update sequencesummary module

.coveragerc

@@ -3,6 +3,7 @@ source=pytorch_transformers
 omit =
     # skip convertion scripts from testing for now
     */convert_*
+    */__main__.py
 [report]
 exclude_lines =
     pragma: no cover

examples/test_examples.py

@@ -48,7 +48,7 @@ class ExamplesTests(unittest.TestCase):
         testargs = ["run_glue.py", "--data_dir=./examples/tests_samples/MRPC/",
                     "--task_name=mrpc", "--do_train", "--do_eval", "--output_dir=./examples/tests_samples/temp_dir",
                     "--train_batch_size=4", "--eval_batch_size=2", "--num_train_epochs=2.0", "--overwrite_output_dir"]
-        model_name = "--model_name=xlnet-large-cased"
+        model_name = "--model_name=bert-base-uncased"
         with patch.object(sys, 'argv', testargs + [model_name]):
             result = run_glue.main()
             for value in result.values():

pytorch_transformers/modeling_gpt2.py

@@ -119,9 +119,12 @@ class GPT2Config(PretrainedConfig):
         layer_norm_epsilon=1e-5,
         initializer_range=0.02,
         predict_special_tokens=True,
+
+        num_labels=1,
         summary_type='token_ids',
         summary_use_proj=True,
         summary_activation=None,
+        summary_proj_to_labels=True,
         summary_first_dropout=0.1,
         **kwargs
     ):
@@ -168,10 +171,13 @@ class GPT2Config(PretrainedConfig):
             self.layer_norm_epsilon = layer_norm_epsilon
             self.initializer_range = initializer_range
             self.predict_special_tokens = predict_special_tokens
+
+            self.num_labels = num_labels
             self.summary_type = summary_type
             self.summary_use_proj = summary_use_proj
             self.summary_activation = summary_activation
             self.summary_first_dropout = summary_first_dropout
+            self.summary_proj_to_labels = summary_proj_to_labels
         else:
             raise ValueError(
                 "First argument must be either a vocabulary size (int)"

pytorch_transformers/modeling_openai.py

@@ -147,9 +147,12 @@ class OpenAIGPTConfig(PretrainedConfig):
         layer_norm_epsilon=1e-5,
         initializer_range=0.02,
         predict_special_tokens=True,
+
+        num_labels=1,
         summary_type='token_ids',
         summary_use_proj=True,
         summary_activation=None,
+        summary_proj_to_labels=True,
         summary_first_dropout=0.1,
         **kwargs
     ):
@@ -199,10 +202,13 @@ class OpenAIGPTConfig(PretrainedConfig):
             self.layer_norm_epsilon = layer_norm_epsilon
             self.initializer_range = initializer_range
             self.predict_special_tokens = predict_special_tokens
+
+            self.num_labels = num_labels
             self.summary_type = summary_type
             self.summary_use_proj = summary_use_proj
             self.summary_activation = summary_activation
             self.summary_first_dropout = summary_first_dropout
+            self.summary_proj_to_labels = summary_proj_to_labels
         else:
             raise ValueError(
                 "First argument must be either a vocabulary size (int)"