[CodeStyle][Typos][D-[13-19]] Fix typos (delet, dependecies, dependecy, decprecated, derivated, descripor, deserailize)

_typos.toml

@@ -136,13 +136,6 @@ defind = 'defind'
 defeine = 'defeine'
 defition = 'defition'
 defination = 'defination'
-delet = 'delet'
-dependecies = 'dependecies'
-dependecy = 'dependecy'
-decprecated = 'decprecated'
-derivated = 'derivated'
-descripor = 'descripor'
-deserailize = 'deserailize'
 Destory = 'Destory'
 DEIVCE = 'DEIVCE'
 dictionnary = 'dictionnary'

paddle/common/flags.cc

@@ -1521,8 +1521,8 @@ PHI_DEFINE_EXPORTED_bool(use_shm_cache,
  * Since Version: 2.6.2
  * Value Range: bool, default=false
  * Example:
- * Note: . If True, mmap_allocator will use file descripor to open shared memory
- * operation.
+ * Note: . If True, mmap_allocator will use file descriptor to open shared
+ * memory operation.
  */
 PHI_DEFINE_EXPORTED_bool(dataloader_use_file_descriptor,
                          false,

paddle/fluid/eager/autograd_meta.h

@@ -56,7 +56,7 @@ using AbstractAutogradMeta = paddle::AbstractAutogradMeta;
  *
  * **/
 
-// No other AutogradMeta class should be derivated from AbstractAutogradMeta.
+// No other AutogradMeta class should be derived from AbstractAutogradMeta.
 // It's only used by
 class AutogradMeta : public AbstractAutogradMeta {
  public:

paddle/fluid/framework/ir/lock_free_optimize_pass.h

@@ -30,7 +30,7 @@ class Graph;
 
 /*
  * Remove the sum op of all gradients of the backward op.
- * And remove the dependecies of the optimizer related to the
+ * And remove the dependencies of the optimizer related to the
  * same backward op.
  *
  * Before this pass:

paddle/phi/infermeta/spmd_rules/reshape.cc

@@ -313,7 +313,7 @@ SpmdInfo ReshapeInferSpmdReverse(const DistMetaTensor& x,
   return {{x_dist_attr}, {out_dist_attr_dst}};
 }
 
-// FIXME(dev): XShape will be decprecated in the future, so we
+// FIXME(dev): XShape will be deprecated in the future, so we
 // need unify inferSpmd into ReshapeInferSpmd function.
 SpmdInfo ReshapeInferSpmdDynamic(const DistMetaTensor& x,
                                  const std::vector<int64_t>& shape) {

python/paddle/distributed/passes/auto_parallel_sharding.py

@@ -1303,7 +1303,7 @@ def _overlap_grad_comm(
                     )
                     idx += 1
 
-                # NOTE(Ruibiao): Why add dependecy here?
+                # NOTE(Ruibiao): Why add dependency here?
                 # It is hack to delay GC for coalesce_var, which significantly reduce memory usage.
                 # With the pattern of reduce_sum + scale, the coalesce_var is used by the reduce_sum
                 # op on the comm-stream, and then released by the scale op on the comp-stream. Since

python/paddle/incubate/distributed/fleet/parameter_server/distribute_transpiler/__init__.py

@@ -869,7 +869,7 @@ def _build_trainer_programs(self, compiled_config):
             # for startup program
             _startup = worker.fake_init_ops_pass(_startup, compiled_config)
             _startup = worker.init_from_server_pass(_startup, compiled_config)
-            _startup = worker.delet_extra_optimizes_pass(
+            _startup = worker.delete_extra_optimizes_pass(
                 _startup, compiled_config
             )
         else:

test/cpp/inference/api/trt_dynamic_shape_test.cc

@@ -295,7 +295,7 @@ TEST(AnalysisPredictor, trt_dynamic) { TestDynamic(true); }
 TEST(AnalysisPredictor, trt_memory_serialize) {
   // serailize
   TestDynamic(true, true, true);
-  // deserailize
+  // deserialize
   TestDynamic(true, false, true);
 }
 TEST(AnalysisPredictor, trt_dynamic2) { TestDynamic2(); }