ -31,7 +31,7 @@ limitations under the License.
 // (tensorflow::table::Table).  Each key is a name of a tensor and its value is
 // a serialized BundleEntryProto.  Each BundleEntryProto describes the metadata
 // of a tensor: which of the "data" files contains the content of a tensor, the
-// offset into that file, checksum, some auxilary data, etc.
+// offset into that file, checksum, some auxiliary data, etc.
 //
 // A tensor bundle can be accessed randomly using a BundleReader.  Usage:
 //
diff --git a/tensorflow/docs_src/performance/xla/index.md b/tensorflow/docs_src/performance/xla/index.md
index d2c18433279..19045b45d92 100644
--- a/tensorflow/docs_src/performance/xla/index.md
+++ b/tensorflow/docs_src/performance/xla/index.md
@@ -65,13 +65,13 @@ The following diagram shows the compilation process in XLA:
   <img src="https://www.tensorflow.org/images/how-does-xla-work.png">
 </div>
 
-XLA comes with several optimizations and analyses that are target-independent,
+XLA comes with several optimizations and analyzes that are target-independent,
 such as [CSE](https://en.wikipedia.org/wiki/Common_subexpression_elimination),
 target-independent operation fusion, and buffer analysis for allocating runtime
 memory for the computation.
 
 After the target-independent step, XLA sends the HLO computation to a backend.
-The backend can perform further HLO-level analyses and optimizations, this time
+The backend can perform further HLO-level analyzes and optimizations, this time
 with target specific information and needs in mind. For example, the XLA GPU
 backend may perform operation fusion beneficial specifically for the GPU
 programming model and determine how to partition the computation into streams.
diff --git a/tensorflow/docs_src/programmers_guide/reading_data.md b/tensorflow/docs_src/programmers_guide/reading_data.md
index 088724337e4..3c31d3a1a70 100644
--- a/tensorflow/docs_src/programmers_guide/reading_data.md
+++ b/tensorflow/docs_src/programmers_guide/reading_data.md
@@ -332,7 +332,7 @@ limit has been reached and no more examples are available.
 
 The last ingredient is the
 @{tf.train.Coordinator}. This is responsible
-for letting all the threads know if anything has signalled a shut down. Most
+for letting all the threads know if anything has signaled a shut down. Most
 commonly this would be because an exception was raised, for example one of the
 threads got an error when running some operation (or an ordinary Python
 exception).
diff --git a/tensorflow/docs_src/programmers_guide/version_semantics.md b/tensorflow/docs_src/programmers_guide/version_semantics.md
index 47fc582387a..cee3b105de4 100644
--- a/tensorflow/docs_src/programmers_guide/version_semantics.md
+++ b/tensorflow/docs_src/programmers_guide/version_semantics.md
@@ -118,7 +118,7 @@ Many users of TensorFlow will be saving graphs and trained models to disk for
 later evaluation or more training, often changing versions of TensorFlow in the
 process.  First, following semver, any graph or checkpoint written out with one
 version of TensorFlow can be loaded and evaluated with a later version of
-TensorFlow with the same major release.  However, we will endeavour to preserve
+TensorFlow with the same major release.  However, we will endeavor to preserve
 backwards compatibility even across major releases when possible, so that the
 serialized files are usable over long periods of time.
 
diff --git a/tensorflow/examples/android/jni/object_tracking/image_utils.h b/tensorflow/examples/android/jni/object_tracking/image_utils.h
index 2d712e77f91..ac9ffd90f8a 100644
--- a/tensorflow/examples/android/jni/object_tracking/image_utils.h
+++ b/tensorflow/examples/android/jni/object_tracking/image_utils.h
@@ -67,7 +67,7 @@ inline static void MarkImage(const int x, const int y, const int radius,
     // reduce the number of iterations required as compared to starting from
     // either 0 and counting up or radius and counting down.
     for (int d_x = radius - d_y; d_x <= radius; ++d_x) {
-      // The first time this critera is met, we know the width of the circle at
+      // The first time this criteria is met, we know the width of the circle at
       // this row (without using sqrt).
       if (squared_y_dist + Square(d_x) >= squared_radius) {
         const int min_x = MAX(x - d_x, 0);
diff --git a/tensorflow/examples/udacity/1_notmnist.ipynb b/tensorflow/examples/udacity/1_notmnist.ipynb
index 5b2966c0062..39674e1aa49 100644
--- a/tensorflow/examples/udacity/1_notmnist.ipynb
+++ b/tensorflow/examples/udacity/1_notmnist.ipynb
@@ -70,7 +70,7 @@
         "colab_type": "text"
       },
       "source": [
-        "First, we'll download the dataset to our local machine. The data consists of characters rendered in a variety of fonts on a 28x28 image. The labels are limited to 'A' through 'J' (10 classes). The training set has about 500k and the testset 19000 labelled examples. Given these sizes, it should be possible to train models quickly on any machine."
+        "First, we'll download the dataset to our local machine. The data consists of characters rendered in a variety of fonts on a 28x28 image. The labels are limited to 'A' through 'J' (10 classes). The training set has about 500k and the testset 19000 labeled examples. Given these sizes, it should be possible to train models quickly on any machine."
       ]
     },
     {
@@ -168,7 +168,7 @@
       },
       "source": [
         "Extract the dataset from the compressed .tar.gz file.\n",
-        "This should give you a set of directories, labelled A through J."
+        "This should give you a set of directories, labeled A through J."
       ]
     },
     {
diff --git a/tensorflow/go/op/wrappers.go b/tensorflow/go/op/wrappers.go
index 4af1386306d..f508b63b138 100644
--- a/tensorflow/go/op/wrappers.go
+++ b/tensorflow/go/op/wrappers.go
@@ -4813,7 +4813,7 @@ type QueueCloseV2Attr func(optionalAttr)
 // QueueCloseV2CancelPendingEnqueues sets the optional cancel_pending_enqueues attribute to value.
 //
 // value: If true, all pending enqueue requests that are
-// blocked on the given queue will be cancelled.
+// blocked on the given queue will be canceled.
 // If not specified, defaults to false
 func QueueCloseV2CancelPendingEnqueues(value bool) QueueCloseV2Attr {
 	return func(m optionalAttr) {
diff --git a/tensorflow/java/src/test/java/org/tensorflow/TensorFlowTest.java b/tensorflow/java/src/test/java/org/tensorflow/TensorFlowTest.java
index 27e2215f62b..a31ea900d1c 100644
--- a/tensorflow/java/src/test/java/org/tensorflow/TensorFlowTest.java
+++ b/tensorflow/java/src/test/java/org/tensorflow/TensorFlowTest.java
@@ -33,7 +33,7 @@ public class TensorFlowTest {
   public void registeredOpList() {
     // Would be nice to actually parse the output as a tensorflow.OpList protocol buffer message,
     // but as of May 2017, bazel support for generating Java code from protocol buffer definitions
-    // was not sorted out. Revisit? Till then, at least excercise the code.
+    // was not sorted out. Revisit? Till then, at least exercise the code.
     assertTrue(TensorFlow.registeredOpList().length > 0);
   }
 }
diff --git a/tensorflow/python/debug/cli/analyzer_cli.py b/tensorflow/python/debug/cli/analyzer_cli.py
index 69b6d9ffdf0..da27f4cebea 100644
--- a/tensorflow/python/debug/cli/analyzer_cli.py
+++ b/tensorflow/python/debug/cli/analyzer_cli.py
@@ -368,7 +368,7 @@ class DebugAnalyzer(object):
   def add_tensor_filter(self, filter_name, filter_callable):
     """Add a tensor filter.
 
-    A tensor filter is a named callable of the siganture:
+    A tensor filter is a named callable of the signature:
       filter_callable(dump_datum, tensor),
 
     wherein dump_datum is an instance of debug_data.DebugTensorDatum carrying
diff --git a/tensorflow/python/debug/cli/debugger_cli_common.py b/tensorflow/python/debug/cli/debugger_cli_common.py
index 9ad49771d1d..12e79ab07a4 100644
--- a/tensorflow/python/debug/cli/debugger_cli_common.py
+++ b/tensorflow/python/debug/cli/debugger_cli_common.py
@@ -840,7 +840,7 @@ class TabCompletionRegistry(object):
 
     Args:
       context_words: A list of context words belonging to the context being
-        registerd. It is a list of str, instead of a single string, to support
+        registered. It is a list of str, instead of a single string, to support
         synonym words triggering the same tab-completion context, e.g.,
         both "drink" and the short-hand "dr" can trigger the same context.
       comp_items: A list of completion items, as a list of str.
diff --git a/tensorflow/python/debug/cli/profile_analyzer_cli.py b/tensorflow/python/debug/cli/profile_analyzer_cli.py
index c08605b92b0..3304194b1cb 100644
--- a/tensorflow/python/debug/cli/profile_analyzer_cli.py
+++ b/tensorflow/python/debug/cli/profile_analyzer_cli.py
@@ -330,7 +330,7 @@ class ProfileAnalyzer(object):
     self._arg_parsers["list_profile"] = ap
 
     ap = argparse.ArgumentParser(
-        description="Print a Python source file wiht line-level profile "
+        description="Print a Python source file with line-level profile "
                     "information",
         usage=argparse.SUPPRESS)
     ap.add_argument(
diff --git a/tensorflow/python/debug/lib/debug_utils.py b/tensorflow/python/debug/lib/debug_utils.py
index 9013cb096d9..f1e972940b7 100644
--- a/tensorflow/python/debug/lib/debug_utils.py
+++ b/tensorflow/python/debug/lib/debug_utils.py
@@ -121,7 +121,7 @@ def watch_graph(run_options,
       are set, the two filtering operations will occur in a logical `AND`
       relation. In other words, a node will be included if and only if it
       hits both whitelists.
-    tensor_dtype_regex_whitelist: Regular-experssion whitelist for Tensor
+    tensor_dtype_regex_whitelist: Regular-expression whitelist for Tensor
       data type, e.g., `"^int.*"`.
       This whitelist operates in logical `AND` relations to the two whitelists
       above.
@@ -210,7 +210,7 @@ def watch_graph_with_blacklists(run_options,
       relation. In other words, a node will be excluded if it hits either of
       the two blacklists; a node will be included if and only if it hits
       neither of the blacklists.
-    tensor_dtype_regex_blacklist: Regular-experssion blacklist for Tensor
+    tensor_dtype_regex_blacklist: Regular-expression blacklist for Tensor
       data type, e.g., `"^int.*"`.
       This blacklist operates in logical `OR` relations to the two whitelists
       above.
diff --git a/tensorflow/python/debug/wrappers/dumping_wrapper.py b/tensorflow/python/debug/wrappers/dumping_wrapper.py
index 0d9b3cfa7eb..63229a85398 100644
--- a/tensorflow/python/debug/wrappers/dumping_wrapper.py
+++ b/tensorflow/python/debug/wrappers/dumping_wrapper.py
@@ -86,7 +86,7 @@ class DumpingDebugWrapperSession(framework.NonInteractiveDebugWrapperSession):
     """Implementation of abstrat method in superclass.
 
     See doc of `NonInteractiveDebugWrapperSession.prepare_run_debug_urls()`
-    for details. This implentation creates a run-specific subdirectory under
+    for details. This implementation creates a run-specific subdirectory under
     self._session_root and stores information regarding run `fetches` and
     `feed_dict.keys()` in the subdirectory.
 
diff --git a/tensorflow/python/debug/wrappers/framework.py b/tensorflow/python/debug/wrappers/framework.py
index ea642adbd1d..2c239038e44 100644
--- a/tensorflow/python/debug/wrappers/framework.py
+++ b/tensorflow/python/debug/wrappers/framework.py
@@ -666,7 +666,7 @@ class WatchOptions(object):
         are set, the two filtering operations will occur in a logical `AND`
         relation. In other words, a node will be included if and only if it
         hits both whitelists.
-      tensor_dtype_regex_whitelist: Regular-experssion whitelist for Tensor
+      tensor_dtype_regex_whitelist: Regular-expression whitelist for Tensor
         data type, e.g., `"^int.*"`.
         This whitelist operates in logical `AND` relations to the two whitelists
         above.
diff --git a/tensorflow/python/estimator/estimator.py b/tensorflow/python/estimator/estimator.py
index 33ffe3d81ed..230f103c5d1 100644
--- a/tensorflow/python/estimator/estimator.py
+++ b/tensorflow/python/estimator/estimator.py
@@ -200,7 +200,7 @@ class Estimator(object):
         error. 'steps' works incrementally. If you call two times
         train(steps=10) then training occurs in total 20 steps. If `OutOfRange`
         or `StopIteration` error occurs in the middle, training stops before 20
-        steps. If you don't want to have incremental behaviour please set
+        steps. If you don't want to have incremental behavior please set
         `max_steps` instead. If set, `max_steps` must be `None`.
       max_steps: Number of total steps for which to train model. If `None`,
         train forever or train until input_fn generates the `OutOfRange` or
diff --git a/tensorflow/python/framework/function.py b/tensorflow/python/framework/function.py
index ac8aee2c83d..08faa3a6d2e 100644
--- a/tensorflow/python/framework/function.py
+++ b/tensorflow/python/framework/function.py
@@ -281,7 +281,7 @@ class _FuncGraph(ops.Graph):
   _FuncGraph overrides ops.Graph's create_op() so that we can keep
   track of every inputs into every op created inside the function.  If
   any input is from other graphs, we keep track of it in self.capture
-  and substitue the input with a place holder.
+  and substitute the input with a place holder.
 
   Each captured input's corresponding place holder is converted into a
   function argument and the caller passes in the captured tensor.
diff --git a/tensorflow/python/kernel_tests/barrier_ops_test.py b/tensorflow/python/kernel_tests/barrier_ops_test.py
index e90543a44b0..7f49c639577 100644
--- a/tensorflow/python/kernel_tests/barrier_ops_test.py
+++ b/tensorflow/python/kernel_tests/barrier_ops_test.py
@@ -402,7 +402,7 @@ class BarrierTest(test.TestCase):
       with self.assertRaisesOpError("is closed"):
         fail_insert_op.run()
 
-      # This op should succeed because the barrier has not cancelled
+      # This op should succeed because the barrier has not canceled
       # pending enqueues
       insert_1_op.run()
       self.assertEquals(size_t.eval(), [3])
@@ -461,7 +461,7 @@ class BarrierTest(test.TestCase):
       with self.assertRaisesOpError("is closed"):
         fail_insert_op.run()
 
-      # This op should fail because the queue is cancelled.
+      # This op should fail because the queue is canceled.
       with self.assertRaisesOpError("is closed"):
         insert_2_op.run()
 
diff --git a/tensorflow/python/kernel_tests/conv_ops_3d_test.py b/tensorflow/python/kernel_tests/conv_ops_3d_test.py
index 04c43ef5fa4..14622ab4678 100644
--- a/tensorflow/python/kernel_tests/conv_ops_3d_test.py
+++ b/tensorflow/python/kernel_tests/conv_ops_3d_test.py
@@ -330,7 +330,7 @@ class Conv3DTest(test.TestCase):
 
     if test.is_gpu_available() and use_gpu:
       data_type = dtypes.float32
-      # TOOD(mjanusz): Modify gradient_checker to also provide max relative
+      # TODO(mjanusz): Modify gradient_checker to also provide max relative
       # error and synchronize the tolerance levels between the tests for forward
       # and backward computations.
       if test.is_gpu_available():
diff --git a/tensorflow/python/kernel_tests/distributions/categorical_test.py b/tensorflow/python/kernel_tests/distributions/categorical_test.py
index 396de45cad3..33db933e82a 100644
--- a/tensorflow/python/kernel_tests/distributions/categorical_test.py
+++ b/tensorflow/python/kernel_tests/distributions/categorical_test.py
@@ -127,7 +127,7 @@ class CategoricalTest(test.TestCase):
       self.assertAllClose(dist.prob(0).eval(), 0.2)
 
   def testCDFWithDynamicEventShape(self):
-    """Test that dynamically-sized events with unkown shape work."""
+    """Test that dynamically-sized events with unknown shape work."""
     batch_size = 2
     histograms = array_ops.placeholder(dtype=dtypes.float32,
                                        shape=(batch_size, None))
diff --git a/tensorflow/python/kernel_tests/metrics_test.py b/tensorflow/python/kernel_tests/metrics_test.py
index cd5bee362d7..543039bdd3d 100644
--- a/tensorflow/python/kernel_tests/metrics_test.py
+++ b/tensorflow/python/kernel_tests/metrics_test.py
@@ -1169,7 +1169,7 @@ class AUCTest(test.TestCase):
       self.assertAlmostEqual(1, auc.eval(), 6)
 
   def np_auc(self, predictions, labels, weights):
-    """Computes the AUC explicitely using Numpy.
+    """Computes the AUC explicitly using Numpy.
 
     Args:
       predictions: an ndarray with shape [N].
diff --git a/tensorflow/python/kernel_tests/substr_op_test.py b/tensorflow/python/kernel_tests/substr_op_test.py
index 0c0710fed43..854394b0dde 100644
--- a/tensorflow/python/kernel_tests/substr_op_test.py
+++ b/tensorflow/python/kernel_tests/substr_op_test.py
@@ -183,7 +183,7 @@ class SubstrOpTest(test.TestCase):
 
     position = np.array([[1, 2, 3], [1, 2, 3], [1, 2, 3]], dtype)
     length = np.array([[2, 3, 4]], dtype)
-    # Should fail: postion/length have different dimensionality
+    # Should fail: position/length have different dimensionality
     with self.assertRaises(ValueError):
       substr_op = string_ops.substr(test_string, position, length)
 
diff --git a/tensorflow/python/kernel_tests/variable_scope_test.py b/tensorflow/python/kernel_tests/variable_scope_test.py
index 245dcc96db7..7108131d53d 100644
--- a/tensorflow/python/kernel_tests/variable_scope_test.py
+++ b/tensorflow/python/kernel_tests/variable_scope_test.py
@@ -115,7 +115,7 @@ class VariableScopeTest(test.TestCase):
           dtypes.int64, dtypes.bool
       ]
 
-      # Use different varibale_name to distinguish various dtypes
+      # Use different variable_name to distinguish various dtypes
       for (i, dtype) in enumerate(types):
         x = variable_scope.get_variable(
             name="x%d" % i, shape=(3, 4), dtype=dtype)
@@ -807,7 +807,7 @@ class VariableScopeWithPartitioningTest(test.TestCase):
           dtypes.int64, dtypes.bool
       ]
 
-      # Use different varibale_name to distinguish various dtypes
+      # Use different variable_name to distinguish various dtypes
       for (i, dtype) in enumerate(types):
         x = variable_scope.get_variable(
             name="x%d" % i,
diff --git a/tensorflow/python/ops/ctc_ops.py b/tensorflow/python/ops/ctc_ops.py
index 4ea4d9ed2dd..477c0d1cb49 100644
--- a/tensorflow/python/ops/ctc_ops.py
+++ b/tensorflow/python/ops/ctc_ops.py
@@ -37,7 +37,7 @@ def ctc_loss(labels, inputs, sequence_length,
   This op implements the CTC loss as presented in the article:
 
   [A. Graves, S. Fernandez, F. Gomez, J. Schmidhuber.
-  Connectionist Temporal Classification: Labelling Unsegmented Sequence Data
+  Connectionist Temporal Classification: Labeling Unsegmented Sequence Data
   with Recurrent Neural Networks. ICML 2006, Pittsburgh, USA, pp. 369-376.](http://www.cs.toronto.edu/~graves/icml_2006.pdf)
 
   Input requirements:
diff --git a/tensorflow/python/ops/data_flow_ops.py b/tensorflow/python/ops/data_flow_ops.py
index eb9d0ba7e32..00f339c3935 100644
--- a/tensorflow/python/ops/data_flow_ops.py
+++ b/tensorflow/python/ops/data_flow_ops.py
@@ -516,7 +516,7 @@ class QueueBase(object):
     that would block will fail immediately.
 
     If `cancel_pending_enqueues` is `True`, all pending requests will also
-    be cancelled.
+    be canceled.
 
     Args:
       cancel_pending_enqueues: (Optional.) A boolean, defaulting to
@@ -988,7 +988,7 @@ class Barrier(object):
     TakeMany operations that would block will fail immediately.
 
     If `cancel_pending_enqueues` is `True`, all pending requests to the
-    underlying queue will also be cancelled, and completing of already
+    underlying queue will also be canceled, and completing of already
     started values is also not acceptable anymore.
 
     Args:
diff --git a/tensorflow/python/ops/distributions/transformed_distribution.py b/tensorflow/python/ops/distributions/transformed_distribution.py
index 09b26a9fb73..1be3819569c 100644
--- a/tensorflow/python/ops/distributions/transformed_distribution.py
+++ b/tensorflow/python/ops/distributions/transformed_distribution.py
@@ -339,7 +339,7 @@ class TransformedDistribution(distribution_lib.Distribution):
             self.distribution.event_shape_tensor()))
 
   def _event_shape(self):
-    # If there's a chance that the event_shape has been overriden, we return
+    # If there's a chance that the event_shape has been overridden, we return
     # what we statically know about the `event_shape_override`. This works
     # because: `_is_maybe_event_override` means `static_override` is `None` or a
     # non-empty list, i.e., we don't statically know the `event_shape` or we do.
@@ -360,7 +360,7 @@ class TransformedDistribution(distribution_lib.Distribution):
         self.distribution.batch_shape_tensor())
 
   def _batch_shape(self):
-    # If there's a chance that the batch_shape has been overriden, we return
+    # If there's a chance that the batch_shape has been overridden, we return
     # what we statically know about the `batch_shape_override`. This works
     # because: `_is_maybe_batch_override` means `static_override` is `None` or a
     # non-empty list, i.e., we don't statically know the `batch_shape` or we do.
diff --git a/tensorflow/python/ops/image_ops_test.py b/tensorflow/python/ops/image_ops_test.py
index 492dbe6d135..5588d18ef1d 100644
--- a/tensorflow/python/ops/image_ops_test.py
+++ b/tensorflow/python/ops/image_ops_test.py
@@ -1449,7 +1449,7 @@ class PadToBoundingBoxTest(test_util.TensorFlowTestCase):
           use_tensor_inputs_options=[False])
 
       # The orignal error message does not contain back slashes. However, they
-      # are added by either the assert op or the runtime. If this behaviour
+      # are added by either the assert op or the runtime. If this behavior
       # changes in the future, the match string will also needs to be changed.
       self._assertRaises(
           x,
@@ -2281,7 +2281,7 @@ class ResizeImageWithCropOrPadTest(test_util.TensorFlowTestCase):
           use_tensor_inputs_options=[False])
 
       # The orignal error message does not contain back slashes. However, they
-      # are added by either the assert op or the runtime. If this behaviour
+      # are added by either the assert op or the runtime. If this behavior
       # changes in the future, the match string will also needs to be changed.
       self._assertRaises(
           x,
diff --git a/tensorflow/python/ops/math_ops.py b/tensorflow/python/ops/math_ops.py
index 1555d19395f..9e989210257 100644
--- a/tensorflow/python/ops/math_ops.py
+++ b/tensorflow/python/ops/math_ops.py
@@ -386,7 +386,7 @@ def sign(x, name=None):
     A `Tensor` or `SparseTensor`, respectively. Has the same type as `x`.
 
   @compatibility(numpy)
-  Equivalent to numpy.sign except for the behaviour for input values of NaN.
+  Equivalent to numpy.sign except for the behavior for input values of NaN.
   @end_compatibility
   """
   with ops.name_scope(name, "Sign", [x]) as name:
diff --git a/tensorflow/python/ops/math_ops_test.py b/tensorflow/python/ops/math_ops_test.py
index 120827d18b0..617d2305bd8 100644
--- a/tensorflow/python/ops/math_ops_test.py
+++ b/tensorflow/python/ops/math_ops_test.py
@@ -413,9 +413,9 @@ class DivAndModTest(test_util.TensorFlowTestCase):
       tf_divs = array_ops.constant(divs)
       tf2_result = (tf_nums // tf_divs * tf_divs + tf_nums % tf_divs).eval()
       np_result = (nums // divs) * divs + (nums % divs)
-      # consistentcy with numpy
+      # Consistent with numpy
       self.assertAllEqual(tf_result, np_result)
-      # consistentcy with two forms of divide
+      # Consistent with two forms of divide
       self.assertAllEqual(tf_result, tf2_result)
       # consistency for truncation form
       tf3_result = (math_ops.truncatediv(nums, divs) * divs +
diff --git a/tensorflow/python/ops/sparse_ops.py b/tensorflow/python/ops/sparse_ops.py
index 3c1ab3248b6..ee08ef534b6 100644
--- a/tensorflow/python/ops/sparse_ops.py
+++ b/tensorflow/python/ops/sparse_ops.py
@@ -1460,7 +1460,7 @@ def sparse_tensor_dense_matmul(sp_a,
   `sp_a.dense_shape` takes on large values.
 
   Below is a rough speed comparison between `sparse_tensor_dense_matmul`,
-  labelled 'sparse', and `matmul`(a_is_sparse=True), labelled 'dense'.  For
+  labeled 'sparse', and `matmul`(a_is_sparse=True), labeled 'dense'.  For
   purposes of the comparison, the time spent converting from a `SparseTensor` to
   a dense `Tensor` is not included, so it is overly conservative with respect to
   the time ratio.
diff --git a/tensorflow/python/ops/special_math_ops.py b/tensorflow/python/ops/special_math_ops.py
index 851fba0beba..b561203bb47 100644
--- a/tensorflow/python/ops/special_math_ops.py
+++ b/tensorflow/python/ops/special_math_ops.py
@@ -424,7 +424,7 @@ def _exponential_space_einsum(equation, *inputs):
   missing_idx = set(idx_out).difference(idx_all)
   if missing_idx:
     raise ValueError(
-        'Unknown ouput axes: %s' % missing_idx
+        'Unknown output axes: %s' % missing_idx
     )
 
   axis_order = {}
diff --git a/tensorflow/python/ops/variable_scope.py b/tensorflow/python/ops/variable_scope.py
index a29ddfa9f2f..aceffd373af 100644
--- a/tensorflow/python/ops/variable_scope.py
+++ b/tensorflow/python/ops/variable_scope.py
@@ -282,7 +282,7 @@ class _VariableStore(object):
 
     # If a *_ref type is passed in an error would be triggered further down the
     # stack. We prevent this using base_dtype to get a non-ref version of the
-    # type, before doing anything else. When _ref types are removed in favour of
+    # type, before doing anything else. When _ref types are removed in favor of
     # resources, this line can be removed.
     try:
       dtype = dtype.base_dtype
diff --git a/tensorflow/python/tools/import_pb_to_tensorboard.py b/tensorflow/python/tools/import_pb_to_tensorboard.py
index caeb04a24bf..2bb055e9786 100644
--- a/tensorflow/python/tools/import_pb_to_tensorboard.py
+++ b/tensorflow/python/tools/import_pb_to_tensorboard.py
@@ -31,7 +31,7 @@ def import_to_tensorboard(model_dir, log_dir):
 
   Args:
     model_dir: The location of the protobuf (`pb`) model to visualize
-    log_dir: The location for the Tensorboard log to begin visualisation from.
+    log_dir: The location for the Tensorboard log to begin visualization from.
 
   Usage:
     Call this function with your model location and desired log directory.
diff --git a/tensorflow/python/training/evaluation.py b/tensorflow/python/training/evaluation.py
index 7c46591d07b..bbaa3931c20 100644
--- a/tensorflow/python/training/evaluation.py
+++ b/tensorflow/python/training/evaluation.py
@@ -113,7 +113,7 @@ def _evaluate_once(checkpoint_path,
 
   One may also consider using a `tf.contrib.training.SummaryAtEndHook` to record
   summaries after the `eval_ops` have run. If `eval_ops` is `None`, the
-  summaries run immedietly after the model checkpoint has been restored.
+  summaries run immediately after the model checkpoint has been restored.
 
   Note that `evaluate_once` creates a local variable used to track the number of
   evaluations run via `tf.contrib.training.get_or_create_eval_step`.
diff --git a/tensorflow/python/util/lazy_loader.py b/tensorflow/python/util/lazy_loader.py
index 34308ff9313..6d2622b1c04 100644
--- a/tensorflow/python/util/lazy_loader.py
+++ b/tensorflow/python/util/lazy_loader.py
@@ -24,7 +24,7 @@ import types
 
 
 class LazyLoader(types.ModuleType):
-  """Lazily import a module, mainly to avoid pulling in large dependancies.
+  """Lazily import a module, mainly to avoid pulling in large dependencies.
 
   `contrib`, and `ffmpeg` are examples of modules that are large and not always
   needed, and this allows them to only be loaded when they are used.
diff --git a/tensorflow/stream_executor/cuda/cuda_diagnostics.h b/tensorflow/stream_executor/cuda/cuda_diagnostics.h
index 5cce6b93656..aa68321acc8 100644
--- a/tensorflow/stream_executor/cuda/cuda_diagnostics.h
+++ b/tensorflow/stream_executor/cuda/cuda_diagnostics.h
@@ -75,7 +75,7 @@ class Diagnostician {
 
   // Given the DSO version number and the driver version file contents, extracts
   // the driver version and compares, warning the user in the case of
-  // incompatability.
+  // incompatibility.
   //
   // This is solely used for more informative log messages when the user is
   // running on a machine that happens to have a libcuda/kernel driver mismatch.
diff --git a/tensorflow/stream_executor/cuda/cuda_driver.h b/tensorflow/stream_executor/cuda/cuda_driver.h
index c5d7d8b32f3..68494aba659 100644
--- a/tensorflow/stream_executor/cuda/cuda_driver.h
+++ b/tensorflow/stream_executor/cuda/cuda_driver.h
@@ -77,7 +77,7 @@ class CUDADriver {
 
   // Destroys a CUDA stream associated with the given context.
   // stream is owned by the caller, must not be null, and *stream is set to null
-  // if the stream is successfuly destroyed.
+  // if the stream is successfully destroyed.
   // http://docs.nvidia.com/cuda/cuda-driver-api/group__CUDA__STREAM.html#group__CUDA__STREAM_1g244c8833de4596bcd31a06cdf21ee758
   static void DestroyStream(CudaContext* context, CUstream *stream);
 
diff --git a/tensorflow/stream_executor/cuda/cuda_event.h b/tensorflow/stream_executor/cuda/cuda_event.h
index 46f0232b1dd..56667e65d38 100644
--- a/tensorflow/stream_executor/cuda/cuda_event.h
+++ b/tensorflow/stream_executor/cuda/cuda_event.h
@@ -46,7 +46,7 @@ class CUDAEvent : public internal::EventInterface {
   // Polls the CUDA platform for the event's current status.
   Event::Status PollForStatus();
 
-  // The underyling CUDA event element.
+  // The underlying CUDA event element.
   const CUevent& cuda_event();
 
  private:
diff --git a/tensorflow/stream_executor/cuda/cuda_gpu_executor.cc b/tensorflow/stream_executor/cuda/cuda_gpu_executor.cc
index c1e72bb5655..43c707730af 100644
--- a/tensorflow/stream_executor/cuda/cuda_gpu_executor.cc
+++ b/tensorflow/stream_executor/cuda/cuda_gpu_executor.cc
@@ -847,7 +847,7 @@ void *CUDAExecutor::CudaContextHack() { return context_; }
 
 CudaContext* CUDAExecutor::cuda_context() { return context_; }
 
-// Attemps to read the NUMA node corresponding to the GPU device's PCI bus out
+// Attempts to read the NUMA node corresponding to the GPU device's PCI bus out
 // of SysFS. Returns -1 if it cannot.
 //
 // For anything more complicated/prod-focused than this, you'll likely want to
diff --git a/tensorflow/stream_executor/plugin.h b/tensorflow/stream_executor/plugin.h
index b1db8b7cb87..0b88b86e2b1 100644
--- a/tensorflow/stream_executor/plugin.h
+++ b/tensorflow/stream_executor/plugin.h
@@ -49,7 +49,7 @@ enum class PluginKind {
 //
 // A PluginConfig may be passed to the StreamExecutor constructor - the plugins
 // described therein will be used to provide BLAS, DNN, FFT, and RNG
-// functionality. Platform-approprate defaults will be used for any un-set
+// functionality. Platform-appropriate defaults will be used for any un-set
 // libraries. If a platform does not support a specified plugin (ex. cuBLAS on
 // an OpenCL executor), then an error will be logged and no plugin operations
 // will succeed.
diff --git a/tensorflow/stream_executor/stream_executor_pimpl.h b/tensorflow/stream_executor/stream_executor_pimpl.h
index 5c52afa7944..780d12c8dce 100644
--- a/tensorflow/stream_executor/stream_executor_pimpl.h
+++ b/tensorflow/stream_executor/stream_executor_pimpl.h
@@ -205,7 +205,7 @@ class StreamExecutor {
   // This should be done before deallocating the region with delete[]/free/etc.
   bool HostMemoryUnregister(void *location) SE_MUST_USE_RESULT;
 
-  // Synchronizes all activity occuring in the StreamExecutor's context (most
+  // Synchronizes all activity occurring in the StreamExecutor's context (most
   // likely a whole device).
   bool SynchronizeAllActivity() SE_MUST_USE_RESULT;
 
@@ -238,7 +238,7 @@ class StreamExecutor {
                                     DeviceMemoryBase *gpu_dst);
 
   // Alternative interface for memcpying from host to device that takes an
-  // array slice. Checks that the destination size can accomodate the host
+  // array slice. Checks that the destination size can accommodate the host
   // slice size.
   template <class T>
   port::Status SynchronousMemcpyH2D(port::ArraySlice<T> host_src,
@@ -253,7 +253,7 @@ class StreamExecutor {
                                     void *host_dst);
 
   // Alternative interface for memcpying from device to host that takes an
-  // array slice. Checks that the destination size can accomodate the host
+  // array slice. Checks that the destination size can accommodate the host
   // slice size.
   template <typename T>
   port::Status SynchronousMemcpyD2H(const DeviceMemory<T> &gpu_src,
diff --git a/tensorflow/tensorboard/components/tf_backend_d3v4/behavior.ts b/tensorflow/tensorboard/components/tf_backend_d3v4/behavior.ts
index ab8a64b7b13..dc47df2a5c2 100644
--- a/tensorflow/tensorboard/components/tf_backend_d3v4/behavior.ts
+++ b/tensorflow/tensorboard/components/tf_backend_d3v4/behavior.ts
@@ -86,7 +86,7 @@ export const BackendBehavior = {
    * Backend reload, which gets metadata on available runs, tags, etc from
    *   the backend.
    * Frontend reload, which loads new data for each chart or visual display.
-   * Backend reload logic is provided by this behaivor. The frontend reload
+   * Backend reload logic is provided by this behavior. The frontend reload
    *   logic should be provided elsewhere, since it is component-specific.
    * To keep things simple and consistent, we do the backend reload first,
    *   and the frontend reload afterwards.
diff --git a/tensorflow/tensorboard/components/tf_dashboard_common/tf-chart-scaffold.html b/tensorflow/tensorboard/components/tf_dashboard_common/tf-chart-scaffold.html
index 83b141cb98a..b9d0a8c39ee 100644
--- a/tensorflow/tensorboard/components/tf_dashboard_common/tf-chart-scaffold.html
+++ b/tensorflow/tensorboard/components/tf_dashboard_common/tf-chart-scaffold.html
@@ -34,7 +34,7 @@ chart() - Returns the underlying chart element.
 reload() - Reloads the data and sends it to the underlying chart.
 
 This element should have a compatible chart plugin element as it's content. The
-plugin is requred to implement two functions:
+plugin is required to implement two functions:
 - setVisibleSeries(names: string[]): a function that receives an array of series
     names as the first parameter, responsible for changing the series currently
     being displayed to only the series in this array.
diff --git a/tensorflow/tensorboard/components/tf_dashboard_common_d3v4/tf-chart-scaffold.html b/tensorflow/tensorboard/components/tf_dashboard_common_d3v4/tf-chart-scaffold.html
index 9cacb7f5c89..a39fb9462ba 100644
--- a/tensorflow/tensorboard/components/tf_dashboard_common_d3v4/tf-chart-scaffold.html
+++ b/tensorflow/tensorboard/components/tf_dashboard_common_d3v4/tf-chart-scaffold.html
@@ -32,7 +32,7 @@ chart() - Returns the underlying chart element.
 reload() - Reloads the data and sends it to the underlying chart.
 
 This element should have a compatible chart plugin element as it's content. The
-plugin is requred to implement two functions:
+plugin is required to implement two functions:
 - setVisibleSeries(names: string[]): a function that receives an array of series
     names as the first parameter, responsible for changing the series currently
     being displayed to only the series in this array.
diff --git a/tensorflow/tensorboard/components/tf_graph_common_d3v4/util.ts b/tensorflow/tensorboard/components/tf_graph_common_d3v4/util.ts
index 7f4d329e795..d0be1d6ba5a 100644
--- a/tensorflow/tensorboard/components/tf_graph_common_d3v4/util.ts
+++ b/tensorflow/tensorboard/components/tf_graph_common_d3v4/util.ts
@@ -68,7 +68,7 @@ module tf.graph.util {
    * progress
    * of the subtask and the subtask message. The parent task should pass a
    * subtracker to its subtasks. The subtask reports its own progress which
-   * becames relative to the main task.
+   * becomes relative to the main task.
    */
   export function getSubtaskTracker(
       parentTracker: ProgressTracker, impactOnTotalProgress: number,
diff --git a/tensorflow/tensorboard/components/tf_graph_d3v4/tf-graph-scene.html b/tensorflow/tensorboard/components/tf_graph_d3v4/tf-graph-scene.html
index 10a65f54d52..35705713b98 100644
--- a/tensorflow/tensorboard/components/tf_graph_d3v4/tf-graph-scene.html
+++ b/tensorflow/tensorboard/components/tf_graph_d3v4/tf-graph-scene.html
@@ -322,7 +322,7 @@ limitations under the License.
 /* --- Annotation --- */
 
 /* only applied for annotations that are not summary or constant.
-(.summary, .constant gets overriden below) */
+(.summary, .constant gets overridden below) */
 ::content .annotation > .annotation-node > * {
   stroke-width: 0.5;
   stroke-dasharray: 1, 1;
diff --git a/tensorflow/tensorboard/components/tf_storage/storage.ts b/tensorflow/tensorboard/components/tf_storage/storage.ts
index 5516ae7ded3..e8d5fa672fb 100644
--- a/tensorflow/tensorboard/components/tf_storage/storage.ts
+++ b/tensorflow/tensorboard/components/tf_storage/storage.ts
@@ -22,7 +22,7 @@ limitations under the License.
  * which TensorBoard uses after like localhost:8000/#events&runPrefix=train*
  * to store state in the URI.
  *
- * It also allows saving the values to localStorage for long-term persistance.
+ * It also allows saving the values to localStorage for long-term persistence.
  */
 module TF.URIStorage {
   type StringDict = {[key: string]: string};
@@ -36,7 +36,7 @@ module TF.URIStorage {
   /**
    * The name of the property for users to set on a Polymer component
    * in order for its stored properties to be stored in the URI unambiguously.
-   * (No need to set this if you want mutliple instances of the component to
+   * (No need to set this if you want multiple instances of the component to
    * share URI state)
    *
    * Example:
@@ -257,7 +257,7 @@ module TF.URIStorage {
    * Convert dictionary of strings into a URI Component.
    * All key value entries get added as key value pairs in the component,
    * with the exception of a key with the TAB value, which if present
-   * gets prepended to the URI Component string for backwards comptability
+   * gets prepended to the URI Component string for backwards compatibility
    * reasons.
    */
   function _dictToComponent(items: StringDict): string {
diff --git a/tensorflow/tensorboard/components/tf_storage_d3v4/storage.ts b/tensorflow/tensorboard/components/tf_storage_d3v4/storage.ts
index 19a27b9cbd1..1b39efc03a1 100644
--- a/tensorflow/tensorboard/components/tf_storage_d3v4/storage.ts
+++ b/tensorflow/tensorboard/components/tf_storage_d3v4/storage.ts
@@ -25,7 +25,7 @@ import * as _ from 'lodash';
  * which TensorBoard uses after like localhost:8000/#events&runPrefix=train*
  * to store state in the URI.
  *
- * It also allows saving the values to localStorage for long-term persistance.
+ * It also allows saving the values to localStorage for long-term persistence.
  */
 type StringDict = {[key: string]: string};
 
@@ -38,7 +38,7 @@ export let TAB = '__tab__';
 /**
  * The name of the property for users to set on a Polymer component
  * in order for its stored properties to be stored in the URI unambiguously.
- * (No need to set this if you want mutliple instances of the component to
+ * (No need to set this if you want multiple instances of the component to
  * share URI state)
  *
  * Example:
@@ -258,7 +258,7 @@ function _writeComponent(component: string) {
  * Convert dictionary of strings into a URI Component.
  * All key value entries get added as key value pairs in the component,
  * with the exception of a key with the TAB value, which if present
- * gets prepended to the URI Component string for backwards comptability
+ * gets prepended to the URI Component string for backwards compatibility
  * reasons.
  */
 function _dictToComponent(items: StringDict): string {
diff --git a/tensorflow/tensorboard/components/vz_data_summary/vz-data-summary.ts b/tensorflow/tensorboard/components/vz_data_summary/vz-data-summary.ts
index 9a4e80c8a98..27faf35f6bb 100644
--- a/tensorflow/tensorboard/components/vz_data_summary/vz-data-summary.ts
+++ b/tensorflow/tensorboard/components/vz_data_summary/vz-data-summary.ts
@@ -118,7 +118,7 @@ function createRect(
   // Set dimensions.
   rect.setAttribute('width', sliceWidth.toString());
   rect.setAttribute('height', height.toString());
-  // Set colour.
+  // Set color.
   rect.setAttribute('fill', color);
 
   return rect;
diff --git a/tensorflow/tensorboard/components/vz_heatmap/vz-heatmap.html b/tensorflow/tensorboard/components/vz_heatmap/vz-heatmap.html
index 4a005422765..09fa8579291 100644
--- a/tensorflow/tensorboard/components/vz_heatmap/vz-heatmap.html
+++ b/tensorflow/tensorboard/components/vz_heatmap/vz-heatmap.html
@@ -20,7 +20,7 @@ limitations under the License.
 
 <!--
 `vz-heatmap` A simple heatmap to visualize 2D data using predefined or user
-defined colour scheme, with a dependency on d3.js. The heatmap automatically
+defined color scheme, with a dependency on d3.js. The heatmap automatically
 fits itself to the width of the container it is placed in.
 
 @element vz-heatmap
@@ -287,7 +287,7 @@ fits itself to the width of the container it is placed in.
             this._generatedColorScale.domain(this.values);
           }
 
-          // If values reset, and data field contains valid data, set colour scale
+          // If values reset, and data field contains valid data, set color scale
           // to use the data extent as domain.
         } else if (this._currentDataValid) {
           this._generatedColorScale.domain(this._dataExtent);
diff --git a/tensorflow/tensorboard/components/vz_heatmap_d3v4/vz-heatmap.html b/tensorflow/tensorboard/components/vz_heatmap_d3v4/vz-heatmap.html
index 8a3503468fa..089692576fa 100644
--- a/tensorflow/tensorboard/components/vz_heatmap_d3v4/vz-heatmap.html
+++ b/tensorflow/tensorboard/components/vz_heatmap_d3v4/vz-heatmap.html
@@ -21,7 +21,7 @@ limitations under the License.
 
 <!--
 `vz-heatmap` A simple heatmap to visualize 2D data using predefined or user
-defined colour scheme, with a dependency on d3.js. The heatmap automatically
+defined color scheme, with a dependency on d3.js. The heatmap automatically
 fits itself to the width of the container it is placed in.
 
 @element vz-heatmap
@@ -288,7 +288,7 @@ fits itself to the width of the container it is placed in.
             this._generatedColorScale.domain(this.values);
           }
 
-          // If values reset, and data field contains valid data, set colour scale
+          // If values reset, and data field contains valid data, set color scale
           // to use the data extent as domain.
         } else if (this._currentDataValid) {
           this._generatedColorScale.domain(this._dataExtent);
diff --git a/tensorflow/tensorboard/components/vz_projector_d3v4/scatterPlotVisualizer3DLabels.ts b/tensorflow/tensorboard/components/vz_projector_d3v4/scatterPlotVisualizer3DLabels.ts
index cbd9785e2f6..7820af0d48d 100644
--- a/tensorflow/tensorboard/components/vz_projector_d3v4/scatterPlotVisualizer3DLabels.ts
+++ b/tensorflow/tensorboard/components/vz_projector_d3v4/scatterPlotVisualizer3DLabels.ts
@@ -38,7 +38,7 @@ const VERTICES_PER_GLYPH = 2 * 3;  // 2 triangles, 3 verts per triangle
  *            bottom center of the word is positioned at (0, 0);
  *    position: The position of the label in worldspace.
  *    vUv: The (u, v) coordinates that index into the glyphs sheet (range 0, 1.)
- *    color: The color of the label (matches the cooresponding point's color.)
+ *    color: The color of the label (matches the corresponding point's color.)
  *    wordShown: Boolean. Whether or not the label is visible.
  */
 
diff --git a/tensorflow/tensorboard/components/vz_projector_d3v4/vz-projector-inspector-panel.html b/tensorflow/tensorboard/components/vz_projector_d3v4/vz-projector-inspector-panel.html
index 7554c322cef..f5087a99967 100644
--- a/tensorflow/tensorboard/components/vz_projector_d3v4/vz-projector-inspector-panel.html
+++ b/tensorflow/tensorboard/components/vz_projector_d3v4/vz-projector-inspector-panel.html
@@ -223,7 +223,7 @@ limitations under the License.
         <span class="option-label">distance</span>
         <div class="options">
           <a class="selected cosine" href="javascript:void(0);">COSINE</a>
-          <a class="euclidean" href="javascript:void(0);">EUCLIDIAN</a>
+          <a class="euclidean" href="javascript:void(0);">EUCLIDEAN</a>
         </div>
       </div>
     </div>
diff --git a/tensorflow/tensorboard/scripts/generate_testdata.py b/tensorflow/tensorboard/scripts/generate_testdata.py
index 64cc9ff2dd4..e8fb9cd6d4b 100644
--- a/tensorflow/tensorboard/scripts/generate_testdata.py
+++ b/tensorflow/tensorboard/scripts/generate_testdata.py
@@ -38,7 +38,7 @@ from tensorflow.python.platform import app
 from tensorflow.python.platform import flags
 from tensorflow.python.summary.writer import writer as writer_lib
 
-tf.flags.DEFINE_string("target", None, """The directoy where serialized data
+tf.flags.DEFINE_string("target", None, """The directory where serialized data
 will be written""")
 
 flags.DEFINE_boolean("overwrite", False, """Whether to remove and overwrite
diff --git a/tensorflow/tools/ci_build/windows/bazel/bazel_test_lib.sh b/tensorflow/tools/ci_build/windows/bazel/bazel_test_lib.sh
index 65d05d8d7d4..21b6fa2bb99 100644
--- a/tensorflow/tools/ci_build/windows/bazel/bazel_test_lib.sh
+++ b/tensorflow/tools/ci_build/windows/bazel/bazel_test_lib.sh
@@ -97,7 +97,7 @@ exclude_cpu_cc_tests="${failing_cpu_cc_tests} + ${broken_cpu_cc_tests}"
 exclude_gpu_cc_tests="${extra_failing_gpu_cc_tests} + ${exclude_cpu_cc_tests}"
 
 # Python tests
-# The first argument is the name of the python test direcotry
+# The first argument is the name of the python test directory
 function get_failing_cpu_py_tests() {
     echo "
     //$1/tensorflow/python:basic_session_run_hooks_test + \
diff --git a/tensorflow/tools/dist_test/python/census_widendeep.py b/tensorflow/tools/dist_test/python/census_widendeep.py
index db56a687f6b..3a557814960 100644
--- a/tensorflow/tools/dist_test/python/census_widendeep.py
+++ b/tensorflow/tools/dist_test/python/census_widendeep.py
@@ -133,7 +133,7 @@ class CensusDataSource(object):
       columns: Columns to retrieve from the data files (A list of strings)
       label_column: Name of the label column
       categorical_columns: Names of the categorical columns (A list of strings)
-      continuous_columns: Names of the continuous columsn (A list of strings)
+      continuous_columns: Names of the continuous columns (A list of strings)
     """
 
     # Retrieve data from disk (if available) or download from the web.
diff --git a/tensorflow/tools/docker/parameterized_docker_build.sh b/tensorflow/tools/docker/parameterized_docker_build.sh
index 886266caaf8..f88af68cdeb 100755
--- a/tensorflow/tools/docker/parameterized_docker_build.sh
+++ b/tensorflow/tools/docker/parameterized_docker_build.sh
@@ -64,7 +64,7 @@
 #
 #   TF_DOCKER_BUILD_OPTIONS
 #     (Optional)
-#     Specifices the desired build options. Defaults to OPT.
+#     Specifies the desired build options. Defaults to OPT.
 
 # Script directory
 SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
diff --git a/tensorflow/tools/docs/py_guide_parser.py b/tensorflow/tools/docs/py_guide_parser.py
index 245643cb32e..216353ecee3 100644
--- a/tensorflow/tools/docs/py_guide_parser.py
+++ b/tensorflow/tools/docs/py_guide_parser.py
@@ -35,7 +35,7 @@ class PyGuideParser(object):
   """Simple parsing of a guide .md file.
 
   Descendants can override the process_*() functions (called by process())
-  to either record infromation from the guide, or call replace_line()
+  to either record information from the guide, or call replace_line()
   to affect the return value of process().
   """
 
diff --git a/tensorflow/tools/gcs_test/python/gcs_smoke.py b/tensorflow/tools/gcs_test/python/gcs_smoke.py
index 615e142c471..51933a52a66 100644
--- a/tensorflow/tools/gcs_test/python/gcs_smoke.py
+++ b/tensorflow/tools/gcs_test/python/gcs_smoke.py
@@ -36,7 +36,7 @@ flags.DEFINE_integer("num_examples", 10, "Number of examples to generate")
 FLAGS = flags.FLAGS
 
 def create_examples(num_examples, input_mean):
-  """Create ExampleProto's containg data."""
+  """Create ExampleProto's containing data."""
   ids = np.arange(num_examples).reshape([num_examples, 1])
   inputs = np.random.randn(num_examples, 1) + input_mean
   target = inputs - input_mean
diff --git a/tensorflow/tools/graph_transforms/README.md b/tensorflow/tools/graph_transforms/README.md
index 7e8c51efe6a..53f7f1685f3 100644
--- a/tensorflow/tools/graph_transforms/README.md
+++ b/tensorflow/tools/graph_transforms/README.md
@@ -998,7 +998,7 @@ There are a few things to know about the `ReplaceMatchingOpTypes` function:
     important nodes are listed in the `output_nodes` argument that's passed into
     each replacement function call. You can disable this checking by setting
     `allow_inconsistencies` to true in the options, but otherwise any
-    replacements that break the graph constraints will be cancelled. If you do
+    replacements that break the graph constraints will be canceled. If you do
     allow inconsistencies, it's your transform's responsibility to fix them up
     before you return your final result. Functions like `RenameNodeInputs` can
     be useful if you are doing wholesale node renaming for example.
diff --git a/tensorflow/tools/graph_transforms/transform_utils.h b/tensorflow/tools/graph_transforms/transform_utils.h
index 54808efa9fb..95cb21d64c6 100644
--- a/tensorflow/tools/graph_transforms/transform_utils.h
+++ b/tensorflow/tools/graph_transforms/transform_utils.h
@@ -106,7 +106,7 @@ void FilterGraphDef(const GraphDef& input_graph_def,
                     std::function<bool(const NodeDef&)> selector,
                     GraphDef* output_graph_def);
 
-// Creates a copy of the input graph, with all occurences of the attributes with
+// Creates a copy of the input graph, with all occurrences of the attributes with
 // the names in the argument removed from the node defs.
 void RemoveAttributes(const GraphDef& input_graph_def,
                       const std::vector<string>& attributes,
diff --git a/third_party/gpus/cuda_configure.bzl b/third_party/gpus/cuda_configure.bzl
index c6deae05b88..12d324e9a14 100644
--- a/third_party/gpus/cuda_configure.bzl
+++ b/third_party/gpus/cuda_configure.bzl
@@ -5,7 +5,7 @@
 
   * `TF_NEED_CUDA`: Whether to enable building with CUDA.
   * `GCC_HOST_COMPILER_PATH`: The GCC host compiler path
-  * `TF_CUDA_CLANG`: Wheter to use clang as a cuda compiler.
+  * `TF_CUDA_CLANG`: Whether to use clang as a cuda compiler.
   * `CLANG_CUDA_COMPILER_PATH`: The clang compiler path that will be used for
     both host and device code compilation if TF_CUDA_CLANG is 1.
   * `CUDA_TOOLKIT_PATH`: The path to the CUDA toolkit. Default is
diff --git a/third_party/hadoop/hdfs.h b/third_party/hadoop/hdfs.h
index 560d8bba0e0..a664f3b50cf 100644
--- a/third_party/hadoop/hdfs.h
+++ b/third_party/hadoop/hdfs.h
@@ -171,7 +171,7 @@ void hdfsFileFreeReadStatistics(struct hdfsReadStatistics *stats);
  * Connect to the hdfs.
  * @param nn   The NameNode.  See hdfsBuilderSetNameNode for details.
  * @param port The port on which the server is listening.
- * @param user the user name (this is hadoop domain user). Or NULL is equivelant
+ * @param user the user name (this is hadoop domain user). Or NULL is equivalent
  * to hhdfsConnect(host, port)
  * @return Returns a handle to the filesystem or NULL on error.
  * @deprecated Use hdfsBuilderConnect instead.
@@ -397,7 +397,7 @@ hdfsFile hdfsOpenFile(hdfsFS fs, const char *path, int flags, int bufferSize,
                       short replication, tSize blocksize);
 
 /**
- * hdfsTruncateFile - Truncate a hdfs file to given lenght.
+ * hdfsTruncateFile - Truncate a hdfs file to given length.
  * @param fs The configured filesystem handle.
  * @param path The full path to the file.
  * @param newlength The size the file is to be truncated to
diff --git a/third_party/py/python_configure.bzl b/third_party/py/python_configure.bzl
index 5f5380c40c2..64ca6ee02f3 100644
--- a/third_party/py/python_configure.bzl
+++ b/third_party/py/python_configure.bzl
@@ -113,7 +113,7 @@ def _symlink_genrule_for_dir(repository_ctx, src_dir, dest_dir, genrule_name):
     # Copy the source directory to the output directory.
     # Creating a junction pointing to the source directory and declaring the
     # outputs like we do on Linux, then expecting that Bazel would pick up these
-    # outputs via the junction is ill-concieved, because Bazel deletes the
+    # outputs via the junction is ill-conceived, because Bazel deletes the
     # output files from previous build prior to executing the action, meaning it
     # would delete the actual files via the old junction.
     cmd_script = "%s-cmd.cmd" % genrule_name
diff --git a/third_party/sycl/sycl/LICENSE.text b/third_party/sycl/sycl/LICENSE.text
index 0c2955c4d76..8d3f050b392 100644
--- a/third_party/sycl/sycl/LICENSE.text
+++ b/third_party/sycl/sycl/LICENSE.text
@@ -67,7 +67,7 @@ you; so please press the "CANCEL" button to cancel your download.
         ComputeCpp within its marketing materials, without the
         express prior written permission of Codeplay.
  4. Support. Codeplay does not provide any guarantees of support for
-    the Software to the user. Codeplay will use reasonable endeavours
+    the Software to the user. Codeplay will use reasonable endeavors
     to respond to users' support requests, for the most recent
     release only, via the community support website at https://
     computecpp.codeplay.com.
@@ -78,7 +78,7 @@ you; so please press the "CANCEL" button to cancel your download.
     copyrights, trade secrets and other proprietary rights in the
     Software, including the rights to make and license the use of all
     copies. To the extent that any patents owned by Codeplay or its
-    licensors relate to any component of the Software, the licence
+    licensors relate to any component of the Software, the license
     granted to the user in accordance with this Agreement allows for
     the lawful use of such patents but only for the purposes of this
     Agreement and not further or otherwise. Therefore, the user may