[tests] remove T5 test skip decorator

tests/test_fx/test_pipeline/test_hf_model/test_t5.py

@@ -2,20 +2,28 @@
 import transformers
 import torch
 from hf_utils import split_model_and_compare_output
+from colossalai.fx.tracer.meta_patch import meta_patched_module
+try:
+    import apex
+
+    @meta_patched_module.register(apex.normalization.FusedRMSNorm)
+    def apex_fused_layernorm(self, input):
+        return torch.empty(input.shape, device='meta')
+except ImportError:
+    pass
 
 BATCH_SIZE = 1
 SEQ_LENGHT = 16
 
 
-@pytest.mark.skip('tracing failed')
 def test_t5():
     MODEL_LIST = [
         transformers.T5Model,
         transformers.T5ForConditionalGeneration,
         transformers.T5EncoderModel,
     ]
 
-    config = transformers.T5Config(d_model=128, num_layers=2)
+    config = transformers.T5Config(vocab_size=100, d_model=128, num_layers=2)
 
     def data_gen():
         input_ids = torch.zeros((BATCH_SIZE, SEQ_LENGHT), dtype=torch.int64)