Fix _check_vaild_topo _check_valid_topo (#61628)

co63oc · web-flow · commit 4709dc313378 · 2024-02-07T12:32:35.000+08:00
diff --git a/python/paddle/distributed/fleet/base/topology.py b/python/paddle/distributed/fleet/base/topology.py
@@ -190,7 +190,7 @@ def __init__(self, topology):
         self.stage_id = self._get_pipe_parallel_id()
 
         assert (
-            self._check_vaild_topo()
+            self._check_valid_topo()
         ), "mp_num: {}, sharding_num: {}, pp_num: {}, dp_num: {}, sep_num: {}".format(
             self.nranks,
             self._mp_degree,
@@ -328,7 +328,7 @@ def get_parallel_mode(self):
             # pp may coexist with mp、sep、dp and sharding
             return ParallelMode.PIPELINE_PARALLEL
 
-    def _check_vaild_topo(self):
+    def _check_valid_topo(self):
         return (
             self._dp_degree
             * self._mp_degree
diff --git a/python/paddle/distributed/fleet/meta_parallel/pipeline_parallel.py b/python/paddle/distributed/fleet/meta_parallel/pipeline_parallel.py
@@ -118,7 +118,7 @@ def _load_micro_batch_impl(self, inputs, micro_step):
                         else None
                     )
                 elif data is not None:
-                    self._check_data_vaild(data)
+                    self._check_data_valid(data)
                     output.append(data[begin:end, :].detach())
                 else:
                     output.append(None)
@@ -133,12 +133,12 @@ def _load_micro_batch_impl(self, inputs, micro_step):
             )
             return inputs[micro_step].detach()
         elif inputs is not None:
-            self._check_data_vaild(inputs)
+            self._check_data_valid(inputs)
             return inputs[begin:end, :].detach()
         else:
             return None
 
-    def _check_data_vaild(self, data):
+    def _check_data_valid(self, data):
         batch_size = data.shape[0]
         assert self._micro_batch_size * self._acc_steps == batch_size, (
             "batch_size needs to be divisible by micro_batch_size. Currently, "
@@ -818,7 +818,7 @@ def _broadcast_final_loss(self):
         if self.is_pipeline_last_stage(ignore_virtual=True):
             assert (
                 self.total_loss is not None
-            ), "train_batch() in last stage should obtain vaild loss"
+            ), "train_batch() in last stage should obtain valid loss"
             loss = (
                 self.total_loss.detach()
                 if not self._delay_scale_loss
diff --git a/python/paddle/distributed/fleet/recompute/recompute.py b/python/paddle/distributed/fleet/recompute/recompute.py
@@ -191,7 +191,7 @@ def forward(ctx, run_function, preserve_rng_state, *args, **kwargs):
     @staticmethod
     def backward(ctx, *args):
         with paddle.base.dygraph.guard():
-            # TODO need to check the recompute calling is vaild or not
+            # TODO need to check the recompute calling is valid or not
 
             # Restore inputs
             inputs = list(ctx.inputs)
diff --git a/python/paddle/optimizer/adam.py b/python/paddle/optimizer/adam.py
@@ -188,17 +188,17 @@ def __init__(
         if not isinstance(beta1, (Variable, Value)):
             if not 0 <= beta1 < 1:
                 raise ValueError(
-                    "Invaild value of beta1, expect beta1 in [0,1)."
+                    "Invalid value of beta1, expect beta1 in [0,1)."
                 )
         if not isinstance(beta2, (Variable, Value)):
             if not 0 <= beta2 < 1:
                 raise ValueError(
-                    "Invaild value of beta2, expect beta2 in [0,1)."
+                    "Invalid value of beta2, expect beta2 in [0,1)."
                 )
         if not isinstance(epsilon, (Variable, Value)):
             if not 0 <= epsilon:
                 raise ValueError(
-                    "Invaild value of epsilon, expect epsilon >= 0."
+                    "Invalid value of epsilon, expect epsilon >= 0."
                 )
         super().__init__(
             learning_rate=learning_rate,
diff --git a/python/paddle/optimizer/adamax.py b/python/paddle/optimizer/adamax.py
@@ -152,11 +152,11 @@ def __init__(
         assert beta2 is not None
         assert epsilon is not None
         if not 0 <= beta1 < 1:
-            raise ValueError("Invaild value of beta1, expect beta1 in [0,1).")
+            raise ValueError("Invalid value of beta1, expect beta1 in [0,1).")
         if not 0 <= beta2 < 1:
-            raise ValueError("Invaild value of beta2, expect beta2 in [0,1).")
+            raise ValueError("Invalid value of beta2, expect beta2 in [0,1).")
         if not 0 <= epsilon:
-            raise ValueError("Invaild value of epsilon, expect epsilon >= 0.")
+            raise ValueError("Invalid value of epsilon, expect epsilon >= 0.")
         super().__init__(
             learning_rate=learning_rate,
             parameters=parameters,
diff --git a/python/paddle/optimizer/adamw.py b/python/paddle/optimizer/adamw.py
@@ -176,11 +176,11 @@ def __init__(
         assert beta2 is not None
         assert epsilon is not None
         if not isinstance(beta1, Value) and not 0 <= beta1 < 1:
-            raise ValueError("Invaild value of beta1, expect beta1 in [0,1).")
+            raise ValueError("Invalid value of beta1, expect beta1 in [0,1).")
         if not isinstance(beta1, Value) and not 0 <= beta2 < 1:
-            raise ValueError("Invaild value of beta2, expect beta2 in [0,1).")
+            raise ValueError("Invalid value of beta2, expect beta2 in [0,1).")
         if not isinstance(beta1, Value) and not 0 <= epsilon:
-            raise ValueError("Invaild value of epsilon, expect epsilon >= 0.")
+            raise ValueError("Invalid value of epsilon, expect epsilon >= 0.")
         if not isinstance(weight_decay, float) and not isinstance(
             weight_decay, (framework.Variable, Value)
         ):
