[transformer][pipeline parallel] fix typo in test (#1370)

* fix typo

* Update test_pipeline_parallel_fwd_bwd.py

tests/L0/run_transformer/test_pipeline_parallel_fwd_bwd.py

@@ -45,7 +45,7 @@ class PipelineParallelForwardBackwardTest(DistributedTestBase):
         forward_only: bool,
         fwd_bwd_func: FwdStepFunc,
         pipeline_model_parallel_world_size: Optional[int],
-        vriatual_pipeline_model_parallel_size: Optional[int],
+        virtual_pipeline_model_parallel_size: Optional[int],
     ) -> None:
         for dtype, deallocate_pipeline_outputs in itertools.product(
             [torch.float32] + _get_autocast_dtypes(), (True, False),
@@ -67,7 +67,7 @@ class PipelineParallelForwardBackwardTest(DistributedTestBase):
             parallel_state.initialize_model_parallel(
                 tensor_model_parallel_size_=tensor_model_parallel_world_size,
                 pipeline_model_parallel_size_=pipeline_model_parallel_world_size,
-                virtual_pipeline_model_parallel_size_=vriatual_pipeline_model_parallel_size,
+                virtual_pipeline_model_parallel_size_=virtual_pipeline_model_parallel_size,
             )
             pp_utils._reconfigure_microbatch_calculator(
                 rank=parallel_state.get_tensor_model_parallel_rank(),
@@ -88,7 +88,7 @@ class PipelineParallelForwardBackwardTest(DistributedTestBase):
             model = build_model(
                 testing_utils.model_provider_func,
                 wrap_with_ddp=True,
-                virtual_pipeline_model_parallel_size=vriatual_pipeline_model_parallel_size,
+                virtual_pipeline_model_parallel_size=virtual_pipeline_model_parallel_size,
                 hidden_size=PipelineParallelForwardBackwardTest.HIDDEN_SIZE,
             )
             _param_groups = _get_params_for_weight_decay_optimization(model)