Black 20 release

src/transformers/configuration_roberta.py

@@ -61,6 +61,5 @@ class RobertaConfig(BertConfig):
     model_type = "roberta"
 
     def __init__(self, pad_token_id=1, bos_token_id=0, eos_token_id=2, **kwargs):
-        """Constructs RobertaConfig.
-        """
+        """Constructs RobertaConfig."""
         super().__init__(pad_token_id=pad_token_id, bos_token_id=bos_token_id, eos_token_id=eos_token_id, **kwargs)

src/transformers/configuration_t5.py

@@ -80,7 +80,10 @@ class T5Config(PretrainedConfig):
         **kwargs
     ):
         super().__init__(
-            pad_token_id=pad_token_id, eos_token_id=eos_token_id, is_encoder_decoder=is_encoder_decoder, **kwargs,
+            pad_token_id=pad_token_id,
+            eos_token_id=eos_token_id,
+            is_encoder_decoder=is_encoder_decoder,
+            **kwargs,
         )
         self.vocab_size = vocab_size
         self.n_positions = n_positions

src/transformers/configuration_utils.py

@@ -29,7 +29,7 @@ logger = logging.get_logger(__name__)
 
 
 class PretrainedConfig(object):
-    r""" Base class for all configuration classes.
+    r"""Base class for all configuration classes.
     Handles a few parameters common to all models' configurations as well as methods for loading/downloading/saving
     configurations.
 

src/transformers/configuration_xlm.py

@@ -191,8 +191,7 @@ class XLMConfig(PretrainedConfig):
         bos_token_id=0,
         **kwargs
     ):
-        """Constructs XLMConfig.
-        """
+        """Constructs XLMConfig."""
         super().__init__(pad_token_id=pad_token_id, bos_token_id=bos_token_id, **kwargs)
         self.vocab_size = vocab_size
         self.emb_dim = emb_dim

src/transformers/configuration_xlnet.py

@@ -162,8 +162,7 @@ class XLNetConfig(PretrainedConfig):
         eos_token_id=2,
         **kwargs
     ):
-        """Constructs XLNetConfig.
-        """
+        """Constructs XLNetConfig."""
         super().__init__(pad_token_id=pad_token_id, bos_token_id=bos_token_id, eos_token_id=eos_token_id, **kwargs)
         self.vocab_size = vocab_size
         self.d_model = d_model

src/transformers/convert_dialogpt_original_pytorch_checkpoint_to_pytorch.py

@@ -27,5 +27,6 @@ if __name__ == "__main__":
         checkpoint_path = os.path.join(args.dialogpt_path, f"{MODEL}_ft.pkl")
         pytorch_dump_folder_path = f"./DialoGPT-{MODEL}"
         convert_dialogpt_checkpoint(
-            checkpoint_path, pytorch_dump_folder_path,
+            checkpoint_path,
+            pytorch_dump_folder_path,
         )

src/transformers/convert_graph_to_onnx.py

@@ -38,24 +38,39 @@ class OnnxConverterArgumentParser(ArgumentParser):
         super().__init__("ONNX Converter")
 
         self.add_argument(
-            "--pipeline", type=str, choices=SUPPORTED_PIPELINES, default="feature-extraction",
+            "--pipeline",
+            type=str,
+            choices=SUPPORTED_PIPELINES,
+            default="feature-extraction",
         )
         self.add_argument(
-            "--model", type=str, required=True, help="Model's id or path (ex: bert-base-cased)",
+            "--model",
+            type=str,
+            required=True,
+            help="Model's id or path (ex: bert-base-cased)",
         )
         self.add_argument("--tokenizer", type=str, help="Tokenizer's id or path (ex: bert-base-cased)")
         self.add_argument(
-            "--framework", type=str, choices=["pt", "tf"], help="Framework for loading the model",
+            "--framework",
+            type=str,
+            choices=["pt", "tf"],
+            help="Framework for loading the model",
         )
         self.add_argument("--opset", type=int, default=11, help="ONNX opset to use")
         self.add_argument(
-            "--check-loading", action="store_true", help="Check ONNX is able to load the model",
+            "--check-loading",
+            action="store_true",
+            help="Check ONNX is able to load the model",
         )
         self.add_argument(
-            "--use-external-format", action="store_true", help="Allow exporting model >= than 2Gb",
+            "--use-external-format",
+            action="store_true",
+            help="Allow exporting model >= than 2Gb",
         )
         self.add_argument(
-            "--quantize", action="store_true", help="Quantize the neural network to be run with int8",
+            "--quantize",
+            action="store_true",
+            help="Quantize the neural network to be run with int8",
         )
         self.add_argument("output")
 
@@ -376,7 +391,10 @@ def quantize(onnx_model_path: Path) -> Path:
     )
 
     quantized_model = quantize(
-        model=onnx_model, quantization_mode=QuantizationMode.IntegerOps, force_fusions=True, symmetric_weight=True,
+        model=onnx_model,
+        quantization_mode=QuantizationMode.IntegerOps,
+        force_fusions=True,
+        symmetric_weight=True,
     )
 
     # Append "-quantized" at the end of the model's name

src/transformers/convert_marian_to_pytorch.py

@@ -255,7 +255,11 @@ license: apache-2.0
 
 
 def write_model_card(
-    hf_model_name: str, repo_root="OPUS-MT-train", save_dir=Path("marian_converted"), dry_run=False, extra_metadata={},
+    hf_model_name: str,
+    repo_root="OPUS-MT-train",
+    save_dir=Path("marian_converted"),
+    dry_run=False,
+    extra_metadata={},
 ) -> str:
     """Copy the most recent model's readme section from opus, and add metadata.
     upload command: aws s3 sync model_card_dir s3://models.huggingface.co/bert/Helsinki-NLP/ --dryrun
@@ -604,7 +608,9 @@ class OpusState:
 
         assert "hidden_size" not in cfg.to_dict()
         load_layers_(
-            model.model.encoder.layers, state_dict, BART_CONVERTER,
+            model.model.encoder.layers,
+            state_dict,
+            BART_CONVERTER,
         )
         load_layers_(model.model.decoder.layers, state_dict, BART_CONVERTER, is_decoder=True)
 

src/transformers/convert_pytorch_checkpoint_to_tf2.py

@@ -108,7 +108,12 @@ if is_torch_available():
 logging.set_verbosity_info()
 
 MODEL_CLASSES = {
-    "bert": (BertConfig, TFBertForPreTraining, BertForPreTraining, BERT_PRETRAINED_CONFIG_ARCHIVE_MAP,),
+    "bert": (
+        BertConfig,
+        TFBertForPreTraining,
+        BertForPreTraining,
+        BERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
+    ),
     "bert-large-uncased-whole-word-masking-finetuned-squad": (
         BertConfig,
         TFBertForQuestionAnswering,
@@ -127,9 +132,24 @@ MODEL_CLASSES = {
         BertForSequenceClassification,
         BERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
     ),
-    "gpt2": (GPT2Config, TFGPT2LMHeadModel, GPT2LMHeadModel, GPT2_PRETRAINED_CONFIG_ARCHIVE_MAP,),
-    "xlnet": (XLNetConfig, TFXLNetLMHeadModel, XLNetLMHeadModel, XLNET_PRETRAINED_CONFIG_ARCHIVE_MAP,),
-    "xlm": (XLMConfig, TFXLMWithLMHeadModel, XLMWithLMHeadModel, XLM_PRETRAINED_CONFIG_ARCHIVE_MAP,),
+    "gpt2": (
+        GPT2Config,
+        TFGPT2LMHeadModel,
+        GPT2LMHeadModel,
+        GPT2_PRETRAINED_CONFIG_ARCHIVE_MAP,
+    ),
+    "xlnet": (
+        XLNetConfig,
+        TFXLNetLMHeadModel,
+        XLNetLMHeadModel,
+        XLNET_PRETRAINED_CONFIG_ARCHIVE_MAP,
+    ),
+    "xlm": (
+        XLMConfig,
+        TFXLMWithLMHeadModel,
+        XLMWithLMHeadModel,
+        XLM_PRETRAINED_CONFIG_ARCHIVE_MAP,
+    ),
     "xlm-roberta": (
         XLMRobertaConfig,
         TFXLMRobertaForMaskedLM,
@@ -148,7 +168,12 @@ MODEL_CLASSES = {
         OpenAIGPTLMHeadModel,
         OPENAI_GPT_PRETRAINED_CONFIG_ARCHIVE_MAP,
     ),
-    "roberta": (RobertaConfig, TFRobertaForMaskedLM, RobertaForMaskedLM, ROBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP,),
+    "roberta": (
+        RobertaConfig,
+        TFRobertaForMaskedLM,
+        RobertaForMaskedLM,
+        ROBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP,
+    ),
     "roberta-large-mnli": (
         RobertaConfig,
         TFRobertaForSequenceClassification,
@@ -179,10 +204,30 @@ MODEL_CLASSES = {
         DistilBertForQuestionAnswering,
         DISTILBERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
     ),
-    "ctrl": (CTRLConfig, TFCTRLLMHeadModel, CTRLLMHeadModel, CTRL_PRETRAINED_CONFIG_ARCHIVE_MAP,),
-    "albert": (AlbertConfig, TFAlbertForPreTraining, AlbertForPreTraining, ALBERT_PRETRAINED_CONFIG_ARCHIVE_MAP,),
-    "t5": (T5Config, TFT5ForConditionalGeneration, T5ForConditionalGeneration, T5_PRETRAINED_CONFIG_ARCHIVE_MAP,),
-    "electra": (ElectraConfig, TFElectraForPreTraining, ElectraForPreTraining, ELECTRA_PRETRAINED_CONFIG_ARCHIVE_MAP,),
+    "ctrl": (
+        CTRLConfig,
+        TFCTRLLMHeadModel,
+        CTRLLMHeadModel,
+        CTRL_PRETRAINED_CONFIG_ARCHIVE_MAP,
+    ),
+    "albert": (
+        AlbertConfig,
+        TFAlbertForPreTraining,
+        AlbertForPreTraining,
+        ALBERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
+    ),
+    "t5": (
+        T5Config,
+        TFT5ForConditionalGeneration,
+        T5ForConditionalGeneration,
+        T5_PRETRAINED_CONFIG_ARCHIVE_MAP,
+    ),
+    "electra": (
+        ElectraConfig,
+        TFElectraForPreTraining,
+        ElectraForPreTraining,
+        ELECTRA_PRETRAINED_CONFIG_ARCHIVE_MAP,
+    ),
 }
 
 

src/transformers/convert_reformer_trax_checkpoint_to_pytorch.py

@@ -49,10 +49,12 @@ def set_layer_weights_in_torch_lsh(weights, torch_layer, hidden_size):
         torch.tensor(np_query_key).transpose(1, 2).contiguous().view(-1, hidden_size),
     )
     set_param(
-        torch_layer.self_attention.value, torch.tensor(np_value).transpose(1, 2).contiguous().view(-1, hidden_size),
+        torch_layer.self_attention.value,
+        torch.tensor(np_value).transpose(1, 2).contiguous().view(-1, hidden_size),
     )
     set_param(
-        torch_layer.output.dense, torch.tensor(np_dense).view(-1, hidden_size).contiguous().transpose(0, 1),
+        torch_layer.output.dense,
+        torch.tensor(np_dense).view(-1, hidden_size).contiguous().transpose(0, 1),
     )
 
 
@@ -64,16 +66,20 @@ def set_layer_weights_in_torch_local(weights, torch_layer, hidden_size):
     np_dense = np.asarray(weights[3])
 
     set_param(
-        torch_layer.self_attention.query, torch.tensor(np_query).transpose(1, 2).contiguous().view(-1, hidden_size),
+        torch_layer.self_attention.query,
+        torch.tensor(np_query).transpose(1, 2).contiguous().view(-1, hidden_size),
     )
     set_param(
-        torch_layer.self_attention.key, torch.tensor(np_key).transpose(1, 2).contiguous().view(-1, hidden_size),
+        torch_layer.self_attention.key,
+        torch.tensor(np_key).transpose(1, 2).contiguous().view(-1, hidden_size),
     )
     set_param(
-        torch_layer.self_attention.value, torch.tensor(np_value).transpose(1, 2).contiguous().view(-1, hidden_size),
+        torch_layer.self_attention.value,
+        torch.tensor(np_value).transpose(1, 2).contiguous().view(-1, hidden_size),
     )
     set_param(
-        torch_layer.output.dense, torch.tensor(np_dense).view(-1, hidden_size).contiguous().transpose(0, 1),
+        torch_layer.output.dense,
+        torch.tensor(np_dense).view(-1, hidden_size).contiguous().transpose(0, 1),
     )
 
 
@@ -83,7 +89,9 @@ def set_block_weights_in_torch(weights, torch_block, hidden_size):
     layer_norm_1_weight = np.asarray(layer_norm_1[0])
     layer_norm_1_bias = np.asarray(layer_norm_1[1])
     set_param(
-        torch_block.attention.layer_norm, torch.tensor(layer_norm_1_weight), torch.tensor(layer_norm_1_bias),
+        torch_block.attention.layer_norm,
+        torch.tensor(layer_norm_1_weight),
+        torch.tensor(layer_norm_1_bias),
     )
 
     # lsh weights + output
@@ -104,7 +112,9 @@ def set_block_weights_in_torch(weights, torch_block, hidden_size):
     layer_norm_2_weight = np.asarray(intermediate_weights[0][0])
     layer_norm_2_bias = np.asarray(intermediate_weights[0][1])
     set_param(
-        torch_block.feed_forward.layer_norm, torch.tensor(layer_norm_2_weight), torch.tensor(layer_norm_2_bias),
+        torch_block.feed_forward.layer_norm,
+        torch.tensor(layer_norm_2_weight),
+        torch.tensor(layer_norm_2_bias),
     )
 
     # intermediate dense
@@ -133,7 +143,8 @@ def set_model_weights_in_torch(weights, torch_model, hidden_size):
     # word embeds
     word_embeddings = np.asarray(weights[1])
     set_param(
-        torch_model_reformer.embeddings.word_embeddings, torch.tensor(word_embeddings),
+        torch_model_reformer.embeddings.word_embeddings,
+        torch.tensor(word_embeddings),
     )
 
     if isinstance(weights[3], tuple):

src/transformers/data/datasets/glue.py

@@ -86,7 +86,10 @@ class GlueDataset(Dataset):
         cached_features_file = os.path.join(
             cache_dir if cache_dir is not None else args.data_dir,
             "cached_{}_{}_{}_{}".format(
-                mode.value, tokenizer.__class__.__name__, str(args.max_seq_length), args.task_name,
+                mode.value,
+                tokenizer.__class__.__name__,
+                str(args.max_seq_length),
+                args.task_name,
             ),
         )
         label_list = self.processor.get_labels()

src/transformers/data/datasets/language_modeling.py

@@ -21,7 +21,11 @@ class TextDataset(Dataset):
     """
 
     def __init__(
-        self, tokenizer: PreTrainedTokenizer, file_path: str, block_size: int, overwrite_cache=False,
+        self,
+        tokenizer: PreTrainedTokenizer,
+        file_path: str,
+        block_size: int,
+        overwrite_cache=False,
     ):
         assert os.path.isfile(file_path), f"Input file path {file_path} not found"
 
@@ -29,7 +33,12 @@ class TextDataset(Dataset):
 
         directory, filename = os.path.split(file_path)
         cached_features_file = os.path.join(
-            directory, "cached_lm_{}_{}_{}".format(tokenizer.__class__.__name__, str(block_size), filename,),
+            directory,
+            "cached_lm_{}_{}_{}".format(
+                tokenizer.__class__.__name__,
+                str(block_size),
+                filename,
+            ),
         )
 
         # Make sure only the first process in distributed training processes the dataset,

src/transformers/data/datasets/squad.py

@@ -119,7 +119,10 @@ class SquadDataset(Dataset):
         cached_features_file = os.path.join(
             cache_dir if cache_dir is not None else args.data_dir,
             "cached_{}_{}_{}_{}".format(
-                mode.value, tokenizer.__class__.__name__, str(args.max_seq_length), version_tag,
+                mode.value,
+                tokenizer.__class__.__name__,
+                str(args.max_seq_length),
+                version_tag,
             ),
         )
 

src/transformers/data/metrics/squad_metrics.py

@@ -589,7 +589,7 @@ def compute_predictions_log_probs(
     tokenizer,
     verbose_logging,
 ):
-    """ XLNet write prediction logic (more complex than Bert's).
+    """XLNet write prediction logic (more complex than Bert's).
     Write final predictions to the json file and log-odds of null if needed.
 
     Requires utils_squad_evaluate.py

src/transformers/data/processors/glue.py

@@ -69,7 +69,10 @@ def glue_convert_examples_to_features(
 if is_tf_available():
 
     def _tf_glue_convert_examples_to_features(
-        examples: tf.data.Dataset, tokenizer: PreTrainedTokenizer, task=str, max_length: Optional[int] = None,
+        examples: tf.data.Dataset,
+        tokenizer: PreTrainedTokenizer,
+        task=str,
+        max_length: Optional[int] = None,
     ) -> tf.data.Dataset:
         """
         Returns:

src/transformers/data/processors/utils.py

@@ -269,7 +269,9 @@ class SingleSentenceClassificationProcessor(DataProcessor):
                 logger.info("Tokenizing example %d", ex_index)
 
             input_ids = tokenizer.encode(
-                example.text_a, add_special_tokens=True, max_length=min(max_length, tokenizer.max_len),
+                example.text_a,
+                add_special_tokens=True,
+                max_length=min(max_length, tokenizer.max_len),
             )
             all_input_ids.append(input_ids)