Fix checkponit checkpoint, etc (#62168)

co63oc · web-flow · commit f471aa136bdf · 2024-02-29T23:37:56.000+08:00
diff --git a/paddle/fluid/operators/activation_op.cc b/paddle/fluid/operators/activation_op.cc
@@ -94,7 +94,7 @@ class ActivationGradOpMaker : public framework::SingleGradOpMaker<T> {
 //     paddle::Tensor dx = this->GetSingleInputGrad("X");
 //     auto* dx_ptr = this->GetOutputPtr(&dx);
 //     std::string dx_name = this->GetOutputName(dx);
-//     VLOG(6) << "Runing hardswish_grad composite func";
+//     VLOG(6) << "Running hardswish_grad composite func";
 //     prim::hardswish_grad<prim::DescTensor>(x, out_grad, dx_ptr);
 //     this->RecoverOutputName(dx, dx_name);
 //   }
@@ -394,19 +394,19 @@ REGISTER_ACTIVATION_OP(mish, Mish, MishFunctor, MishGradFunctor);
 /* ==========================  register checkpoint ===========================*/
 REGISTER_OP_VERSION(leaky_relu)
     .AddCheckpoint(
-        R"ROC(fix leaky_relu, bahavior changed when alpha < 0 or alpha > 1)ROC",
+        R"ROC(fix leaky_relu, behavior changed when alpha < 0 or alpha > 1)ROC",
         paddle::framework::compatible::OpVersionDesc()
             .BugfixWithBehaviorChanged(
-                "leaky_relu calculate formula before checkponit: out = max(x, "
+                "leaky_relu calculate formula before checkpoint: out = max(x, "
                 "alpha * x); after checkpoint: out = x if x > 0 else alpha * "
                 "x"));
 
 REGISTER_OP_VERSION(hard_shrink)
     .AddCheckpoint(
-        R"ROC(fix hard_shrink, bahavior changed when threshold<0)ROC",
+        R"ROC(fix hard_shrink, behavior changed when threshold<0)ROC",
         paddle::framework::compatible::OpVersionDesc()
             .BugfixWithBehaviorChanged(
-                "hard_shrink calculate formula before checkponit: out = x * "
+                "hard_shrink calculate formula before checkpoint: out = x * "
                 "((x < -threshold) + (x > threshold)); after checkpoint: out = "
                 "x * (((x < -threshold) + (x > threshold)) > 0)"));
 
diff --git a/paddle/fluid/operators/activation_op.h b/paddle/fluid/operators/activation_op.h
@@ -371,7 +371,7 @@ struct AbsGradGradFunctor : public BaseActivationFunctor<T> {
 
 // TODO(dengkaipeng): double gradient calculation for Square/Sqrt need
 // DOut(dy) as input(not output), tensor extraction is different from
-// others. Impliment extraction kernel separately here.
+// others. Implement extraction kernel separately here.
 inline void ExtractDoubleGradTensorWithInputDOut(
     const framework::ExecutionContext& ctx,
     const phi::DenseTensor** X,
diff --git a/paddle/fluid/operators/assign_value_op.h b/paddle/fluid/operators/assign_value_op.h
@@ -29,7 +29,7 @@ typename std::enable_if<std::is_same<T, bool>::value>::type CopyVectorToTensor(
     const char* value_name,
     phi::DenseTensor* out,
     const framework::ExecutionContext& ctx) {
-  // phi::DenseTensore dtype is vector<bool>, it will be converted to
+  // phi::DenseTensor dtype is vector<bool>, it will be converted to
   //  vector<int>.
   //  at the same time, we can not use vector<bool> to hold the value, because
   //  the c++ use bit value to replace byte value.
diff --git a/paddle/fluid/operators/attention_lstm_op.cc b/paddle/fluid/operators/attention_lstm_op.cc
@@ -488,7 +488,7 @@ class AttentionLSTMKernel : public framework::OpKernel<T> {
 
         // gate act: sigmoid
         act_gate(D3, lstm_out_data, lstm_out_data);
-        // candicate act: tanh
+        // candidate act: tanh
         act_cand(D, lstm_out_data + D3, lstm_out_data + D3);
 
         // a = forget * prev_cell
diff --git a/paddle/fluid/operators/batch_norm_op.cc b/paddle/fluid/operators/batch_norm_op.cc
@@ -308,11 +308,11 @@ void BatchNormOpMaker::Make() {
                 "to true or is_test true. the behavior is equivalent. "
                 "In train mode, when setting use_global_stats True, the "
                 "global mean and variance are also used during train time, "
-                "the BN acts as scaling and shiffting.")
+                "the BN acts as scaling and shifting.")
       .SetDefault(false);
   AddAttr<bool>("trainable_statistics",
                 "(bool, default false) Whether to calculate mean and variance "
-                "in test mode. If setting true in test mode, mean and variace "
+                "in test mode. If setting true in test mode, mean and variance "
                 "will be calculated by current batch statistics.")
       .SetDefault(false);
   AddComment(R"DOC(
@@ -586,7 +586,7 @@ class BatchNormCompositeGradOpMaker : public prim::CompositeGradOpMakerBase {
     auto use_global_stats = this->Attr<bool>("use_global_stats");
     auto trainable_statistics = this->Attr<bool>("trainable_statistics");
 
-    VLOG(3) << "Runing batch_norm composite func";
+    VLOG(3) << "Running batch_norm composite func";
     prim::batch_norm_grad<prim::DescTensor>(x,
                                             scale,
                                             bias,
diff --git a/paddle/fluid/operators/beam_search_decode_op_def.h b/paddle/fluid/operators/beam_search_decode_op_def.h
@@ -27,7 +27,7 @@ using LoDTensorArray = framework::LoDTensorArray;
 
 // all the lod have 2 levels.
 // The first is source level, the second is sentence level.
-// source level describe how many prefixes (branchs) for each source sentence
+// source level describe how many prefixes (branches) for each source sentence
 // (beam). sentence level describe how these candidates belong to the prefixes.
 const size_t kSourceLevel = 0;
 const size_t kSentenceLevel = 1;
diff --git a/paddle/fluid/operators/chunk_eval_op.h b/paddle/fluid/operators/chunk_eval_op.h
@@ -199,7 +199,7 @@ class ChunkEvalKernel : public framework::OpKernel<T> {
     const int64_t* inference_data = inference->data<int64_t>();
     const int64_t* label_data = label->data<int64_t>();
     T* precision_data = precision->mutable_data<T>(place);
-    T* racall_data = recall->mutable_data<T>(place);
+    T* recall_data = recall->mutable_data<T>(place);
     T* f1_data = f1->mutable_data<T>(place);
     int64_t* num_infer_chunks_data =
         num_infer_chunks->mutable_data<int64_t>(place);
@@ -280,14 +280,14 @@ class ChunkEvalKernel : public framework::OpKernel<T> {
                           ? 0
                           : static_cast<T>(*num_correct_chunks_data) /
                                 (*num_infer_chunks_data);
-    *racall_data = !(*num_label_chunks_data)
+    *recall_data = !(*num_label_chunks_data)
                        ? 0
                        : static_cast<T>(*num_correct_chunks_data) /
                              (*num_label_chunks_data);
     *f1_data = !(*num_correct_chunks_data)
                    ? 0
-                   : 2 * (*precision_data) * (*racall_data) /
-                         ((*precision_data) + (*racall_data));
+                   : 2 * (*precision_data) * (*recall_data) /
+                         ((*precision_data) + (*recall_data));
   }
 
   void EvalOneSeq(const int64_t* output,
