[CI] fix typo with tests components

op_builder/utils.py

@@ -36,7 +36,7 @@ def get_cuda_version_in_pytorch() -> List[int]:
         torch_cuda_minor = torch.version.cuda.split(".")[1]
     except:
         raise ValueError(
-            "[extension] Cannot retrive the CUDA version in the PyTorch binary given by torch.version.cuda")
+            "[extension] Cannot retrieve the CUDA version in the PyTorch binary given by torch.version.cuda")
     return torch_cuda_major, torch_cuda_minor
 
 

tests/components_to_test/albert.py

@@ -28,7 +28,7 @@ def bert_model_builder(checkpoint: bool = False):
         print('building AlbertForSequenceClassification model')
 
         # adapting huggingface BertForSequenceClassification for single unitest calling interface
-        class ModelAaptor(AlbertForSequenceClassification):
+        class ModelAdaptor(AlbertForSequenceClassification):
 
             def forward(self, input_ids, labels):
                 """
@@ -37,23 +37,23 @@ def forward(self, input_ids, labels):
                 """
                 return super().forward(input_ids=input_ids, labels=labels)[0]
 
-        model = ModelAaptor(config)
+        model = ModelAdaptor(config)
         # if checkpoint and version.parse(transformers.__version__) >= version.parse("4.11.0"):
         #     model.gradient_checkpointing_enable()
 
         return model
 
-    is_distrbuted = torch.distributed.is_initialized()
+    is_distributed = torch.distributed.is_initialized()
     trainloader = get_bert_data_loader(n_class=vocab_size,
                                        batch_size=2,
                                        total_samples=10000,
                                        sequence_length=sequence_length,
-                                       is_distrbuted=is_distrbuted)
+                                       is_distributed=is_distributed)
     testloader = get_bert_data_loader(n_class=vocab_size,
                                       batch_size=2,
                                       total_samples=10000,
                                       sequence_length=sequence_length,
-                                      is_distrbuted=is_distrbuted)
+                                      is_distributed=is_distributed)
 
     criterion = None
     return bert_model_builder, trainloader, testloader, torch.optim.Adam, criterion

tests/components_to_test/beit.py

@@ -27,7 +27,7 @@ def generate(self):
 @non_distributed_component_funcs.register(name='beit')
 def get_training_components():
 
-    def model_buider(checkpoint=False):
+    def model_builder(checkpoint=False):
         model = Beit(img_size=DummyDataLoader.img_size,
                      num_classes=DummyDataLoader.num_class,
                      embed_dim=32,
@@ -39,4 +39,4 @@ def model_buider(checkpoint=False):
     testloader = DummyDataLoader()
 
     criterion = torch.nn.CrossEntropyLoss()
-    return model_buider, trainloader, testloader, torch.optim.Adam, criterion
+    return model_builder, trainloader, testloader, torch.optim.Adam, criterion

tests/components_to_test/bert.py

@@ -13,7 +13,7 @@ def get_bert_data_loader(
         total_samples,
         sequence_length,
         device=torch.device('cpu:0'),
-        is_distrbuted=False,
+        is_distributed=False,
 ):
     train_data = torch.randint(
         low=0,
@@ -24,7 +24,7 @@ def get_bert_data_loader(
     )
     train_label = torch.randint(low=0, high=2, size=(total_samples,), device=device, dtype=torch.long)
     train_dataset = torch.utils.data.TensorDataset(train_data, train_label)
-    if is_distrbuted:
+    if is_distributed:
         sampler = torch.utils.data.distributed.DistributedSampler(train_dataset)
     else:
         sampler = SequentialSampler(train_dataset)
@@ -52,8 +52,8 @@ def bert_model_builder(checkpoint: bool = False):
                             attention_probs_dropout_prob=0.)
         print('building BertForSequenceClassification model')
 
-        # adapting huggingface BertForSequenceClassification for single unitest calling interface
-        class ModelAaptor(BertForSequenceClassification):
+        # adapting huggingface BertForSequenceClassification for single unittest calling interface
+        class ModelAdaptor(BertForSequenceClassification):
 
             def forward(self, input_ids, labels):
                 """
@@ -62,23 +62,23 @@ def forward(self, input_ids, labels):
                 """
                 return super().forward(input_ids=input_ids, labels=labels)[0]
 
-        model = ModelAaptor(config)
+        model = ModelAdaptor(config)
         if checkpoint and version.parse(transformers.__version__) >= version.parse("4.11.0"):
             model.gradient_checkpointing_enable()
 
         return model
 
-    is_distrbuted = torch.distributed.is_initialized()
+    is_distributed = torch.distributed.is_initialized()
     trainloader = get_bert_data_loader(n_class=vocab_size,
                                        batch_size=2,
                                        total_samples=10000,
                                        sequence_length=sequence_length,
-                                       is_distrbuted=is_distrbuted)
+                                       is_distributed=is_distributed)
     testloader = get_bert_data_loader(n_class=vocab_size,
                                       batch_size=2,
                                       total_samples=10000,
                                       sequence_length=sequence_length,
-                                      is_distrbuted=is_distrbuted)
+                                      is_distributed=is_distributed)
 
     criterion = None
     return bert_model_builder, trainloader, testloader, torch.optim.Adam, criterion

tests/components_to_test/registry.py

@@ -9,10 +9,10 @@ def __init__(self):
     def register(self, name):
         assert name not in self._registry
 
-        def _regsiter(callable_):
+        def _register(callable_):
             self._registry[name] = callable_
 
-        return _regsiter
+        return _register
 
     def get_callable(self, name: str):
         return self._registry[name]
@@ -34,6 +34,6 @@ def __next__(self):
 
 
 non_distributed_component_funcs = Registry()
-model_paralle_component_funcs = Registry()
+model_parallel_component_funcs = Registry()
 
-__all__ = ['non_distributed_component_funcs', 'model_paralle_component_funcs']
+__all__ = ['non_distributed_component_funcs', 'model_parallel_component_funcs']