Fix communicatoin communication (#60995)

paddle/fluid/operators/cinn/cinn_instruction_run_op.cc

@@ -94,7 +94,7 @@ Both the input and output of this operator are a set of variables
 which are the input and output arguments of the bound cinn instruction respectively.
 In addition, there is an attribute named 'cached_index' should be
 set necessarily to get the CinnCompiledObject where the instruction is included
-and 'instruction_index' is fetch the instruction object from complied runtime prograrm.
+and 'instruction_index' is fetch the instruction object from complied runtime program.
 
 It accomplishes the execution of the instruction according to the following steps:
   0. Set the shapes ot the output variables at InferShape function with

paddle/fluid/operators/cinn/cinn_launch_context.cc

@@ -199,7 +199,7 @@ void CinnLaunchContext::BuildVarNameMap(
       paddle2cinn_varmap_.size(),
       cinn2paddle_varmap_.size(),
       platform::errors::PreconditionNotMet(
-          "Size of variables is not euqal, paddle[%ld] vs cinn[%ld]",
+          "Size of variables is not equal, paddle[%ld] vs cinn[%ld]",
           paddle2cinn_varmap_.size(),
           cinn2paddle_varmap_.size()));
 }
@@ -426,7 +426,7 @@ std::unique_ptr<framework::ProgramDesc> CinnLaunchContext::BuildCompiledProgram(
   //   to the new VarDesc.
   //   (2) For all variables, the shape, data type of their VarDescs
   //   are set by values of the corresponding compiled tensors,
-  //   including the in/out variables where the equiality between their tensors
+  //   including the in/out variables where the equality between their tensors
   //   and the CINN compiled ones is verified in corresponding cinn_launch_op.
   for (auto&& arg : cinn_argument_names_) {
     const std::string& var_name = cinn2paddle_varmap_.at(arg);

paddle/fluid/operators/cinn/cinn_launch_context.h

@@ -86,7 +86,7 @@ class CinnLaunchContext {
   // Return whether a Paddle variable used in cinn execution
   bool IsVariableUsed(const std::string& var_name) const;
 
-  // Check the equiality in type and dimension between the tensor
+  // Check the equality in type and dimension between the tensor
   // in Paddle and the compiled tensor returned by CINN of a same variable
   void CheckTensorEquivalent(const std::string& var_name,
                              const phi::DenseTensor& paddle_tensor);
@@ -96,7 +96,7 @@ class CinnLaunchContext {
     return skip_eager_vars_;
   }
 
-  // Redirect the name of a Paddle variable to the orignal if it was inplaced
+  // Redirect the name of a Paddle variable to the original if it was inplaced
   std::string RedirectVarName(const std::string& var_name) const;
 
   // Return internal variable names list
@@ -157,7 +157,7 @@ class CinnLaunchContext {
   // a list of internal variable names in Paddle
   std::unordered_set<std::string> internal_var_names_;
   // In CINN, there are two variables(in/out) mapped to the one inplaced
-  // variable of Paddle. To resovle this conflict, we add a output counterpart
+  // variable of Paddle. To resolve this conflict, we add a output counterpart
   // in Paddle with the name suffixed by @InplaceOut.
   // This set stores which Paddle variable names are inplaced.
   std::unordered_set<std::string> inplace_var_names_;
@@ -179,14 +179,14 @@ class CinnLaunchContext {
   std::vector<std::string> skip_eager_vars_;
 
   // because a cinn_pod_value_t does not own a cinn_buffer_t object,
-  // an extra stroage is necessary to keep those objects and they can
+  // an extra storage is necessary to keep those objects and they can
   // not be released until the runtime program finish execution.
   std::vector<std::unique_ptr<cinn_buffer_t>> hold_buffers_;
   // this map saves all execution arguments with their cinn names as key,
   // and it is passed to the Execute interface of a cinn runtime program.
   std::map<std::string, cinn_pod_value_t> name2argument_;
   // this map saves all execution arguments with paddle variables as key,
-  // this map conbine name2argument_ and paddle2cinn_varmap_
+  // this map combine name2argument_ and paddle2cinn_varmap_
   std::map<std::string, cinn_pod_value_t> paddle2argument_;
 };
 

paddle/fluid/operators/cinn/cinn_launch_op.cc

@@ -54,7 +54,7 @@ void DebugCinnCompiledResult(const CinnCompiledObject& result) {
   const auto& cinn_scope = *(result.scope);
   const auto& paddle2cinn_varmap = result.paddle2cinn_varmap;
 
-  VLOG(4) << "Compiled runtime_program instrunction size:["
+  VLOG(4) << "Compiled runtime_program instruction size:["
           << cinn_runtime_program->size() << "]";
 
   std::vector<std::string> infos;

paddle/fluid/operators/collective/c_comm_init_multitrainer_op.cc

@@ -83,7 +83,7 @@ class CCommInitMultiTrainerOpMaker : public framework::OpProtoAndCheckerMaker {
     AddComment(R"DOC(
 CCommInitMultiTrainer operator
 
-Initialize collective communicatoin context within this trainer
+Initialize collective communication context within this trainer
 )DOC");
     AddAttr<int>("ntrainers",
                  "(int) The number of trainers of distributed trainers");

paddle/fluid/operators/collective/c_gen_xccl_id_op.cc

@@ -73,7 +73,7 @@ class CGenXCCLIdOp : public framework::OperatorBase {
 class CGenXCCLIdOpMaker : public framework::OpProtoAndCheckerMaker {
  public:
   void Make() override {
-    AddOutput("Out", "Raw variable contains a XCCL UniqueId instaces.");
+    AddOutput("Out", "Raw variable contains a XCCL UniqueId instances.");
     AddComment(R"DOC(
 CGenXCCLId operator
 

paddle/fluid/operators/collective/global_gather_op.cu.cc

@@ -319,7 +319,7 @@ struct GlobalGatherProcessGroupFunctor<phi::GPUContext, T> {
   }
 };
 
-template <typename T, typename DeivceContext>
+template <typename T, typename DeviceContext>
 class GlobalGatherOpCUDAKernel : public framework::OpKernel<T> {
  public:
   void Compute(const framework::ExecutionContext& ctx) const override {

paddle/fluid/operators/controlflow/depend_op.cc

@@ -79,7 +79,7 @@ y = opB(x)
 
 if tensor b and tensor x has some inner dependency, for example, x share data with b,
 we need to add explicit dependency for x <- b, otherwise the these two operators may
-be executed parellel in static graph. We can use depend op as below,
+be executed parallel in static graph. We can use depend op as below,
 
 b = opA(a)
 x = depend(x, b)

paddle/fluid/operators/controlflow/feed_op.cc

@@ -155,7 +155,7 @@ class FeedOp : public framework::OperatorWithKernel {
             feed_sparse_tensor.GetIndicesDict());
       } else {
         PADDLE_THROW(
-            phi::errors::Unimplemented("Only support DenseTnesor, Strings, and "
+            phi::errors::Unimplemented("Only support DenseTensor, Strings, and "
                                        "SparseCooTensor for feed op now."));
       }
     }

paddle/fluid/operators/detection/bipartite_match_op.cc

@@ -254,7 +254,7 @@ class BipartiteMatchOpMaker : public framework::OpProtoAndCheckerMaker {
         "represented by each row and each column. For example, assumed one "
         "entity is A with shape [K], another entity is B with shape [M]. The "
         "DistMat[i][j] is the distance between A[i] and B[j]. The bigger "
-        "the distance is, the better macthing the pairs are. Please note, "
+        "the distance is, the better matching the pairs are. Please note, "
         "This tensor can contain LoD information to represent a batch of "
         "inputs. One instance of this batch can contain different numbers of "
         "entities.");

paddle/fluid/operators/detection/collect_fpn_proposals_op.cc

@@ -123,7 +123,7 @@ class CollectFpnProposalsOpMaker : public framework::OpProtoAndCheckerMaker {
                  "Select post_nms_topN RoIs from"
                  " all images and all fpn layers");
     AddComment(R"DOC(
-This operator concats all proposals from different images
+This operator concatenates all proposals from different images
  and different FPN levels. Then sort all of those proposals
 by objectness confidence. Select the post_nms_topN RoIs in
  total. Finally, re-sort the RoIs in the order of batch index.

paddle/fluid/operators/detection/density_prior_box_op.cc

@@ -83,8 +83,8 @@ class DensityPriorBoxOp : public framework::OperatorWithKernel {
             fixed_sizes.size(),
             densities.size()));
     size_t num_priors = 0;
-    for (auto densitie : densities) {
-      num_priors += (fixed_ratios.size()) * (pow(densitie, 2));  // NOLINT
+    for (auto density : densities) {
+      num_priors += (fixed_ratios.size()) * (pow(density, 2));  // NOLINT
     }
     if (!flatten) {
       std::vector<int64_t> dim_vec(4);