Fix typos performace performance (#70787)

co63oc · web-flow · commit 01e22b465a0a · 2025-01-20T11:12:52.000+08:00
* Fix

* ci
diff --git a/paddle/cinn/ir/group_schedule/tactic/tile_broadcast_tactic.cc b/paddle/cinn/ir/group_schedule/tactic/tile_broadcast_tactic.cc
@@ -367,7 +367,7 @@ void TileBroadcastTactic::Apply(ir::IRSchedule* sch,
   FuseAxisGroups(sch, block_id);
 
   // Do tiling.
-  // To achieve best performace, we apply different tiling tepmlates based on
+  // To achieve best performance, we apply different tiling templates based on
   // low_broadcast_size. The key is which axis to allocate inner loop:
   // 1. For small size:
   //        [B, P, B<=256]
diff --git a/paddle/common/flags.cc b/paddle/common/flags.cc
@@ -898,7 +898,7 @@ PHI_DEFINE_EXPORTED_bool(enable_neighbor_list_use_uva,
  */
 PHI_DEFINE_EXPORTED_double(graph_neighbor_size_percent,
                            1.0,
-                           "It controls whether precent of neighbor_size.");
+                           "It controls whether percent of neighbor_size.");
 
 /**
  * Distributed related FLAG
diff --git a/paddle/fluid/framework/fleet/ps_gpu_wrapper.cc b/paddle/fluid/framework/fleet/ps_gpu_wrapper.cc
@@ -1103,7 +1103,7 @@ void PSGPUWrapper::FilterPull(std::shared_ptr<HeterContext> gpu_task,
                       nullptr,
                       common::errors::InvalidArgument(
                           "The shard values after deduplication should not "
-                          "be nullptr, but got %d at positon %d.",
+                          "be nullptr, but got %d at position %d.",
                           shard_values[dedup_size],
                           dedup_size));
     ++dedup_size;
diff --git a/paddle/fluid/pir/transforms/pd_op_to_kernel_pass.cc b/paddle/fluid/pir/transforms/pd_op_to_kernel_pass.cc
@@ -3361,7 +3361,7 @@ void ProcessBlock(
       new_arg.set_attribute(name, attr);
     }
     if (auto dense_tensor_type = arg.type().dyn_cast<DenseTensorType>()) {
-      auto place_attr = arg.attribute<PlaceAttribute>("palce");
+      auto place_attr = arg.attribute<PlaceAttribute>("place");
       auto var_place = place_attr ? place_attr.data() : phi::Place();
       new_arg.set_type(
           AllocatedDenseTensorType::get(ctx, var_place, dense_tensor_type));
@@ -3371,7 +3371,7 @@ void ProcessBlock(
   if (phi::is_accelerat_place(place)) {
     for (auto& [keyword, arg] : block->kwargs()) {
       if (auto dense_tensor_type = arg.type().dyn_cast<DenseTensorType>()) {
-        auto place_attr = arg.attribute<PlaceAttribute>("palce");
+        auto place_attr = arg.attribute<PlaceAttribute>("place");
         if (place_attr && place_attr.data() == place) continue;
         auto dtype = dense_tensor_type.dtype();
         phi::KernelKey shadow_key{paddle::experimental::get_accelerat_backend(),
diff --git a/paddle/fluid/pybind/pir.cc b/paddle/fluid/pybind/pir.cc
@@ -1445,14 +1445,14 @@ void BindValue(py::module *m) {
       .def_property(
           "place_attr",
           [](Value self) -> phi::Place {
-            auto palce_attr = self.attribute<PlaceAttribute>("place");
-            return palce_attr ? palce_attr.data() : phi::Place();
+            auto place_attr = self.attribute<PlaceAttribute>("place");
+            return place_attr ? place_attr.data() : phi::Place();
           },
           [](Value self, const phi::Place &place) {
             // auto place = CastPyArg2Place(place_obj.release().ptr(), 1);
             auto place_attr =
                 dialect::PlaceAttribute::get(pir::IrContext::Instance(), place);
-            self.set_attribute("palce", place_attr);
+            self.set_attribute("place", place_attr);
           })
       .def("initialized",
            [](Value self) {
diff --git a/paddle/phi/kernels/fusion/cutlass/memory_efficient_attention/gemm/find_default_mma.h b/paddle/phi/kernels/fusion/cutlass/memory_efficient_attention/gemm/find_default_mma.h
@@ -71,7 +71,7 @@ template <
     typename InstructionShape,
     /// Number of stages used in the pipelined mainloop
     int Stages,
-    /// Operation perfomed by GEMM
+    /// Operation performed by GEMM
     typename Operator,
     typename Enable_ = void>
 struct FindDefaultMma {
diff --git a/paddle/phi/kernels/gpu/lars_momentum_kernel.cu b/paddle/phi/kernels/gpu/lars_momentum_kernel.cu
@@ -520,10 +520,10 @@ void LarsMomentumKernel(
     /* Implementation of lars optimizer consists of following two steps:
       1. Figure out the L2 norm statistic result of grad data and param data.
       2. Update param and velocity with usage of L2 norm statistic result.
-    Step1 and step2 can be merged with api provided by nvida
+    Step1 and step2 can be merged with api provided by nvidia
       cudaLaunchCooperativeKernel:
-      - The thread quantity shall less than pyhsical SM limited threads
-      - Launche as thread-block can synchronizlly execute. */
+      - The thread quantity shall less than physical SM limited threads
+      - Launches as thread-block can synchronizlly execute. */
     cudaOccupancyMaxActiveBlocksPerMultiprocessor(
         &num_blocks_per_sm,
         MergedMomentumLarsKernel<T, MT>,
diff --git a/python/paddle/distributed/auto_parallel/sharding.py b/python/paddle/distributed/auto_parallel/sharding.py
@@ -202,7 +202,7 @@ def apply_gradients(self, params_grads):
 
             assert (
                 param_dist_attr.partial_dims == set()
-            ), f"Sharding fusion do not support parital parameter. but received {param.name} : {param}."
+            ), f"Sharding fusion do not support partial parameter. but received {param.name} : {param}."
             assert (
                 param_dist_attr.dims_mapping == grad_dist_attr.dims_mapping
             ), f"Parameter and grad should have same dims_mapping. but received name:{param.name}, parameter:{param}, grad: {grad}."
diff --git a/python/paddle/distributed/auto_parallel/static/reshard_funcs/nd_mesh_reshard_func.py b/python/paddle/distributed/auto_parallel/static/reshard_funcs/nd_mesh_reshard_func.py
@@ -177,7 +177,7 @@ def reshard(self, src_dist_attr, dst_dist_attr, src_value, dst_type):
                 if partial_dim in out_partial_status:
                     if out_partial_status[partial_dim] != partial_type:
                         raise NotImplementedError(
-                            f"Reshard tensor from one partial type {partial_type} to another parital type {out_partial_status[partial_dim]} is not supported yet."
+                            f"Reshard tensor from one partial type {partial_type} to another partial type {out_partial_status[partial_dim]} is not supported yet."
                         )
                     continue
 
diff --git a/test/cpp/pir/cinn/tile_config_performance_test.cc b/test/cpp/pir/cinn/tile_config_performance_test.cc
@@ -560,7 +560,7 @@ void TestPerformanceForTileConfig(int spatial_left_bound,
   constexpr int kMaxTestNum = 1;
   constexpr float kEarlyStopThreshold = 1.2;
   // number of nodes in cudaGraph for test, which is defined in
-  // performace_statistician.h as graph_nodes_num_. This parameter is set to
+  // performance_statistician.h as graph_nodes_num_. This parameter is set to
   // make measure results corresponding to one launch for better readability
   constexpr int kGraphNum = 25;
 
diff --git a/test/ir/pir/fused_pass/onednn/test_onednn_placement.py b/test/ir/pir/fused_pass/onednn/test_onednn_placement.py
@@ -260,7 +260,7 @@ def test_check_output(self):
 
 
 # This case is for testing layout transformation
-class TestConv2dAddPlacmentPass(PassTest):
+class TestConv2dAddPlacementPass(PassTest):
     def is_program_valid(self, program=None):
         return True
 
diff --git a/test/legacy_test/test_imperative_partitial_backward.py b/test/legacy_test/test_imperative_partitial_backward.py
@@ -20,8 +20,8 @@
 from paddle import base
 
 
-class TestImperativePartitialBackward(unittest.TestCase):
-    def test_partitial_backward(self):
+class TestImperativePartialBackward(unittest.TestCase):
+    def test_partial_backward(self):
         with base.dygraph.guard():
             x = np.random.randn(2, 4, 5).astype("float32")
             x = paddle.to_tensor(x)

Original file line number	Diff line number	Diff line change
`@@ -177,7 +177,7 @@ def reshard(self, src_dist_attr, dst_dist_attr, src_value, dst_type):`
`177`	`177`	`if partial_dim in out_partial_status:`
`178`	`178`	`if out_partial_status[partial_dim] != partial_type:`
`179`	`179`	`raise NotImplementedError(`
`180`		`- f"Reshard tensor from one partial type {partial_type} to another parital type {out_partial_status[partial_dim]} is not supported yet."`
	`180`	`+ f"Reshard tensor from one partial type {partial_type} to another partial type {out_partial_status[partial_dim]} is not supported yet."`
`181`	`181`	`)`
`182`	`182`	`continue`
`183`	`183`