make style

* trim seq_len below 1024 if there are columns full of pad_token_id
* Centralize trim_batch so SummarizationDataset can use it too

.circleci/config.yml

@@ -85,6 +85,8 @@ jobs:
         parallelism: 1
         steps:
             - checkout
+            # we need a version of isort with https://github.com/timothycrosley/isort/pull/1000
+            - run: sudo pip install git+git://github.com/timothycrosley/isort.git@e63ae06ec7d70b06df9e528357650281a3d3ec22#egg=isort
             - run: sudo pip install .[tf,torch,quality]
             - run: black --check --line-length 119 --target-version py35 examples templates tests src utils
             - run: isort --check-only --recursive examples templates tests src utils

docs/source/conf.py

@@ -26,7 +26,7 @@ author = u'huggingface'
 # The short X.Y version
 version = u''
 # The full version, including alpha/beta/rc tags
-release = u'2.5.1'
+release = u'2.6.0'
 
 
 # -- General configuration ---------------------------------------------------

docs/source/index.rst

@@ -103,3 +103,4 @@ The library currently contains PyTorch and Tensorflow implementations, pre-train
     model_doc/xlmroberta
     model_doc/flaubert
     model_doc/bart
+    model_doc/t5

docs/source/model_doc/t5.rst

@@ -0,0 +1,69 @@
+T5
+----------------------------------------------------
+**DISCLAIMER:** This model is still a work in progress, if you see something strange,
+file a `Github Issue <https://github.com/huggingface/transformers/issues/new?assignees=&labels=&template=bug-report.md&title>`_
+
+Overview
+~~~~~
+The T5 model was presented in `Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer <https://arxiv.org/pdf/1910.10683.pdf>`_ by Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, Peter J. Liu in 
+Here the abstract: 
+
+*Transfer learning, where a model is first pre-trained on a data-rich task before being fine-tuned on a downstream task, has emerged as a powerful technique in natural language processing (NLP). The effectiveness of transfer learning has given rise to a diversity of approaches, methodology, and practice. 
+In this paper, we explore the landscape of transfer learning techniques for NLP by introducing a unified framework that converts every language problem into a text-to-text format. 
+Our systematic study compares pre-training objectives, architectures, unlabeled datasets, transfer approaches, and other factors on dozens of language understanding tasks. 
+By combining the insights from our exploration with scale and our new "Colossal Clean Crawled Corpus", we achieve state-of-the-art results on many benchmarks covering summarization, question answering, text classification, and more. 
+To facilitate future work on transfer learning for NLP, we release our dataset, pre-trained models, and code.*
+
+The Authors' code can be found `here <https://github.com/google-research/text-to-text-transfer-transformer>`_ .
+
+Tips
+~~~~~~~~~~~~~~~~~~~~
+- T5 is an encoder-decoder model pre-trained on a multi-task mixture of unsupervised 
+  and supervised tasks and which each task is cast as a sequence to sequence task. 
+  Therefore T5 works well on a variety of tasks out-of-the-box by prepending a different prefix to the input corresponding to each task, e.g.: for translation: *translate English to German: ..., summarize: ...*.
+  For more information about the which prefix to use, it is easiest to look into Appendix D of the `paper <https://arxiv.org/pdf/1910.10683.pdf>`_ .
+- For sequence to sequence generation, it is recommended to use ``T5ForConditionalGeneration.generate()``. The method takes care of feeding the encoded input via cross-attention layers to the decoder and auto-regressively generating the decoder output.
+- T5 uses relative scalar embeddings. Encoder input padding can be done on the left and on the right.
+
+
+T5Config
+~~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: transformers.T5Config
+    :members:
+
+
+T5Tokenizer
+~~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: transformers.T5Tokenizer
+    :members: build_inputs_with_special_tokens, get_special_tokens_mask,
+        create_token_type_ids_from_sequences, save_vocabulary
+
+
+T5Model
+~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: transformers.T5Model
+    :members:
+
+
+T5ForConditionalGeneration
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: transformers.T5ForConditionalGeneration
+    :members:
+
+
+TFBertModel
+~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: transformers.TFT5Model
+    :members:
+
+
+TFT5ForConditionalGeneration
+~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: transformers.TFT5ForConditionalGeneration
+    :members:

examples/ner/run_ner.py

@@ -31,6 +31,7 @@ from torch.utils.data.distributed import DistributedSampler
 from tqdm import tqdm, trange
 
 from transformers import (
+    MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING,
     WEIGHTS_NAME,
     AdamW,
     AutoConfig,
@@ -38,7 +39,6 @@ from transformers import (
     AutoTokenizer,
     get_linear_schedule_with_warmup,
 )
-from transformers.modeling_auto import MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING
 from utils_ner import convert_examples_to_features, get_labels, read_examples_from_file
 
 
@@ -52,6 +52,7 @@ logger = logging.getLogger(__name__)
 
 MODEL_CONFIG_CLASSES = list(MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING.keys())
 MODEL_TYPES = tuple(conf.model_type for conf in MODEL_CONFIG_CLASSES)
+
 ALL_MODELS = sum((tuple(conf.pretrained_config_archive_map.keys()) for conf in MODEL_CONFIG_CLASSES), ())
 
 TOKENIZER_ARGS = ["do_lower_case", "strip_accents", "keep_accents", "use_fast"]

examples/ner/run_tf_ner.py

@@ -13,16 +13,11 @@ from seqeval import metrics
 
 from transformers import (
     TF2_WEIGHTS_NAME,
-    BertConfig,
-    BertTokenizer,
-    DistilBertConfig,
-    DistilBertTokenizer,
+    TF_MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING,
+    AutoConfig,
+    AutoTokenizer,
     GradientAccumulator,
-    RobertaConfig,
-    RobertaTokenizer,
-    TFBertForTokenClassification,
-    TFDistilBertForTokenClassification,
-    TFRobertaForTokenClassification,
+    TFAutoModelForTokenClassification,
     create_optimizer,
 )
 from utils_ner import convert_examples_to_features, get_labels, read_examples_from_file
@@ -34,22 +29,17 @@ except ImportError:
     from fastprogress.fastprogress import master_bar, progress_bar
 
 
-ALL_MODELS = sum(
-    (tuple(conf.pretrained_config_archive_map.keys()) for conf in (BertConfig, RobertaConfig, DistilBertConfig)), ()
-)
+MODEL_CONFIG_CLASSES = list(TF_MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING.keys())
+MODEL_TYPES = tuple(conf.model_type for conf in MODEL_CONFIG_CLASSES)
 
-MODEL_CLASSES = {
-    "bert": (BertConfig, TFBertForTokenClassification, BertTokenizer),
-    "roberta": (RobertaConfig, TFRobertaForTokenClassification, RobertaTokenizer),
-    "distilbert": (DistilBertConfig, TFDistilBertForTokenClassification, DistilBertTokenizer),
-}
+ALL_MODELS = sum((tuple(conf.pretrained_config_archive_map.keys()) for conf in MODEL_CONFIG_CLASSES), (),)
 
 
 flags.DEFINE_string(
     "data_dir", None, "The input data dir. Should contain the .conll files (or other data files) " "for the task."
 )
 
-flags.DEFINE_string("model_type", None, "Model type selected in the list: " + ", ".join(MODEL_CLASSES.keys()))
+flags.DEFINE_string("model_type", None, "Model type selected in the list: " + ", ".join(MODEL_TYPES))
 
 flags.DEFINE_string(
     "model_name_or_path",
@@ -509,8 +499,7 @@ def main(_):
     labels = get_labels(args["labels"])
     num_labels = len(labels) + 1
     pad_token_label_id = 0
-    config_class, model_class, tokenizer_class = MODEL_CLASSES[args["model_type"]]
-    config = config_class.from_pretrained(
+    config = AutoConfig.from_pretrained(
         args["config_name"] if args["config_name"] else args["model_name_or_path"],
         num_labels=num_labels,
         cache_dir=args["cache_dir"] if args["cache_dir"] else None,
@@ -520,14 +509,14 @@ def main(_):
 
     # Training
     if args["do_train"]:
-        tokenizer = tokenizer_class.from_pretrained(
+        tokenizer = AutoTokenizer.from_pretrained(
             args["tokenizer_name"] if args["tokenizer_name"] else args["model_name_or_path"],
             do_lower_case=args["do_lower_case"],
             cache_dir=args["cache_dir"] if args["cache_dir"] else None,
         )
 
         with strategy.scope():
-            model = model_class.from_pretrained(
+            model = TFAutoModelForTokenClassification.from_pretrained(
                 args["model_name_or_path"],
                 from_pt=bool(".bin" in args["model_name_or_path"]),
                 config=config,
@@ -562,7 +551,7 @@ def main(_):
 
     # Evaluation
     if args["do_eval"]:
-        tokenizer = tokenizer_class.from_pretrained(args["output_dir"], do_lower_case=args["do_lower_case"])
+        tokenizer = AutoTokenizer.from_pretrained(args["output_dir"], do_lower_case=args["do_lower_case"])
         checkpoints = []
         results = []
 
@@ -584,7 +573,7 @@ def main(_):
             global_step = checkpoint.split("-")[-1] if re.match(".*checkpoint-[0-9]", checkpoint) else "final"
 
             with strategy.scope():
-                model = model_class.from_pretrained(checkpoint)
+                model = TFAutoModelForTokenClassification.from_pretrained(checkpoint)
 
             y_true, y_pred, eval_loss = evaluate(
                 args, strategy, model, tokenizer, labels, pad_token_label_id, mode="dev"
@@ -611,8 +600,8 @@ def main(_):
                         writer.write("\n")
 
     if args["do_predict"]:
-        tokenizer = tokenizer_class.from_pretrained(args["output_dir"], do_lower_case=args["do_lower_case"])
-        model = model_class.from_pretrained(args["output_dir"])
+        tokenizer = AutoTokenizer.from_pretrained(args["output_dir"], do_lower_case=args["do_lower_case"])
+        model = TFAutoModelForTokenClassification.from_pretrained(args["output_dir"])
         eval_batch_size = args["per_device_eval_batch_size"] * args["n_device"]
         predict_dataset, _ = load_and_cache_examples(
             args, tokenizer, labels, pad_token_label_id, eval_batch_size, mode="test"

examples/requirements.txt

@@ -3,3 +3,6 @@ tensorboard
 scikit-learn
 seqeval
 psutil
+sacrebleu
+rouge-score
+tensorflow_datasets

examples/run_bart_prof.sh

@@ -1 +0,0 @@
-python benchmarks.py --models bart-large-cnn --batch_sizes 2 --torch

examples/run_glue.py

@@ -30,32 +30,12 @@ from torch.utils.data.distributed import DistributedSampler
 from tqdm import tqdm, trange
 
 from transformers import (
+    MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING,
     WEIGHTS_NAME,
     AdamW,
-    AlbertConfig,
-    AlbertForSequenceClassification,
-    AlbertTokenizer,
-    BertConfig,
-    BertForSequenceClassification,
-    BertTokenizer,
-    DistilBertConfig,
-    DistilBertForSequenceClassification,
-    DistilBertTokenizer,
-    FlaubertConfig,
-    FlaubertForSequenceClassification,
-    FlaubertTokenizer,
-    RobertaConfig,
-    RobertaForSequenceClassification,
-    RobertaTokenizer,
-    XLMConfig,
-    XLMForSequenceClassification,
-    XLMRobertaConfig,
-    XLMRobertaForSequenceClassification,
-    XLMRobertaTokenizer,
-    XLMTokenizer,
-    XLNetConfig,
-    XLNetForSequenceClassification,
-    XLNetTokenizer,
+    AutoConfig,
+    AutoModelForSequenceClassification,
+    AutoTokenizer,
     get_linear_schedule_with_warmup,
 )
 from transformers import glue_compute_metrics as compute_metrics
@@ -72,33 +52,10 @@ except ImportError:
 
 logger = logging.getLogger(__name__)
 
-ALL_MODELS = sum(
-    (
-        tuple(conf.pretrained_config_archive_map.keys())
-        for conf in (
-            BertConfig,
-            XLNetConfig,
-            XLMConfig,
-            RobertaConfig,
-            DistilBertConfig,
-            AlbertConfig,
-            XLMRobertaConfig,
-            FlaubertConfig,
-        )
-    ),
-    (),
-)
+MODEL_CONFIG_CLASSES = list(MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING.keys())
+MODEL_TYPES = tuple(conf.model_type for conf in MODEL_CONFIG_CLASSES)
 
-MODEL_CLASSES = {
-    "bert": (BertConfig, BertForSequenceClassification, BertTokenizer),
-    "xlnet": (XLNetConfig, XLNetForSequenceClassification, XLNetTokenizer),
-    "xlm": (XLMConfig, XLMForSequenceClassification, XLMTokenizer),
-    "roberta": (RobertaConfig, RobertaForSequenceClassification, RobertaTokenizer),
-    "distilbert": (DistilBertConfig, DistilBertForSequenceClassification, DistilBertTokenizer),
-    "albert": (AlbertConfig, AlbertForSequenceClassification, AlbertTokenizer),
-    "xlmroberta": (XLMRobertaConfig, XLMRobertaForSequenceClassification, XLMRobertaTokenizer),
-    "flaubert": (FlaubertConfig, FlaubertForSequenceClassification, FlaubertTokenizer),
-}
+ALL_MODELS = sum((tuple(conf.pretrained_config_archive_map.keys()) for conf in MODEL_CONFIG_CLASSES), (),)
 
 
 def set_seed(args):
@@ -442,7 +399,7 @@ def main():
         default=None,
         type=str,
         required=True,
-        help="Model type selected in the list: " + ", ".join(MODEL_CLASSES.keys()),
+        help="Model type selected in the list: " + ", ".join(MODEL_TYPES),
     )
     parser.add_argument(
         "--model_name_or_path",
@@ -622,19 +579,18 @@ def main():
         torch.distributed.barrier()  # Make sure only the first process in distributed training will download model & vocab
 
     args.model_type = args.model_type.lower()
-    config_class, model_class, tokenizer_class = MODEL_CLASSES[args.model_type]
-    config = config_class.from_pretrained(
+    config = AutoConfig.from_pretrained(
         args.config_name if args.config_name else args.model_name_or_path,
         num_labels=num_labels,
         finetuning_task=args.task_name,
         cache_dir=args.cache_dir if args.cache_dir else None,
     )
-    tokenizer = tokenizer_class.from_pretrained(
+    tokenizer = AutoTokenizer.from_pretrained(
         args.tokenizer_name if args.tokenizer_name else args.model_name_or_path,
         do_lower_case=args.do_lower_case,
         cache_dir=args.cache_dir if args.cache_dir else None,
     )
-    model = model_class.from_pretrained(
+    model = AutoModelForSequenceClassification.from_pretrained(
         args.model_name_or_path,
         from_tf=bool(".ckpt" in args.model_name_or_path),
         config=config,
@@ -673,14 +629,14 @@ def main():
         torch.save(args, os.path.join(args.output_dir, "training_args.bin"))
 
         # Load a trained model and vocabulary that you have fine-tuned
-        model = model_class.from_pretrained(args.output_dir)
-        tokenizer = tokenizer_class.from_pretrained(args.output_dir)
+        model = AutoModelForSequenceClassification.from_pretrained(args.output_dir)
+        tokenizer = AutoTokenizer.from_pretrained(args.output_dir)
         model.to(args.device)
 
     # Evaluation
     results = {}
     if args.do_eval and args.local_rank in [-1, 0]:
-        tokenizer = tokenizer_class.from_pretrained(args.output_dir, do_lower_case=args.do_lower_case)
+        tokenizer = AutoTokenizer.from_pretrained(args.output_dir, do_lower_case=args.do_lower_case)
         checkpoints = [args.output_dir]
         if args.eval_all_checkpoints:
             checkpoints = list(
@@ -692,7 +648,7 @@ def main():
             global_step = checkpoint.split("-")[-1] if len(checkpoints) > 1 else ""
             prefix = checkpoint.split("/")[-1] if checkpoint.find("checkpoint") != -1 else ""
 
-            model = model_class.from_pretrained(checkpoint)
+            model = AutoModelForSequenceClassification.from_pretrained(checkpoint)
             model.to(args.device)
             result = evaluate(args, model, tokenizer, prefix=prefix)
             result = dict((k + "_{}".format(global_step), v) for k, v in result.items())

examples/run_language_modeling.py

@@ -38,28 +38,14 @@ from torch.utils.data.distributed import DistributedSampler
 from tqdm import tqdm, trange
 
 from transformers import (
+    MODEL_WITH_LM_HEAD_MAPPING,
     WEIGHTS_NAME,
     AdamW,
-    BertConfig,
-    BertForMaskedLM,
-    BertTokenizer,
-    CamembertConfig,
-    CamembertForMaskedLM,
-    CamembertTokenizer,
-    DistilBertConfig,
-    DistilBertForMaskedLM,
-    DistilBertTokenizer,
-    GPT2Config,
-    GPT2LMHeadModel,
-    GPT2Tokenizer,
-    OpenAIGPTConfig,
-    OpenAIGPTLMHeadModel,
-    OpenAIGPTTokenizer,
+    AutoConfig,
+    AutoModelWithLMHead,
+    AutoTokenizer,
     PreTrainedModel,
     PreTrainedTokenizer,
-    RobertaConfig,
-    RobertaForMaskedLM,
-    RobertaTokenizer,
     get_linear_schedule_with_warmup,
 )
 
@@ -73,14 +59,8 @@ except ImportError:
 logger = logging.getLogger(__name__)
 
 
-MODEL_CLASSES = {
-    "gpt2": (GPT2Config, GPT2LMHeadModel, GPT2Tokenizer),
-    "openai-gpt": (OpenAIGPTConfig, OpenAIGPTLMHeadModel, OpenAIGPTTokenizer),
-    "bert": (BertConfig, BertForMaskedLM, BertTokenizer),
-    "roberta": (RobertaConfig, RobertaForMaskedLM, RobertaTokenizer),
-    "distilbert": (DistilBertConfig, DistilBertForMaskedLM, DistilBertTokenizer),
-    "camembert": (CamembertConfig, CamembertForMaskedLM, CamembertTokenizer),
-}
+MODEL_CONFIG_CLASSES = list(MODEL_WITH_LM_HEAD_MAPPING.keys())
+MODEL_TYPES = tuple(conf.model_type for conf in MODEL_CONFIG_CLASSES)
 
 
 class TextDataset(Dataset):
@@ -693,23 +673,26 @@ def main():
     if args.local_rank not in [-1, 0]:
         torch.distributed.barrier()  # Barrier to make sure only the first process in distributed training download model & vocab
 
-    config_class, model_class, tokenizer_class = MODEL_CLASSES[args.model_type]
-
     if args.config_name:
-        config = config_class.from_pretrained(args.config_name, cache_dir=args.cache_dir)
+        config = AutoConfig.from_pretrained(args.config_name, cache_dir=args.cache_dir)
     elif args.model_name_or_path:
-        config = config_class.from_pretrained(args.model_name_or_path, cache_dir=args.cache_dir)
+        config = AutoConfig.from_pretrained(args.model_name_or_path, cache_dir=args.cache_dir)
     else:
-        config = config_class()
+        # When we release a pip version exposing CONFIG_MAPPING,
+        # we can do `config = CONFIG_MAPPING[args.model_type]()`.
+        raise ValueError(
+            "You are instantiating a new config instance from scratch. This is not supported, but you can do it from another script, save it,"
+            "and load it from here, using --config_name"
+        )
 
     if args.tokenizer_name:
-        tokenizer = tokenizer_class.from_pretrained(args.tokenizer_name, cache_dir=args.cache_dir)
+        tokenizer = AutoTokenizer.from_pretrained(args.tokenizer_name, cache_dir=args.cache_dir)
     elif args.model_name_or_path:
-        tokenizer = tokenizer_class.from_pretrained(args.model_name_or_path, cache_dir=args.cache_dir)
+        tokenizer = AutoTokenizer.from_pretrained(args.model_name_or_path, cache_dir=args.cache_dir)
     else:
         raise ValueError(
-            "You are instantiating a new {} tokenizer. This is not supported, but you can do it from another script, save it,"
-            "and load it from here, using --tokenizer_name".format(tokenizer_class.__name__)
+            "You are instantiating a new tokenizer from scratch. This is not supported, but you can do it from another script, save it,"
+            "and load it from here, using --tokenizer_name"
         )
 
     if args.block_size <= 0:
@@ -719,7 +702,7 @@ def main():
         args.block_size = min(args.block_size, tokenizer.max_len)
 
     if args.model_name_or_path:
-        model = model_class.from_pretrained(
+        model = AutoModelWithLMHead.from_pretrained(
             args.model_name_or_path,
             from_tf=bool(".ckpt" in args.model_name_or_path),
             config=config,
@@ -727,7 +710,7 @@ def main():
         )
     else:
         logger.info("Training new model from scratch")
-        model = model_class(config=config)
+        model = AutoModelWithLMHead.from_config(config)
 
     model.to(args.device)
 
@@ -768,8 +751,8 @@ def main():
         torch.save(args, os.path.join(args.output_dir, "training_args.bin"))
 
         # Load a trained model and vocabulary that you have fine-tuned
-        model = model_class.from_pretrained(args.output_dir)
-        tokenizer = tokenizer_class.from_pretrained(args.output_dir)
+        model = AutoModelWithLMHead.from_pretrained(args.output_dir)
+        tokenizer = AutoTokenizer.from_pretrained(args.output_dir)
         model.to(args.device)
 
     # Evaluation
@@ -786,7 +769,7 @@ def main():
             global_step = checkpoint.split("-")[-1] if len(checkpoints) > 1 else ""
             prefix = checkpoint.split("/")[-1] if checkpoint.find("checkpoint") != -1 else ""
 
-            model = model_class.from_pretrained(checkpoint)
+            model = AutoModelWithLMHead.from_pretrained(checkpoint)
             model.to(args.device)
             result = evaluate(args, model, tokenizer, prefix=prefix)
             result = dict((k + "_{}".format(global_step), v) for k, v in result.items())

examples/run_squad.py

@@ -30,29 +30,12 @@ from torch.utils.data.distributed import DistributedSampler
 from tqdm import tqdm, trange
 
 from transformers import (
+    MODEL_FOR_QUESTION_ANSWERING_MAPPING,
     WEIGHTS_NAME,
     AdamW,
-    AlbertConfig,
-    AlbertForQuestionAnswering,
-    AlbertTokenizer,
-    BertConfig,
-    BertForQuestionAnswering,
-    BertTokenizer,
-    CamembertConfig,
-    CamembertForQuestionAnswering,
-    CamembertTokenizer,
-    DistilBertConfig,
-    DistilBertForQuestionAnswering,
-    DistilBertTokenizer,
-    RobertaConfig,
-    RobertaForQuestionAnswering,
-    RobertaTokenizer,
-    XLMConfig,
-    XLMForQuestionAnswering,
-    XLMTokenizer,
-    XLNetConfig,
-    XLNetForQuestionAnswering,
-    XLNetTokenizer,
+    AutoConfig,
+    AutoModelForQuestionAnswering,
+    AutoTokenizer,
     get_linear_schedule_with_warmup,
     squad_convert_examples_to_features,
 )
@@ -72,23 +55,10 @@ except ImportError:
 
 logger = logging.getLogger(__name__)
 
-ALL_MODELS = sum(
-    (
-        tuple(conf.pretrained_config_archive_map.keys())
-        for conf in (BertConfig, CamembertConfig, RobertaConfig, XLNetConfig, XLMConfig)
-    ),
-    (),
-)
+MODEL_CONFIG_CLASSES = list(MODEL_FOR_QUESTION_ANSWERING_MAPPING.keys())
+MODEL_TYPES = tuple(conf.model_type for conf in MODEL_CONFIG_CLASSES)
 
-MODEL_CLASSES = {
-    "bert": (BertConfig, BertForQuestionAnswering, BertTokenizer),
-    "camembert": (CamembertConfig, CamembertForQuestionAnswering, CamembertTokenizer),
-    "roberta": (RobertaConfig, RobertaForQuestionAnswering, RobertaTokenizer),
-    "xlnet": (XLNetConfig, XLNetForQuestionAnswering, XLNetTokenizer),
-    "xlm": (XLMConfig, XLMForQuestionAnswering, XLMTokenizer),
-    "distilbert": (DistilBertConfig, DistilBertForQuestionAnswering, DistilBertTokenizer),
-    "albert": (AlbertConfig, AlbertForQuestionAnswering, AlbertTokenizer),
-}
+ALL_MODELS = sum((tuple(conf.pretrained_config_archive_map.keys()) for conf in MODEL_CONFIG_CLASSES), (),)
 
 
 def set_seed(args):
@@ -513,7 +483,7 @@ def main():
         default=None,
         type=str,
         required=True,
-        help="Model type selected in the list: " + ", ".join(MODEL_CLASSES.keys()),
+        help="Model type selected in the list: " + ", ".join(MODEL_TYPES),
     )
     parser.add_argument(
         "--model_name_or_path",
@@ -757,17 +727,16 @@ def main():
         torch.distributed.barrier()
 
     args.model_type = args.model_type.lower()
-    config_class, model_class, tokenizer_class = MODEL_CLASSES[args.model_type]
-    config = config_class.from_pretrained(
+    config = AutoConfig.from_pretrained(
         args.config_name if args.config_name else args.model_name_or_path,
         cache_dir=args.cache_dir if args.cache_dir else None,
     )
-    tokenizer = tokenizer_class.from_pretrained(
+    tokenizer = AutoTokenizer.from_pretrained(
         args.tokenizer_name if args.tokenizer_name else args.model_name_or_path,
         do_lower_case=args.do_lower_case,
         cache_dir=args.cache_dir if args.cache_dir else None,
     )
-    model = model_class.from_pretrained(
+    model = AutoModelForQuestionAnswering.from_pretrained(
         args.model_name_or_path,
         from_tf=bool(".ckpt" in args.model_name_or_path),
         config=config,
@@ -817,8 +786,8 @@ def main():
         torch.save(args, os.path.join(args.output_dir, "training_args.bin"))
 
         # Load a trained model and vocabulary that you have fine-tuned
-        model = model_class.from_pretrained(args.output_dir)  # , force_download=True)
-        tokenizer = tokenizer_class.from_pretrained(args.output_dir, do_lower_case=args.do_lower_case)
+        model = AutoModelForQuestionAnswering.from_pretrained(args.output_dir)  # , force_download=True)
+        tokenizer = AutoTokenizer.from_pretrained(args.output_dir, do_lower_case=args.do_lower_case)
         model.to(args.device)
 
     # Evaluation - we can ask to evaluate all the checkpoints (sub-directories) in a directory
@@ -842,7 +811,7 @@ def main():
         for checkpoint in checkpoints:
             # Reload the model
             global_step = checkpoint.split("-")[-1] if len(checkpoints) > 1 else ""
-            model = model_class.from_pretrained(checkpoint)  # , force_download=True)
+            model = AutoModelForQuestionAnswering.from_pretrained(checkpoint)  # , force_download=True)
             model.to(args.device)
 
             # Evaluate

examples/summarization/bart/README.md

@@ -14,6 +14,19 @@ python evaluate_cnn.py <path_to_test.source> cnn_test_summaries.txt
 ```
 the default batch size, 8, fits in 16GB GPU memory, but may need to be adjusted to fit your system.
 
+
+### Training
+
+
+
+After downloading the CNN and Daily Mail datasets, preprocess the dataset:
+```commandline
+git clone https://github.com/artmatsak/cnn-dailymail
+cd cnn-dailymail && python make_datafiles.py ../cnn/stories/ ../dailymail/stories/
+```
+
+Run the training script: `run_train.sh`
+
 ### Where is the code?
 The core model is in `src/transformers/modeling_bart.py`. This directory only contains examples.
 

examples/summarization/bart/run_bart_sum.py

@@ -0,0 +1,172 @@
+import argparse
+import glob
+import logging
+import os
+import time
+
+import torch
+from torch.utils.data import DataLoader
+
+from transformer_base import BaseTransformer, add_generic_args, generic_train, get_linear_schedule_with_warmup
+from utils import SummarizationDataset
+
+
+logger = logging.getLogger(__name__)
+
+
+class BartSystem(BaseTransformer):
+
+    mode = "language-modeling"
+
+    def __init__(self, hparams):
+        super(BartSystem, self).__init__(hparams, num_labels=None, mode=self.mode)
+
+    def forward(
+        self, input_ids, attention_mask=None, decoder_input_ids=None, decoder_attention_mask=None, lm_labels=None
+    ):
+        return self.model(
+            input_ids,
+            attention_mask=attention_mask,
+            decoder_input_ids=decoder_input_ids,
+            decoder_attention_mask=decoder_attention_mask,
+            lm_labels=lm_labels,
+        )
+
+    def _step(self, batch):
+        y = batch["target_ids"]
+        y_ids = y[:, :-1].contiguous()
+        lm_labels = y[:, 1:].clone()
+        lm_labels[y[:, 1:] == self.tokenizer.pad_token_id] = -100
+        outputs = self(
+            input_ids=batch["source_ids"],
+            attention_mask=batch["source_mask"],
+            decoder_input_ids=y_ids,
+            lm_labels=lm_labels,
+        )
+
+        loss = outputs[0]
+
+        return loss
+
+    def training_step(self, batch, batch_idx):
+        loss = self._step(batch)
+
+        tensorboard_logs = {"train_loss": loss}
+        return {"loss": loss, "log": tensorboard_logs}
+
+    def validation_step(self, batch, batch_idx):
+        loss = self._step(batch)
+        return {"val_loss": loss}
+
+    def validation_end(self, outputs):
+        avg_loss = torch.stack([x["val_loss"] for x in outputs]).mean()
+        tensorboard_logs = {"val_loss": avg_loss}
+        return {"avg_val_loss": avg_loss, "log": tensorboard_logs}
+
+    def test_step(self, batch, batch_idx):
+        generated_ids = self.model.generate(
+            batch["source_ids"],
+            attention_mask=batch["source_mask"],
+            num_beams=1,
+            max_length=80,
+            repetition_penalty=2.5,
+            length_penalty=1.0,
+            early_stopping=True,
+        )
+        preds = [
+            self.tokenizer.decode(g, skip_special_tokens=True, clean_up_tokenization_spaces=True)
+            for g in generated_ids
+        ]
+        target = [
+            self.tokenizer.decode(t, skip_special_tokens=True, clean_up_tokenization_spaces=True)
+            for t in batch["target_ids"]
+        ]
+        loss = self._step(batch)
+
+        return {"val_loss": loss, "preds": preds, "target": target}
+
+    def test_end(self, outputs):
+        return self.validation_end(outputs)
+
+    def test_epoch_end(self, outputs):
+        output_test_predictions_file = os.path.join(self.hparams.output_dir, "test_predictions.txt")
+        output_test_targets_file = os.path.join(self.hparams.output_dir, "test_targets.txt")
+        # write predictions and targets for later rouge evaluation.
+        with open(output_test_predictions_file, "w+") as p_writer, open(output_test_targets_file, "w+") as t_writer:
+            for output_batch in outputs:
+                p_writer.writelines(s + "\n" for s in output_batch["preds"])
+                t_writer.writelines(s + "\n" for s in output_batch["target"])
+            p_writer.close()
+            t_writer.close()
+
+        return self.test_end(outputs)
+
+    def train_dataloader(self):
+        train_dataset = SummarizationDataset(
+            self.tokenizer, data_dir=self.hparams.data_dir, type_path="train", block_size=self.hparams.max_seq_length
+        )
+        dataloader = DataLoader(train_dataset, batch_size=self.hparams.train_batch_size)
+        t_total = (
+            (len(dataloader.dataset) // (self.hparams.train_batch_size * max(1, self.hparams.n_gpu)))
+            // self.hparams.gradient_accumulation_steps
+            * float(self.hparams.num_train_epochs)
+        )
+        scheduler = get_linear_schedule_with_warmup(
+            self.opt, num_warmup_steps=self.hparams.warmup_steps, num_training_steps=t_total
+        )
+        self.lr_scheduler = scheduler
+        return dataloader
+
+    def val_dataloader(self):
+        val_dataset = SummarizationDataset(
+            self.tokenizer, data_dir=self.hparams.data_dir, type_path="val", block_size=self.hparams.max_seq_length
+        )
+        return DataLoader(val_dataset, batch_size=self.hparams.eval_batch_size)
+
+    def test_dataloader(self):
+        test_dataset = SummarizationDataset(
+            self.tokenizer, data_dir=self.hparams.data_dir, type_path="test", block_size=self.hparams.max_seq_length
+        )
+        return DataLoader(test_dataset, batch_size=self.hparams.eval_batch_size)
+
+    @staticmethod
+    def add_model_specific_args(parser, root_dir):
+        BaseTransformer.add_model_specific_args(parser, root_dir)
+        # Add BART specific options
+        parser.add_argument(
+            "--max_seq_length",
+            default=1024,
+            type=int,
+            help="The maximum total input sequence length after tokenization. Sequences longer "
+            "than this will be truncated, sequences shorter will be padded.",
+        )
+
+        parser.add_argument(
+            "--data_dir",
+            default=None,
+            type=str,
+            required=True,
+            help="The input data dir. Should contain the dataset files for the CNN/DM summarization task.",
+        )
+        return parser
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    add_generic_args(parser, os.getcwd())
+    parser = BartSystem.add_model_specific_args(parser, os.getcwd())
+    args = parser.parse_args()
+
+    # If output_dir not provided, a folder will be generated in pwd
+    if args.output_dir is None:
+        args.output_dir = os.path.join("./results", f"{args.task}_{args.model_type}_{time.strftime('%Y%m%d_%H%M%S')}",)
+        os.makedirs(args.output_dir)
+
+    model = BartSystem(args)
+    trainer = generic_train(model, args)
+
+    # Optionally, predict on dev set and write to output_dir
+    if args.do_predict:
+        checkpoints = list(sorted(glob.glob(os.path.join(args.output_dir, "checkpointepoch=*.ckpt"), recursive=True)))
+        BartSystem.load_from_checkpoint(checkpoints[-1])
+        trainer.test(model)

examples/summarization/bart/run_train.sh

@@ -0,0 +1,23 @@
+# Install newest ptl.
+pip install -U git+http://github.com/PyTorchLightning/pytorch-lightning/
+
+
+export OUTPUT_DIR_NAME=bart_sum
+export CURRENT_DIR=${PWD}
+export OUTPUT_DIR=${CURRENT_DIR}/${OUTPUT_DIR_NAME}
+
+# Make output directory if it doesn't exist
+mkdir -p $OUTPUT_DIR
+
+# Add parent directory to python path to access transformer_base.py
+export PYTHONPATH="../../":"${PYTHONPATH}"
+
+python run_bart_sum.py \
+--data_dir=./cnn-dailymail/cnn_dm \
+--model_type=bart \
+--model_name_or_path=bart-large \
+--learning_rate=3e-5 \
+--train_batch_size=4 \
+--eval_batch_size=4 \
+--output_dir=$OUTPUT_DIR \
+--do_train

examples/summarization/bart/utils.py

@@ -0,0 +1,43 @@
+import os
+
+from torch.utils.data import Dataset
+
+
+class SummarizationDataset(Dataset):
+    def __init__(self, tokenizer, data_dir="./cnn-dailymail/cnn_dm/", type_path="train", block_size=1024):
+        super(SummarizationDataset,).__init__()
+        self.tokenizer = tokenizer
+
+        self.source = []
+        self.target = []
+
+        print("loading " + type_path + " source.")
+
+        with open(os.path.join(data_dir, type_path + ".source"), "r") as f:
+            for text in f.readlines():  # each text is a line and a full story
+                tokenized = tokenizer.batch_encode_plus(
+                    [text], max_length=block_size, pad_to_max_length=True, return_tensors="pt"
+                )
+                self.source.append(tokenized)
+            f.close()
+
+        print("loading " + type_path + " target.")
+
+        with open(os.path.join(data_dir, type_path + ".target"), "r") as f:
+            for text in f.readlines():  # each text is a line and a summary
+                tokenized = tokenizer.batch_encode_plus(
+                    [text], max_length=56, pad_to_max_length=True, return_tensors="pt"
+                )
+                self.target.append(tokenized)
+            f.close()
+
+    def __len__(self):
+        return len(self.source)
+
+    def __getitem__(self, index):
+        source_ids = self.source[index]["input_ids"].squeeze()
+        target_ids = self.target[index]["input_ids"].squeeze()
+
+        src_mask = self.source[index]["attention_mask"].squeeze()  # might need to squeeze
+
+        return {"source_ids": source_ids, "source_mask": src_mask, "target_ids": target_ids}

examples/summarization/t5/README.md

@@ -0,0 +1,25 @@
+***This script evaluates the the multitask pre-trained checkpoint for ``t5-large`` (see paper [here](https://arxiv.org/pdf/1910.10683.pdf)) on the CNN/Daily Mail test dataset. Please note that the results in the paper were attained using a model fine-tuned on summarization, so that results will be worse here by approx. 0.5 ROUGE points***
+
+### Get the CNN Data
+First, you need to download the CNN data. It's about ~400 MB and can be downloaded by 
+running 
+
+```bash
+python download_cnn_daily_mail.py cnn_articles_input_data.txt cnn_articles_reference_summaries.txt
+```
+
+You should confirm that each file has 11490 lines:
+
+```bash
+wc -l cnn_articles_input_data.txt # should print 11490
+wc -l cnn_articles_reference_summaries.txt # should print 11490
+```
+
+### Usage
+
+To create summaries for each article in dataset, run:
+```bash
+python evaluate_cnn.py cnn_articles_input_data.txt cnn_generated_articles_summaries.txt cnn_articles_reference_summaries.txt rouge_score.txt
+```
+The default batch size, 8, fits in 16GB GPU memory, but may need to be adjusted to fit your system.
+The rouge scores "rouge1, rouge2, rougeL" are automatically created and saved in ``rouge_score.txt``.

examples/summarization/t5/download_cnn_daily_mail.py

@@ -0,0 +1,31 @@
+import argparse
+from pathlib import Path
+
+import tensorflow_datasets as tfds
+
+
+def main(input_path, reference_path, data_dir):
+    cnn_ds = tfds.load("cnn_dailymail", split="test", shuffle_files=False, data_dir=data_dir)
+    cnn_ds_iter = tfds.as_numpy(cnn_ds)
+
+    test_articles_file = Path(input_path).open("w")
+    test_summaries_file = Path(reference_path).open("w")
+
+    for example in cnn_ds_iter:
+        test_articles_file.write(example["article"].decode("utf-8") + "\n")
+        test_articles_file.flush()
+        test_summaries_file.write(example["highlights"].decode("utf-8").replace("\n", " ") + "\n")
+        test_summaries_file.flush()
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("input_path", type=str, help="where to save the articles input data")
+    parser.add_argument(
+        "reference_path", type=str, help="where to save the reference summaries",
+    )
+    parser.add_argument(
+        "--data_dir", type=str, default="~/tensorflow_datasets", help="where to save the tensorflow datasets.",
+    )
+    args = parser.parse_args()
+    main(args.input_path, args.reference_path, args.data_dir)

examples/summarization/t5/evaluate_cnn.py

@@ -0,0 +1,95 @@
+import argparse
+from pathlib import Path
+
+import torch
+from tqdm import tqdm
+
+from rouge_score import rouge_scorer, scoring
+from transformers import T5ForConditionalGeneration, T5Tokenizer
+
+
+def chunks(lst, n):
+    """Yield successive n-sized chunks from lst."""
+    for i in range(0, len(lst), n):
+        yield lst[i : i + n]
+
+
+def generate_summaries(lns, output_file_path, batch_size, device):
+    output_file = Path(output_file_path).open("w")
+
+    model = T5ForConditionalGeneration.from_pretrained("t5-large")
+    model.to(device)
+
+    tokenizer = T5Tokenizer.from_pretrained("t5-large")
+
+    # update config with summarization specific params
+    task_specific_params = model.config.task_specific_params
+    if task_specific_params is not None:
+        model.config.update(task_specific_params.get("summarization", {}))
+
+    for batch in tqdm(list(chunks(lns, batch_size))):
+        batch = [model.config.prefix + text for text in batch]
+
+        dct = tokenizer.batch_encode_plus(batch, max_length=512, return_tensors="pt", pad_to_max_length=True)
+        input_ids = dct["input_ids"].to(device)
+        attention_mask = dct["attention_mask"].to(device)
+
+        summaries = model.generate(input_ids=input_ids, attention_mask=attention_mask)
+        dec = [tokenizer.decode(g, skip_special_tokens=True, clean_up_tokenization_spaces=False) for g in summaries]
+
+        for hypothesis in dec:
+            output_file.write(hypothesis + "\n")
+            output_file.flush()
+
+
+def calculate_rouge(output_lns, reference_lns, score_path):
+    score_file = Path(score_path).open("w")
+    scorer = rouge_scorer.RougeScorer(["rouge1", "rouge2", "rougeL"], use_stemmer=True)
+    aggregator = scoring.BootstrapAggregator()
+
+    for reference_ln, output_ln in zip(reference_lns, output_lns):
+        scores = scorer.score(reference_ln, output_ln)
+        aggregator.add_scores(scores)
+
+    result = aggregator.aggregate()
+    score_file.write(
+        "ROUGE_1: \n{} \n\n ROUGE_2: \n{} \n\n ROUGE_L: \n{} \n\n".format(
+            result["rouge1"], result["rouge2"], result["rougeL"]
+        )
+    )
+
+
+def run_generate():
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "input_path", type=str, help="like cnn_dm/test_articles_input.txt",
+    )
+    parser.add_argument(
+        "output_path", type=str, help="where to save summaries",
+    )
+    parser.add_argument("reference_path", type=str, help="like cnn_dm/test_reference_summaries.txt")
+    parser.add_argument(
+        "score_path", type=str, help="where to save the rouge score",
+    )
+    parser.add_argument(
+        "--batch_size", type=int, default=8, required=False, help="batch size: how many to summarize at a time",
+    )
+    parser.add_argument(
+        "--no_cuda", default=False, type=bool, help="Whether to force the execution on CPU.",
+    )
+
+    args = parser.parse_args()
+    args.device = torch.device("cuda" if torch.cuda.is_available() and not args.no_cuda else "cpu")
+
+    source_lns = [x.rstrip() for x in open(args.input_path).readlines()]
+
+    generate_summaries(source_lns, args.output_path, args.batch_size, args.device)
+
+    output_lns = [x.rstrip() for x in open(args.output_path).readlines()]
+    reference_lns = [x.rstrip() for x in open(args.reference_path).readlines()]
+
+    calculate_rouge(output_lns, reference_lns, args.score_path)
+
+
+if __name__ == "__main__":
+    run_generate()

examples/summarization/t5/test_t5_examples.py

@@ -0,0 +1,29 @@
+import logging
+import sys
+import tempfile
+import unittest
+from pathlib import Path
+from unittest.mock import patch
+
+from .evaluate_cnn import run_generate
+
+
+articles = ["New York (CNN)When Liana Barrientos was 23 years old, she got married in Westchester County."]
+
+logging.basicConfig(level=logging.DEBUG)
+
+logger = logging.getLogger()
+
+
+class TestT5Examples(unittest.TestCase):
+    def test_t5_cli(self):
+        stream_handler = logging.StreamHandler(sys.stdout)
+        logger.addHandler(stream_handler)
+        tmp = Path(tempfile.gettempdir()) / "utest_generations.hypo"
+        with tmp.open("w") as f:
+            f.write("\n".join(articles))
+        testargs = ["evaluate_cnn.py", str(tmp), "output.txt", str(tmp), "score.txt"]
+        with patch.object(sys, "argv", testargs):
+            run_generate()
+            self.assertTrue(Path("output.txt").exists())
+            self.assertTrue(Path("score.txt").exists())

examples/transformer_base.py

@@ -53,10 +53,9 @@ class BaseTransformer(pl.LightningModule):
         super(BaseTransformer, self).__init__()
         self.hparams = hparams
         self.hparams.model_type = self.hparams.model_type.lower()
-
         config = AutoConfig.from_pretrained(
             self.hparams.config_name if self.hparams.config_name else self.hparams.model_name_or_path,
-            num_labels=num_labels,
+            **({"num_labels": num_labels} if num_labels is not None else {}),
             cache_dir=self.hparams.cache_dir if self.hparams.cache_dir else None,
         )
         tokenizer = AutoTokenizer.from_pretrained(

examples/translation/t5/README.md

@@ -0,0 +1,51 @@
+***This script evaluates the multitask pre-trained checkpoint for ``t5-base`` (see paper [here](https://arxiv.org/pdf/1910.10683.pdf)) on the English to German WMT dataset. Please note that the results in the paper were attained using a model fine-tuned on translation, so that results will be worse here by approx. 1.5 BLEU points***
+
+### Intro
+
+This example shows how T5 (here the official [paper](https://arxiv.org/abs/1910.10683)) can be
+evaluated on the WMT English-German dataset.
+
+### Get the WMT Data
+
+To be able to reproduce the authors' results on WMT English to German, you first need to download 
+the WMT14 en-de news datasets.
+Go on Stanford's official NLP [website](https://nlp.stanford.edu/projects/nmt/) and find "newstest2013.en" and "newstest2013.de" under WMT'14 English-German data or download the dataset directly via:
+
+```bash
+curl https://nlp.stanford.edu/projects/nmt/data/wmt14.en-de/newstest2013.en > newstest2013.en
+curl https://nlp.stanford.edu/projects/nmt/data/wmt14.en-de/newstest2013.de > newstest2013.de
+```
+
+You should have 3000 sentence in each file. You can verify this by running:
+
+```bash
+wc -l newstest2013.en  # should give 3000
+```
+
+### Usage
+
+Let's check the longest and shortest sentence in our file to find reasonable decoding hyperparameters: 
+
+Get the longest and shortest sentence:
+
+```bash 
+awk '{print NF}' newstest2013.en | sort -n | head -1 # shortest sentence has 1 word
+awk '{print NF}' newstest2013.en | sort -n | tail -1 # longest sentence has 106 words
+```
+
+We will set our `max_length` to ~3 times the longest sentence and leave `min_length` to its default value of 0.
+We decode with beam search `num_beams=4` as proposed in the paper. Also as is common in beam search we set `early_stopping=True` and `length_penalty=2.0`.
+
+To create translation for each in dataset and get a final BLEU score, run:
+```bash
+python evaluate_wmt.py <path_to_newstest2013.en> newstest2013_de_translations.txt <path_to_newstest2013.de> newsstest2013_en_de_bleu.txt
+```
+the default batch size, 16, fits in 16GB GPU memory, but may need to be adjusted to fit your system.
+
+### Where is the code?
+The core model is in `src/transformers/modeling_t5.py`. This directory only contains examples.
+
+### BLEU Scores
+
+The BLEU score is calculated using [sacrebleu](https://github.com/mjpost/sacreBLEU) by mjpost.
+To get the BLEU score we used 

examples/translation/t5/evaluate_wmt.py

@@ -0,0 +1,90 @@
+import argparse
+from pathlib import Path
+
+import torch
+from tqdm import tqdm
+
+from sacrebleu import corpus_bleu
+from transformers import T5ForConditionalGeneration, T5Tokenizer
+
+
+def chunks(lst, n):
+    """Yield successive n-sized chunks from lst."""
+    for i in range(0, len(lst), n):
+        yield lst[i : i + n]
+
+
+def generate_translations(lns, output_file_path, batch_size, device):
+    output_file = Path(output_file_path).open("w")
+
+    model = T5ForConditionalGeneration.from_pretrained("t5-base")
+    model.to(device)
+
+    tokenizer = T5Tokenizer.from_pretrained("t5-base")
+
+    # update config with summarization specific params
+    task_specific_params = model.config.task_specific_params
+    if task_specific_params is not None:
+        model.config.update(task_specific_params.get("translation_en_to_de", {}))
+
+    for batch in tqdm(list(chunks(lns, batch_size))):
+        batch = [model.config.prefix + text for text in batch]
+
+        dct = tokenizer.batch_encode_plus(batch, max_length=512, return_tensors="pt", pad_to_max_length=True)
+
+        input_ids = dct["input_ids"].to(device)
+        attention_mask = dct["attention_mask"].to(device)
+
+        translations = model.generate(input_ids=input_ids, attention_mask=attention_mask)
+        dec = [tokenizer.decode(g, skip_special_tokens=True, clean_up_tokenization_spaces=False) for g in translations]
+
+        for hypothesis in dec:
+            output_file.write(hypothesis + "\n")
+            output_file.flush()
+
+
+def calculate_bleu_score(output_lns, refs_lns, score_path):
+    bleu = corpus_bleu(output_lns, [refs_lns])
+    result = "BLEU score: {}".format(bleu.score)
+    score_file = Path(score_path).open("w")
+    score_file.write(result)
+
+
+def run_generate():
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "input_path", type=str, help="like wmt/newstest2013.en",
+    )
+    parser.add_argument(
+        "output_path", type=str, help="where to save translation",
+    )
+    parser.add_argument(
+        "reference_path", type=str, help="like wmt/newstest2013.de",
+    )
+    parser.add_argument(
+        "score_path", type=str, help="where to save the bleu score",
+    )
+    parser.add_argument(
+        "--batch_size", type=int, default=16, required=False, help="batch size: how many to summarize at a time",
+    )
+    parser.add_argument(
+        "--no_cuda", default=False, type=bool, help="Whether to force the execution on CPU.",
+    )
+
+    args = parser.parse_args()
+    args.device = torch.device("cuda" if torch.cuda.is_available() and not args.no_cuda else "cpu")
+
+    dash_pattern = (" ##AT##-##AT## ", "-")
+
+    input_lns = [x.strip().replace(dash_pattern[0], dash_pattern[1]) for x in open(args.input_path).readlines()]
+
+    generate_translations(input_lns, args.output_path, args.batch_size, args.device)
+
+    output_lns = [x.strip() for x in open(args.output_path).readlines()]
+    refs_lns = [x.strip().replace(dash_pattern[0], dash_pattern[1]) for x in open(args.reference_path).readlines()]
+
+    calculate_bleu_score(output_lns, refs_lns, args.score_path)
+
+
+if __name__ == "__main__":
+    run_generate()

examples/translation/t5/test_t5_examples.py

@@ -0,0 +1,28 @@
+import logging
+import sys
+import tempfile
+import unittest
+from pathlib import Path
+from unittest.mock import patch
+
+from .evaluate_wmt import run_generate
+
+
+text = [" New York (CNN)When Liana Barrientos was 23 years old, she got married in Westchester County."]
+
+logging.basicConfig(level=logging.DEBUG)
+
+logger = logging.getLogger()
+
+
+class TestT5Examples(unittest.TestCase):
+    def test_t5_cli(self):
+        stream_handler = logging.StreamHandler(sys.stdout)
+        logger.addHandler(stream_handler)
+        tmp = Path(tempfile.gettempdir()) / "utest_generations.hypo"
+        with tmp.open("w") as f:
+            f.write("\n".join(text))
+        testargs = ["evaluate_cnn.py", str(tmp), "output.txt", str(tmp), "score.txt"]
+        with patch.object(sys, "argv", testargs):
+            run_generate()
+            self.assertTrue(Path("output.txt").exists())

examples/utils_multiple_choice.py

@@ -320,7 +320,9 @@ def convert_examples_to_features(
             else:
                 text_b = example.question + " " + ending
 
-            inputs = tokenizer.encode_plus(text_a, text_b, add_special_tokens=True, max_length=max_length,)
+            inputs = tokenizer.encode_plus(
+                text_a, text_b, add_special_tokens=True, max_length=max_length, return_token_type_ids=True
+            )
             if "num_truncated_tokens" in inputs and inputs["num_truncated_tokens"] > 0:
                 logger.info(
                     "Attention! you are cropping tokens (swag task is ok). "

model_cards/dbmdz/bert-base-turkish-128k-cased/README.md

@@ -0,0 +1,76 @@
+---
+language: turkish
+---
+
+# 🤗 + 📚 dbmdz Turkish BERT model
+
+In this repository the MDZ Digital Library team (dbmdz) at the Bavarian State
+Library open sources a cased model for Turkish 🎉
+
+# 🇹🇷 BERTurk
+
+BERTurk is a community-driven cased BERT model for Turkish.
+
+Some datasets used for pretraining and evaluation are contributed from the
+awesome Turkish NLP community, as well as the decision for the model name: BERTurk.
+
+## Stats
+
+The current version of the model is trained on a filtered and sentence
+segmented version of the Turkish [OSCAR corpus](https://traces1.inria.fr/oscar/),
+a recent Wikipedia dump, various [OPUS corpora](http://opus.nlpl.eu/) and a
+special corpus provided by [Kemal Oflazer](http://www.andrew.cmu.edu/user/ko/).
+
+The final training corpus has a size of 35GB and 44,04,976,662 tokens.
+
+Thanks to Google's TensorFlow Research Cloud (TFRC) we could train a cased model
+on a TPU v3-8 for 2M steps.
+
+For this model we use a vocab size of 128k.
+
+## Model weights
+
+Currently only PyTorch-[Transformers](https://github.com/huggingface/transformers)
+compatible weights are available. If you need access to TensorFlow checkpoints,
+please raise an issue!
+
+| Model                                | Downloads
+| ------------------------------------ | ---------------------------------------------------------------------------------------------------------------
+| `dbmdz/bert-base-turkish-128k-cased` | [`config.json`](https://cdn.huggingface.co/dbmdz/bert-base-turkish-128k-cased/config.json) • [`pytorch_model.bin`](https://cdn.huggingface.co/dbmdz/bert-base-turkish-128k-cased/pytorch_model.bin) • [`vocab.txt`](https://cdn.huggingface.co/dbmdz/bert-base-turkish-128k-cased/vocab.txt)
+
+## Usage
+
+With Transformers >= 2.3 our BERTurk cased model can be loaded like:
+
+```python
+from transformers import AutoModel, AutoTokenizer
+
+tokenizer = AutoTokenizer.from_pretrained("dbmdz/bert-base-turkish-128k-cased")
+model = AutoModel.from_pretrained("dbmdz/bert-base-turkish-128k-cased")
+```
+
+## Results
+
+For results on PoS tagging or NER tasks, please refer to
+[this repository](https://github.com/stefan-it/turkish-bert).
+
+# Huggingface model hub
+
+All models are available on the [Huggingface model hub](https://huggingface.co/dbmdz).
+
+# Contact (Bugs, Feedback, Contribution and more)
+
+For questions about our BERT models just open an issue
+[here](https://github.com/dbmdz/berts/issues/new) 🤗
+
+# Acknowledgments
+
+Thanks to [Kemal Oflazer](http://www.andrew.cmu.edu/user/ko/) for providing us
+additional large corpora for Turkish. Many thanks to Reyyan Yeniterzi for providing
+us the Turkish NER dataset for evaluation.
+
+Research supported with Cloud TPUs from Google's TensorFlow Research Cloud (TFRC).
+Thanks for providing access to the TFRC ❤️
+
+Thanks to the generous support from the [Hugging Face](https://huggingface.co/) team,
+it is possible to download both cased and uncased models from their S3 storage 🤗

model_cards/dbmdz/bert-base-turkish-128k-uncased/README.md

@@ -0,0 +1,76 @@
+---
+language: turkish
+---
+
+# 🤗 + 📚 dbmdz Turkish BERT model
+
+In this repository the MDZ Digital Library team (dbmdz) at the Bavarian State
+Library open sources an uncased model for Turkish 🎉
+
+# 🇹🇷 BERTurk
+
+BERTurk is a community-driven uncased BERT model for Turkish.
+
+Some datasets used for pretraining and evaluation are contributed from the
+awesome Turkish NLP community, as well as the decision for the model name: BERTurk.
+
+## Stats
+
+The current version of the model is trained on a filtered and sentence
+segmented version of the Turkish [OSCAR corpus](https://traces1.inria.fr/oscar/),
+a recent Wikipedia dump, various [OPUS corpora](http://opus.nlpl.eu/) and a
+special corpus provided by [Kemal Oflazer](http://www.andrew.cmu.edu/user/ko/).
+
+The final training corpus has a size of 35GB and 44,04,976,662 tokens.
+
+Thanks to Google's TensorFlow Research Cloud (TFRC) we could train an uncased model
+on a TPU v3-8 for 2M steps.
+
+For this model we use a vocab size of 128k.
+
+## Model weights
+
+Currently only PyTorch-[Transformers](https://github.com/huggingface/transformers)
+compatible weights are available. If you need access to TensorFlow checkpoints,
+please raise an issue!
+
+| Model                                  | Downloads
+| -------------------------------------- | ---------------------------------------------------------------------------------------------------------------
+| `dbmdz/bert-base-turkish-128k-uncased` | [`config.json`](https://cdn.huggingface.co/dbmdz/bert-base-turkish-128k-uncased/config.json) • [`pytorch_model.bin`](https://cdn.huggingface.co/dbmdz/bert-base-turkish-128k-uncased/pytorch_model.bin) • [`vocab.txt`](https://cdn.huggingface.co/dbmdz/bert-base-turkish-128k-uncased/vocab.txt)
+
+## Usage
+
+With Transformers >= 2.3 our BERTurk uncased model can be loaded like:
+
+```python
+from transformers import AutoModel, AutoTokenizer
+
+tokenizer = AutoTokenizer.from_pretrained("dbmdz/bert-base-turkish-128k-uncased")
+model = AutoModel.from_pretrained("dbmdz/bert-base-turkish-128k-uncased")
+```
+
+## Results
+
+For results on PoS tagging or NER tasks, please refer to
+[this repository](https://github.com/stefan-it/turkish-bert).
+
+# Huggingface model hub
+
+All models are available on the [Huggingface model hub](https://huggingface.co/dbmdz).
+
+# Contact (Bugs, Feedback, Contribution and more)
+
+For questions about our BERT models just open an issue
+[here](https://github.com/dbmdz/berts/issues/new) 🤗
+
+# Acknowledgments
+
+Thanks to [Kemal Oflazer](http://www.andrew.cmu.edu/user/ko/) for providing us
+additional large corpora for Turkish. Many thanks to Reyyan Yeniterzi for providing
+us the Turkish NER dataset for evaluation.
+
+Research supported with Cloud TPUs from Google's TensorFlow Research Cloud (TFRC).
+Thanks for providing access to the TFRC ❤️
+
+Thanks to the generous support from the [Hugging Face](https://huggingface.co/) team,
+it is possible to download both cased and uncased models from their S3 storage 🤗

model_cards/dbmdz/bert-base-turkish-uncased/README.md

@@ -0,0 +1,74 @@
+---
+language: turkish
+---
+
+# 🤗 + 📚 dbmdz Turkish BERT model
+
+In this repository the MDZ Digital Library team (dbmdz) at the Bavarian State
+Library open sources an uncased model for Turkish 🎉
+
+# 🇹🇷 BERTurk
+
+BERTurk is a community-driven uncased BERT model for Turkish.
+
+Some datasets used for pretraining and evaluation are contributed from the
+awesome Turkish NLP community, as well as the decision for the model name: BERTurk.
+
+## Stats
+
+The current version of the model is trained on a filtered and sentence
+segmented version of the Turkish [OSCAR corpus](https://traces1.inria.fr/oscar/),
+a recent Wikipedia dump, various [OPUS corpora](http://opus.nlpl.eu/) and a
+special corpus provided by [Kemal Oflazer](http://www.andrew.cmu.edu/user/ko/).
+
+The final training corpus has a size of 35GB and 44,04,976,662 tokens.
+
+Thanks to Google's TensorFlow Research Cloud (TFRC) we could train an uncased model
+on a TPU v3-8 for 2M steps.
+
+## Model weights
+
+Currently only PyTorch-[Transformers](https://github.com/huggingface/transformers)
+compatible weights are available. If you need access to TensorFlow checkpoints,
+please raise an issue!
+
+| Model                             | Downloads
+| --------------------------------- | ---------------------------------------------------------------------------------------------------------------
+| `dbmdz/bert-base-turkish-uncased`   | [`config.json`](https://cdn.huggingface.co/dbmdz/bert-base-turkish-uncased/config.json) • [`pytorch_model.bin`](https://cdn.huggingface.co/dbmdz/bert-base-turkish-uncased/pytorch_model.bin) • [`vocab.txt`](https://cdn.huggingface.co/dbmdz/bert-base-turkish-uncased/vocab.txt)
+
+## Usage
+
+With Transformers >= 2.3 our BERTurk uncased model can be loaded like:
+
+```python
+from transformers import AutoModel, AutoTokenizer
+
+tokenizer = AutoTokenizer.from_pretrained("dbmdz/bert-base-turkish-uncased")
+model = AutoModel.from_pretrained("dbmdz/bert-base-turkish-uncased")
+```
+
+## Results
+
+For results on PoS tagging or NER tasks, please refer to
+[this repository](https://github.com/stefan-it/turkish-bert).
+
+# Huggingface model hub
+
+All models are available on the [Huggingface model hub](https://huggingface.co/dbmdz).
+
+# Contact (Bugs, Feedback, Contribution and more)
+
+For questions about our BERT models just open an issue
+[here](https://github.com/dbmdz/berts/issues/new) 🤗
+
+# Acknowledgments
+
+Thanks to [Kemal Oflazer](http://www.andrew.cmu.edu/user/ko/) for providing us
+additional large corpora for Turkish. Many thanks to Reyyan Yeniterzi for providing
+us the Turkish NER dataset for evaluation.
+
+Research supported with Cloud TPUs from Google's TensorFlow Research Cloud (TFRC).
+Thanks for providing access to the TFRC ❤️
+
+Thanks to the generous support from the [Hugging Face](https://huggingface.co/) team,
+it is possible to download both cased and uncased models from their S3 storage 🤗

model_cards/elgeish/cs224n-squad2.0-albert-base-v2/README.md

@@ -0,0 +1,74 @@
+## CS224n SQuAD2.0 Project Dataset
+The goal of this model is to save CS224n students GPU time when establising
+baselines to beat for the [Default Final Project](http://web.stanford.edu/class/cs224n/project/default-final-project-handout.pdf).
+The training set used to fine-tune this model is the same as
+the [official one](https://rajpurkar.github.io/SQuAD-explorer/); however,
+evaluation and model selection were performed using roughly half of the official
+dev set, 6078 examples, picked at random. The data files can be found at
+<https://github.com/elgeish/squad/tree/master/data> — this is the Winter 2020
+version. Given that the official SQuAD2.0 dev set contains the project's test
+set, students must make sure not to use the official SQuAD2.0 dev set in any way
+— including the use of models fine-tuned on the official SQuAD2.0, since they
+used the official SQuAD2.0 dev set for model selection.
+
+## Results
+```json
+{
+  "exact": 78.94044093451794,
+  "f1": 81.7724930324639,
+  "total": 6078,
+  "HasAns_exact": 76.28865979381443,
+  "HasAns_f1": 82.20385314478195,
+  "HasAns_total": 2910,
+  "NoAns_exact": 81.37626262626263,
+  "NoAns_f1": 81.37626262626263,
+  "NoAns_total": 3168,
+  "best_exact": 78.95689371503784,
+  "best_exact_thresh": 0.0,
+  "best_f1": 81.78894581298378,
+  "best_f1_thresh": 0.0
+}
+```
+
+## Notable Arguments
+```json
+{
+  "do_lower_case": true,
+  "doc_stride": 128,
+  "fp16": false,
+  "fp16_opt_level": "O1",
+  "gradient_accumulation_steps": 24,
+  "learning_rate": 3e-05,
+  "max_answer_length": 30,
+  "max_grad_norm": 1,
+  "max_query_length": 64,
+  "max_seq_length": 384,
+  "model_name_or_path": "albert-base-v2",
+  "model_type": "albert",
+  "num_train_epochs": 3,
+  "per_gpu_train_batch_size": 8,
+  "save_steps": 5000,
+  "seed": 42,
+  "train_batch_size": 8,
+  "version_2_with_negative": true,
+  "warmup_steps": 0,
+  "weight_decay": 0
+}
+```
+
+## Environment Setup
+```json
+{
+  "transformers": "2.5.1",
+  "pytorch": "1.4.0=py3.6_cuda10.1.243_cudnn7.6.3_0",
+  "python": "3.6.5=hc3d631a_2",
+  "os": "Linux 4.15.0-1060-aws #62-Ubuntu SMP Tue Feb 11 21:23:22 UTC 2020 x86_64 x86_64 x86_64 GNU/Linux",
+  "gpu": "Tesla V100-SXM2-16GB"
+}
+```
+
+## Related Models
+* [elgeish/cs224n-squad2.0-albert-large-v2](https://huggingface.co/elgeish/cs224n-squad2.0-albert-large-v2)
+* [elgeish/cs224n-squad2.0-albert-xxlarge-v1](https://huggingface.co/elgeish/cs224n-squad2.0-albert-xxlarge-v1)
+* [elgeish/cs224n-squad2.0-distilbert-base-uncased](https://huggingface.co/elgeish/cs224n-squad2.0-distilbert-base-uncased)
+* [elgeish/cs224n-squad2.0-roberta-base](https://huggingface.co/elgeish/cs224n-squad2.0-roberta-base)

model_cards/elgeish/cs224n-squad2.0-albert-large-v2/README.md

@@ -0,0 +1,74 @@
+## CS224n SQuAD2.0 Project Dataset
+The goal of this model is to save CS224n students GPU time when establising
+baselines to beat for the [Default Final Project](http://web.stanford.edu/class/cs224n/project/default-final-project-handout.pdf).
+The training set used to fine-tune this model is the same as
+the [official one](https://rajpurkar.github.io/SQuAD-explorer/); however,
+evaluation and model selection were performed using roughly half of the official
+dev set, 6078 examples, picked at random. The data files can be found at
+<https://github.com/elgeish/squad/tree/master/data> — this is the Winter 2020
+version. Given that the official SQuAD2.0 dev set contains the project's test
+set, students must make sure not to use the official SQuAD2.0 dev set in any way
+— including the use of models fine-tuned on the official SQuAD2.0, since they
+used the official SQuAD2.0 dev set for model selection.
+
+## Results
+```json
+{
+  "exact": 79.2694965449161,
+  "f1": 82.50844352970152,
+  "total": 6078,
+  "HasAns_exact": 74.87972508591065,
+  "HasAns_f1": 81.64478342732858,
+  "HasAns_total": 2910,
+  "NoAns_exact": 83.30176767676768,
+  "NoAns_f1": 83.30176767676768,
+  "NoAns_total": 3168,
+  "best_exact": 79.2694965449161,
+  "best_exact_thresh": 0.0,
+  "best_f1": 82.50844352970155,
+  "best_f1_thresh": 0.0
+}
+```
+
+## Notable Arguments
+```json
+{
+  "do_lower_case": true,
+  "doc_stride": 128,
+  "fp16": false,
+  "fp16_opt_level": "O1",
+  "gradient_accumulation_steps": 1,
+  "learning_rate": 3e-05,
+  "max_answer_length": 30,
+  "max_grad_norm": 1,
+  "max_query_length": 64,
+  "max_seq_length": 384,
+  "model_name_or_path": "albert-large-v2",
+  "model_type": "albert",
+  "num_train_epochs": 5,
+  "per_gpu_train_batch_size": 8,
+  "save_steps": 5000,
+  "seed": 42,
+  "train_batch_size": 8,
+  "version_2_with_negative": true,
+  "warmup_steps": 0,
+  "weight_decay": 0
+}
+```
+
+## Environment Setup
+```json
+{
+  "transformers": "2.5.1",
+  "pytorch": "1.4.0=py3.6_cuda10.1.243_cudnn7.6.3_0",
+  "python": "3.6.5=hc3d631a_2",
+  "os": "Linux 4.15.0-1060-aws #62-Ubuntu SMP Tue Feb 11 21:23:22 UTC 2020 x86_64 x86_64 x86_64 GNU/Linux",
+  "gpu": "Tesla V100-SXM2-16GB"
+}
+```
+
+## Related Models
+* [elgeish/cs224n-squad2.0-albert-base-v2](https://huggingface.co/elgeish/cs224n-squad2.0-albert-base-v2)
+* [elgeish/cs224n-squad2.0-albert-xxlarge-v1](https://huggingface.co/elgeish/cs224n-squad2.0-albert-xxlarge-v1)
+* [elgeish/cs224n-squad2.0-distilbert-base-uncased](https://huggingface.co/elgeish/cs224n-squad2.0-distilbert-base-uncased)
+* [elgeish/cs224n-squad2.0-roberta-base](https://huggingface.co/elgeish/cs224n-squad2.0-roberta-base)

model_cards/elgeish/cs224n-squad2.0-albert-xxlarge-v1/README.md

@@ -0,0 +1,74 @@
+## CS224n SQuAD2.0 Project Dataset
+The goal of this model is to save CS224n students GPU time when establising
+baselines to beat for the [Default Final Project](http://web.stanford.edu/class/cs224n/project/default-final-project-handout.pdf).
+The training set used to fine-tune this model is the same as
+the [official one](https://rajpurkar.github.io/SQuAD-explorer/); however,
+evaluation and model selection were performed using roughly half of the official
+dev set, 6078 examples, picked at random. The data files can be found at
+<https://github.com/elgeish/squad/tree/master/data> — this is the Winter 2020
+version. Given that the official SQuAD2.0 dev set contains the project's test
+set, students must make sure not to use the official SQuAD2.0 dev set in any way
+— including the use of models fine-tuned on the official SQuAD2.0, since they
+used the official SQuAD2.0 dev set for model selection.
+
+## Results
+```json
+{
+  "exact": 85.93287265547877,
+  "f1": 88.91258331187983,
+  "total": 6078,
+  "HasAns_exact": 84.36426116838489,
+  "HasAns_f1": 90.58786301361013,
+  "HasAns_total": 2910,
+  "NoAns_exact": 87.37373737373737,
+  "NoAns_f1": 87.37373737373737,
+  "NoAns_total": 3168,
+  "best_exact": 85.93287265547877,
+  "best_exact_thresh": 0.0,
+  "best_f1": 88.91258331187993,
+  "best_f1_thresh": 0.0
+}
+```
+
+## Notable Arguments
+```json
+{
+  "do_lower_case": true,
+  "doc_stride": 128,
+  "fp16": false,
+  "fp16_opt_level": "O1",
+  "gradient_accumulation_steps": 24,
+  "learning_rate": 3e-05,
+  "max_answer_length": 30,
+  "max_grad_norm": 1,
+  "max_query_length": 64,
+  "max_seq_length": 512,
+  "model_name_or_path": "albert-xxlarge-v1",
+  "model_type": "albert",
+  "num_train_epochs": 4,
+  "per_gpu_train_batch_size": 1,
+  "save_steps": 1000,
+  "seed": 42,
+  "train_batch_size": 1,
+  "version_2_with_negative": true,
+  "warmup_steps": 814,
+  "weight_decay": 0
+}
+```
+
+## Environment Setup
+```json
+{
+  "transformers": "2.5.1",
+  "pytorch": "1.4.0=py3.6_cuda10.1.243_cudnn7.6.3_0",
+  "python": "3.6.5=hc3d631a_2",
+  "os": "Linux 4.15.0-1060-aws #62-Ubuntu SMP Tue Feb 11 21:23:22 UTC 2020 x86_64 x86_64 x86_64 GNU/Linux",
+  "gpu": "Tesla V100-SXM2-16GB"
+}
+```
+
+## Related Models
+* [elgeish/cs224n-squad2.0-albert-base-v2](https://huggingface.co/elgeish/cs224n-squad2.0-albert-base-v2)
+* [elgeish/cs224n-squad2.0-albert-large-v2](https://huggingface.co/elgeish/cs224n-squad2.0-albert-large-v2)
+* [elgeish/cs224n-squad2.0-distilbert-base-uncased](https://huggingface.co/elgeish/cs224n-squad2.0-distilbert-base-uncased)
+* [elgeish/cs224n-squad2.0-roberta-base](https://huggingface.co/elgeish/cs224n-squad2.0-roberta-base)

model_cards/elgeish/cs224n-squad2.0-distilbert-base-uncased/README.md

@@ -0,0 +1,74 @@
+## CS224n SQuAD2.0 Project Dataset
+The goal of this model is to save CS224n students GPU time when establising
+baselines to beat for the [Default Final Project](http://web.stanford.edu/class/cs224n/project/default-final-project-handout.pdf).
+The training set used to fine-tune this model is the same as
+the [official one](https://rajpurkar.github.io/SQuAD-explorer/); however,
+evaluation and model selection were performed using roughly half of the official
+dev set, 6078 examples, picked at random. The data files can be found at
+<https://github.com/elgeish/squad/tree/master/data> — this is the Winter 2020
+version. Given that the official SQuAD2.0 dev set contains the project's test
+set, students must make sure not to use the official SQuAD2.0 dev set in any way
+— including the use of models fine-tuned on the official SQuAD2.0, since they
+used the official SQuAD2.0 dev set for model selection.
+
+## Results
+```json
+{
+  "exact": 65.16946363935504,
+  "f1": 67.87348075352251,
+  "total": 6078,
+  "HasAns_exact": 69.51890034364261,
+  "HasAns_f1": 75.16667217179045,
+  "HasAns_total": 2910,
+  "NoAns_exact": 61.17424242424242,
+  "NoAns_f1": 61.17424242424242,
+  "NoAns_total": 3168,
+  "best_exact": 65.16946363935504,
+  "best_exact_thresh": 0.0,
+  "best_f1": 67.87348075352243,
+  "best_f1_thresh": 0.0
+}
+```
+
+## Notable Arguments
+```json
+{
+  "do_lower_case": true,
+  "doc_stride": 128,
+  "fp16": false,
+  "fp16_opt_level": "O1",
+  "gradient_accumulation_steps": 24,
+  "learning_rate": 3e-05,
+  "max_answer_length": 30,
+  "max_grad_norm": 1,
+  "max_query_length": 64,
+  "max_seq_length": 384,
+  "model_name_or_path": "distilbert-base-uncased-distilled-squad",
+  "model_type": "distilbert",
+  "num_train_epochs": 4,
+  "per_gpu_train_batch_size": 32,
+  "save_steps": 5000,
+  "seed": 42,
+  "train_batch_size": 32,
+  "version_2_with_negative": true,
+  "warmup_steps": 0,
+  "weight_decay": 0
+}
+```
+
+## Environment Setup
+```json
+{
+  "transformers": "2.5.1",
+  "pytorch": "1.4.0=py3.6_cuda10.1.243_cudnn7.6.3_0",
+  "python": "3.6.5=hc3d631a_2",
+  "os": "Linux 4.15.0-1060-aws #62-Ubuntu SMP Tue Feb 11 21:23:22 UTC 2020 x86_64 x86_64 x86_64 GNU/Linux",
+  "gpu": "Tesla V100-SXM2-16GB"
+}
+```
+
+## Related Models
+* [elgeish/cs224n-squad2.0-albert-base-v2](https://huggingface.co/elgeish/cs224n-squad2.0-albert-base-v2)
+* [elgeish/cs224n-squad2.0-albert-large-v2](https://huggingface.co/elgeish/cs224n-squad2.0-albert-large-v2)
+* [elgeish/cs224n-squad2.0-albert-xxlarge-v1](https://huggingface.co/elgeish/cs224n-squad2.0-albert-xxlarge-v1)
+* [elgeish/cs224n-squad2.0-roberta-base](https://huggingface.co/elgeish/cs224n-squad2.0-roberta-base)

model_cards/elgeish/cs224n-squad2.0-roberta-base/README.md

@@ -0,0 +1,74 @@
+## CS224n SQuAD2.0 Project Dataset
+The goal of this model is to save CS224n students GPU time when establising
+baselines to beat for the [Default Final Project](http://web.stanford.edu/class/cs224n/project/default-final-project-handout.pdf).
+The training set used to fine-tune this model is the same as
+the [official one](https://rajpurkar.github.io/SQuAD-explorer/); however,
+evaluation and model selection were performed using roughly half of the official
+dev set, 6078 examples, picked at random. The data files can be found at
+<https://github.com/elgeish/squad/tree/master/data> — this is the Winter 2020
+version. Given that the official SQuAD2.0 dev set contains the project's test
+set, students must make sure not to use the official SQuAD2.0 dev set in any way
+— including the use of models fine-tuned on the official SQuAD2.0, since they
+used the official SQuAD2.0 dev set for model selection.
+
+## Results
+```json
+{
+  "exact": 75.32082922013821,
+  "f1": 78.66699523704254,
+  "total": 6078,
+  "HasAns_exact": 74.84536082474227,
+  "HasAns_f1": 81.83436324767868,
+  "HasAns_total": 2910,
+  "NoAns_exact": 75.75757575757575,
+  "NoAns_f1": 75.75757575757575,
+  "NoAns_total": 3168,
+  "best_exact": 75.32082922013821,
+  "best_exact_thresh": 0.0,
+  "best_f1": 78.66699523704266,
+  "best_f1_thresh": 0.0
+}
+```
+
+## Notable Arguments
+```json
+{
+  "do_lower_case": true,
+  "doc_stride": 128,
+  "fp16": false,
+  "fp16_opt_level": "O1",
+  "gradient_accumulation_steps": 24,
+  "learning_rate": 3e-05,
+  "max_answer_length": 30,
+  "max_grad_norm": 1,
+  "max_query_length": 64,
+  "max_seq_length": 384,
+  "model_name_or_path": "roberta-base",
+  "model_type": "roberta",
+  "num_train_epochs": 4,
+  "per_gpu_train_batch_size": 16,
+  "save_steps": 5000,
+  "seed": 42,
+  "train_batch_size": 16,
+  "version_2_with_negative": true,
+  "warmup_steps": 0,
+  "weight_decay": 0
+}
+```
+
+## Environment Setup
+```json
+{
+  "transformers": "2.5.1",
+  "pytorch": "1.4.0=py3.6_cuda10.1.243_cudnn7.6.3_0",
+  "python": "3.6.5=hc3d631a_2",
+  "os": "Linux 4.15.0-1060-aws #62-Ubuntu SMP Tue Feb 11 21:23:22 UTC 2020 x86_64 x86_64 x86_64 GNU/Linux",
+  "gpu": "Tesla V100-SXM2-16GB"
+}
+```
+
+## Related Models
+* [elgeish/cs224n-squad2.0-albert-base-v2](https://huggingface.co/elgeish/cs224n-squad2.0-albert-base-v2)
+* [elgeish/cs224n-squad2.0-albert-large-v2](https://huggingface.co/elgeish/cs224n-squad2.0-albert-large-v2)
+* [elgeish/cs224n-squad2.0-albert-xxlarge-v1](https://huggingface.co/elgeish/cs224n-squad2.0-albert-xxlarge-v1)
+* [elgeish/cs224n-squad2.0-distilbert-base-uncased](https://huggingface.co/elgeish/cs224n-squad2.0-distilbert-base-uncased)

model_cards/gsarti/biobert-nli/README.md

@@ -0,0 +1,37 @@
+# BioBERT-NLI
+
+This is the model [BioBERT](https://github.com/dmis-lab/biobert) [1] fine-tuned on the [SNLI](https://nlp.stanford.edu/projects/snli/) and the [MultiNLI](https://www.nyu.edu/projects/bowman/multinli/) datasets using the [`sentence-transformers` library](https://github.com/UKPLab/sentence-transformers/) to produce universal sentence embeddings [2].
+
+The model uses the original BERT wordpiece vocabulary and was trained using the **average pooling strategy** and a **softmax loss**.
+
+**Base model**: `monologg/biobert_v1.1_pubmed` from HuggingFace's `AutoModel`.
+
+**Training time**: ~6 hours on the NVIDIA Tesla P100 GPU provided in Kaggle Notebooks.
+
+**Parameters**:
+
+| Parameter        | Value |
+|------------------|-------|
+| Batch size       | 64    |
+| Training steps   | 30000 |
+| Warmup steps     | 1450  |
+| Lowercasing      | False |
+| Max. Seq. Length | 128   |
+
+**Performances**: The performance was evaluated on the test portion of the [STS dataset](http://ixa2.si.ehu.es/stswiki/index.php/STSbenchmark) using Spearman rank correlation and compared to the performances of a general BERT base model obtained with the same procedure to verify their similarity.
+
+| Model                         | Score       |
+|-------------------------------|-------------|
+| `biobert-nli` (this)          | 73.40       |
+| `gsarti/scibert-nli`          | 74.50       |
+| `bert-base-nli-mean-tokens`[3]| 77.12       |
+
+An example usage for similarity-based scientific paper retrieval is provided in the [Covid Papers Browser](https://github.com/gsarti/covid-papers-browser) repository.
+
+**References:**
+
+[1] J. Lee et al, [BioBERT: a pre-trained biomedical language representation model for biomedical text mining](https://academic.oup.com/bioinformatics/article/36/4/1234/5566506)
+
+[2] A. Conneau et al., [Supervised Learning of Universal Sentence Representations from Natural Language Inference Data](https://www.aclweb.org/anthology/D17-1070/)
+
+[3] N. Reimers et I. Gurevych, [Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks](https://www.aclweb.org/anthology/D19-1410/)

model_cards/gsarti/scibert-nli/README.md

@@ -2,26 +2,28 @@
 
 This is the model [SciBERT](https://github.com/allenai/scibert) [1] fine-tuned on the [SNLI](https://nlp.stanford.edu/projects/snli/) and the [MultiNLI](https://www.nyu.edu/projects/bowman/multinli/) datasets using the [`sentence-transformers` library](https://github.com/UKPLab/sentence-transformers/) to produce universal sentence embeddings [2].
 
-The model uses the original `scivocab` wordpiece vocabulary and was trained using the **average pooling strategy** and a **softmax loss**. 
+The model uses the original `scivocab` wordpiece vocabulary and was trained using the **average pooling strategy** and a **softmax loss**.
 
-**Base model**: `allenai/scibert-scivocab-cased` from HuggingFace AutoModel
+**Base model**: `allenai/scibert-scivocab-cased` from HuggingFace's `AutoModel`.
+
+**Training time**: ~4 hours on the NVIDIA Tesla P100 GPU provided in Kaggle Notebooks.
 
 **Parameters**:
 
-| Parameter      | Value |
-|----------------|-------|
-| Batch size     | 64    |
-| Training steps | 20000 |
-| Warmup steps   | 1450  |
+| Parameter        | Value |
+|------------------|-------|
+| Batch size       | 64    |
+| Training steps   | 20000 |
+| Warmup steps     | 1450  |
+| Lowercasing      | True  |
+| Max. Seq. Length | 128   |
 
 **Performances**: The performance was evaluated on the test portion of the [STS dataset](http://ixa2.si.ehu.es/stswiki/index.php/STSbenchmark) using Spearman rank correlation and compared to the performances of a general BERT base model obtained with the same procedure to verify their similarity.
 
-
-| Model                       | Score       |
-|-----------------------------|-------------|
-| `scibert-nli` (ours)        | 74.50       |
-| `bert-base-nli-mean-tokens` | 77.12       |
-
+| Model                         | Score       |
+|-------------------------------|-------------|
+| `scibert-nli` (this)          | 74.50       |
+| `bert-base-nli-mean-tokens`[3]| 77.12       |
 
 An example usage for similarity-based scientific paper retrieval is provided in the [Covid Papers Browser](https://github.com/gsarti/covid-papers-browser) repository.
 
@@ -30,3 +32,5 @@ An example usage for similarity-based scientific paper retrieval is provided in
 [1] I. Beltagy et al, [SciBERT: A Pretrained Language Model for Scientific Text](https://www.aclweb.org/anthology/D19-1371/)
 
 [2] A. Conneau et al., [Supervised Learning of Universal Sentence Representations from Natural Language Inference Data](https://www.aclweb.org/anthology/D17-1070/)
+
+[3] N. Reimers et I. Gurevych, [Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks](https://www.aclweb.org/anthology/D19-1410/)

model_cards/huseinzol05/bert-base-bahasa-cased/README.md

@@ -32,13 +32,54 @@ Preprocessing steps can reproduce from here, [Malaya/pretrained-model/preprocess
 You can use this model by installing `torch` or `tensorflow` and Huggingface library `transformers`. And you can use it directly by initializing it like this:  
 
 ```python
-from transformers import XLNetTokenizer, BertModel
+from transformers import AlbertTokenizer, BertModel
 
 model = BertModel.from_pretrained('huseinzol05/bert-base-bahasa-cased')
-tokenizer = XLNetTokenizer.from_pretrained('huseinzol05/bert-base-bahasa-cased')
+tokenizer = AlbertTokenizer.from_pretrained(
+    'huseinzol05/bert-base-bahasa-cased',
+    unk_token = '[UNK]',
+    pad_token = '[PAD]',
+    do_lower_case = False,
+)
 ```
 
-We use [google/sentencepiece](https://github.com/google/sentencepiece) to train the tokenizer, so to use it, need to load from `XLNetTokenizer`.
+We use [google/sentencepiece](https://github.com/google/sentencepiece) to train the tokenizer, so to use it, need to load from `AlbertTokenizer`.
+
+## Example using AutoModelWithLMHead
+
+```python
+from transformers import AlbertTokenizer, AutoModelWithLMHead, pipeline
+
+model = AutoModelWithLMHead.from_pretrained('huseinzol05/bert-base-bahasa-cased')
+tokenizer = AlbertTokenizer.from_pretrained(
+    'huseinzol05/bert-base-bahasa-cased',
+    unk_token = '[UNK]',
+    pad_token = '[PAD]',
+    do_lower_case = False,
+)
+fill_mask = pipeline('fill-mask', model = model, tokenizer = tokenizer)
+print(fill_mask('makan ayam dengan [MASK]'))
+```
+
+Output is,
+
+```text
+[{'sequence': '[CLS] makan ayam dengan rendang[SEP]',
+  'score': 0.10812027007341385,
+  'token': 2446},
+ {'sequence': '[CLS] makan ayam dengan kicap[SEP]',
+  'score': 0.07653367519378662,
+  'token': 12928},
+ {'sequence': '[CLS] makan ayam dengan nasi[SEP]',
+  'score': 0.06839974224567413,
+  'token': 450},
+ {'sequence': '[CLS] makan ayam dengan ayam[SEP]',
+  'score': 0.059544261544942856,
+  'token': 638},
+ {'sequence': '[CLS] makan ayam dengan sayur[SEP]',
+  'score': 0.05294966697692871,
+  'token': 1639}]
+```
 
 ## Results
 

model_cards/huseinzol05/xlnet-base-bahasa-cased/README.md

@@ -0,0 +1,64 @@
+---
+language: malay
+---
+
+# Bahasa XLNet Model
+
+Pretrained XLNet base language model for Malay and Indonesian. 
+
+## Pretraining Corpus
+
+`XLNET-base-bahasa-cased` model was pretrained on ~1.8 Billion words. We trained on both standard and social media language structures, and below is list of data we trained on,
+
+1. [dumping wikipedia](https://github.com/huseinzol05/Malaya-Dataset#wikipedia-1).
+2. [local instagram](https://github.com/huseinzol05/Malaya-Dataset#instagram).
+3. [local twitter](https://github.com/huseinzol05/Malaya-Dataset#twitter-1).
+4. [local news](https://github.com/huseinzol05/Malaya-Dataset#public-news).
+5. [local parliament text](https://github.com/huseinzol05/Malaya-Dataset#parliament).
+6. [local singlish/manglish text](https://github.com/huseinzol05/Malaya-Dataset#singlish-text).
+7. [IIUM Confession](https://github.com/huseinzol05/Malaya-Dataset#iium-confession).
+8. [Wattpad](https://github.com/huseinzol05/Malaya-Dataset#wattpad).
+9. [Academia PDF](https://github.com/huseinzol05/Malaya-Dataset#academia-pdf).
+
+Preprocessing steps can reproduce from here, [Malaya/pretrained-model/preprocess](https://github.com/huseinzol05/Malaya/tree/master/pretrained-model/preprocess).
+
+## Pretraining details
+
+- This model was trained using zihangdai XLNet's github [repository](https://github.com/zihangdai/xlnet) on 3 Titan V100 32GB VRAM.
+- All steps can reproduce from here, [Malaya/pretrained-model/xlnet](https://github.com/huseinzol05/Malaya/tree/master/pretrained-model/xlnet).
+
+## Load Pretrained Model
+
+You can use this model by installing `torch` or `tensorflow` and Huggingface library `transformers`. And you can use it directly by initializing it like this:  
+
+```python
+from transformers import XLNetTokenizer, XLNetModel
+
+model = XLNetModel.from_pretrained('huseinzol05/xlnet-base-bahasa-cased')
+tokenizer = XLNetTokenizer.from_pretrained(
+    'huseinzol05/xlnet-base-bahasa-cased', do_lower_case = False
+)
+```
+
+## Example using AutoModelWithLMHead
+
+```python
+from transformers import AlbertTokenizer, AutoModelWithLMHead, pipeline
+
+model = AutoModelWithLMHead.from_pretrained('huseinzol05/xlnet-base-bahasa-cased')
+tokenizer = XLNetTokenizer.from_pretrained(
+    'huseinzol05/xlnet-base-bahasa-cased', do_lower_case = False
+)
+fill_mask = pipeline('fill-mask', model = model, tokenizer = tokenizer)
+print(fill_mask('makan ayam dengan [MASK]'))
+```
+
+## Results
+
+For further details on the model performance, simply checkout accuracy page from Malaya, https://malaya.readthedocs.io/en/latest/Accuracy.html, we compared with traditional models.
+
+## Acknowledgement
+
+Thanks to [Im Big](https://www.facebook.com/imbigofficial/), [LigBlou](https://www.facebook.com/ligblou), [Mesolitica](https://mesolitica.com/) and [KeyReply](https://www.keyreply.com/) for sponsoring AWS, Google and GPU clouds to train XLNet for Bahasa. 
+
+

model_cards/mrm8488/GPT-2-finetuned-CORD19/README.md

@@ -0,0 +1,60 @@
+---
+language: english
+thumbnail:
+---
+
+# GPT-2 + CORD19 dataset : 🦠 ✍ ⚕
+
+**GPT-2** fine-tuned on **biorxiv_medrxiv** and **comm_use_subset files** from [CORD-19](https://www.kaggle.com/allen-institute-for-ai/CORD-19-research-challenge) dataset.
+
+
+## Datasets details:
+
+| Dataset                | # Files |
+| ---------------------- | ----- |
+| biorxiv_medrxiv        | 885  |
+| comm_use_subse         | 9K   |
+
+## Model training
+
+The model was trained on a Tesla P100 GPU and 25GB of RAM with the following command:
+
+```bash
+
+export TRAIN_FILE=/path/to/dataset/train.txt
+
+python run_language_modeling.py \
+    --model_type gpt2 \
+    --model_name_or_path gpt2 \
+    --do_train \
+    --train_data_file $TRAIN_FILE \
+    --num_train_epochs 4 \
+    --output_dir model_output \
+    --overwrite_output_dir \
+    --save_steps 10000 \
+    --per_gpu_train_batch_size 3
+```
+
+<img alt="training loss" src="https://svgshare.com/i/JTf.svg' title='GTP-2-finetuned-CORDS19-loss" width="600" height="300" />
+
+## Model in action / Example of usage: ✒
+
+You can get the following script [here](https://github.com/huggingface/transformers/blob/master/examples/run_generation.py)
+
+```bash
+python run_generation.py \
+    --model_type gpt2 \
+    --model_name_or_path mrm8488/GPT-2-finetuned-CORD19 \
+    --length 200
+```
+```txt
+# Input: the effects of COVID-19 on the lungs
+# Output: === GENERATED SEQUENCE 1 ===
+the effects of COVID-19 on the lungs are currently debated (86). The role of this virus in the pathogenesis of pneumonia and lung cancer is still debated. MERS-CoV is also known to cause acute respiratory distress syndrome (87) and is associated with increased expression of pulmonary fibrosis markers (88). Thus, early airway inflammation may play an important role in the pathogenesis of coronavirus pneumonia and may contribute to the severe disease and/or mortality observed in coronavirus patients.
+Pneumonia is an acute, often fatal disease characterized by severe edema, leakage of oxygen and bronchiolar inflammation. Viruses include coronaviruses, and the role of oxygen depletion is complicated by lung injury and fibrosis in the lung, in addition to susceptibility to other lung diseases. The progression of the disease may be variable, depending on the lung injury, pathologic role, prognosis, and the immune status of the patient. Inflammatory responses to respiratory viruses cause various pathologies of the respiratory
+```
+
+
+> Created by [Manuel Romero/@mrm8488](https://twitter.com/mrm8488) | [LinkedIn](https://www.linkedin.com/in/manuel-romero-cs/)
+
+> Made with <span style="color: #e25555;">&hearts;</span> in Spain

model_cards/mrm8488/bert-spanish-cased-finetuned-pos/README.md

@@ -5,7 +5,7 @@ thumbnail: https://i.imgur.com/jgBdimh.png
 
 # Spanish BERT (BETO) + POS
 
-This model is a fine-tuned on [NER-C](https://www.kaggle.com/nltkdata/conll-corpora) Of the Spanish BERT cased [(BETO)](https://github.com/dccuchile/beto) for **POS** (Part of Speech tagging) downstream task.
+This model is a fine-tuned on Spanish [CONLL CORPORA](https://www.kaggle.com/nltkdata/conll-corpora) version of the Spanish BERT cased [(BETO)](https://github.com/dccuchile/beto) for **POS** (Part of Speech tagging) downstream task.
 
 ## Details of the downstream task (POS) - Dataset
 

model_cards/mrm8488/xlm-multi-finetuned-xquadv1/README.md

@@ -93,8 +93,8 @@ from transformers import pipeline
 
 qa_pipeline = pipeline(
     "question-answering",
-    model="mrm8488/bert-multi-uncased-finetuned-xquadv1",
-    tokenizer="bert-multi-uncased-finetuned-xquadv1"
+    model="mrm8488/xlm-multi-finetuned-xquadv1",
+    tokenizer="mrm8488/xlm-multi-finetuned-xquadv1"
 )
 
 # English
@@ -114,7 +114,7 @@ qa_pipeline({
 
 #Output: {'answer': 'работал в репозитории hugginface /','end': 76, 'score': 0.00012340750456964894, 'start': 42}
 ```
-Try it on a Colab:
+Try it on a Colab (*Do not forget to change the model and tokenizer path in the Colab if necessary*):
 
 <a href="https://colab.research.google.com/github/mrm8488/shared_colab_notebooks/blob/master/Try_mrm8488_xquad_finetuned_uncased_model.ipynb" target="_parent"><img src="https://camo.githubusercontent.com/52feade06f2fecbf006889a904d221e6a730c194/68747470733a2f2f636f6c61622e72657365617263682e676f6f676c652e636f6d2f6173736574732f636f6c61622d62616467652e737667" alt="Open In Colab" data-canonical-src="https://colab.research.google.com/assets/colab-badge.svg"></a>
 

model_cards/twmkn9/albert-base-v2-squad2/README.md

@@ -1,22 +1,24 @@
-This model is ALBERT base v2 trained on SQuAD v2 as:
+This model is [ALBERT base v2](https://huggingface.co/albert-base-v2) trained on SQuAD v2 as:
 
 ```
-python run_squad.py 
---model_type albert 
---model_name_or_path albert-base-v2 
---do_train 
---do_eval 
---overwrite_cache 
---do_lower_case 
---version_2_with_negative 
---train_file $SQUAD_DIR/train-v2.0.json 
---predict_file $SQUAD_DIR/dev-v2.0.json 
---per_gpu_train_batch_size 8 
---num_train_epochs 3 
---learning_rate 3e-5 
---max_seq_length 384 
---doc_stride 128 
---output_dir ./tmp/albert_base_fine/
+export SQUAD_DIR=../../squad2
+python3 run_squad.py 
+    --model_type albert 
+    --model_name_or_path albert-base-v2 
+    --do_train 
+    --do_eval 
+    --overwrite_cache 
+    --do_lower_case 
+    --version_2_with_negative 
+    --save_steps 100000 
+    --train_file $SQUAD_DIR/train-v2.0.json 
+    --predict_file $SQUAD_DIR/dev-v2.0.json 
+    --per_gpu_train_batch_size 8 
+    --num_train_epochs 3 
+    --learning_rate 3e-5 
+    --max_seq_length 384 
+    --doc_stride 128 
+    --output_dir ./tmp/albert_fine/
 ```
 
 Performance on a dev subset is close to the original paper:

model_cards/twmkn9/bert-base-uncased-squad2/README.md

@@ -1,22 +1,24 @@
-This model is BERT base uncased trained on SQuAD v2 as:
+This model is [BERT base uncased](https://huggingface.co/bert-base-uncased) trained on SQuAD v2 as:
 
 ```
-python run_squad.py 
---model_type bert 
---model_name_or_path bert-base-uncased
---do_train 
---do_eval 
---overwrite_cache 
---do_lower_case 
---version_2_with_negative 
---train_file $SQUAD_DIR/train-v2.0.json 
---predict_file $SQUAD_DIR/dev-v2.0.json 
---per_gpu_train_batch_size 8 
---num_train_epochs 3 
---learning_rate 3e-5 
---max_seq_length 384 
---doc_stride 128 
---output_dir ./tmp/bert_base_fine/
+export SQUAD_DIR=../../squad2
+python3 run_squad.py 
+    --model_type bert 
+    --model_name_or_path bert-base-uncased 
+    --do_train 
+    --do_eval 
+    --overwrite_cache 
+    --do_lower_case 
+    --version_2_with_negative 
+    --save_steps 100000 
+    --train_file $SQUAD_DIR/train-v2.0.json 
+    --predict_file $SQUAD_DIR/dev-v2.0.json 
+    --per_gpu_train_batch_size 8 
+    --num_train_epochs 3 
+    --learning_rate 3e-5 
+    --max_seq_length 384 
+    --doc_stride 128 
+    --output_dir ./tmp/bert_fine_tuned/
 ```
 
 Performance on a dev subset is close to the original paper:

model_cards/twmkn9/distilbert-base-uncased-squad2/README.md

@@ -0,0 +1,45 @@
+This model is [Distilbert base uncased](https://huggingface.co/distilbert-base-uncased) trained on SQuAD v2 as:
+
+```
+export SQUAD_DIR=../../squad2
+python3 run_squad.py 
+    --model_type distilbert 
+    --model_name_or_path distilbert-base-uncased
+    --do_train 
+    --do_eval 
+    --overwrite_cache 
+    --do_lower_case 
+    --version_2_with_negative 
+    --save_steps 100000 
+    --train_file $SQUAD_DIR/train-v2.0.json 
+    --predict_file $SQUAD_DIR/dev-v2.0.json 
+    --per_gpu_train_batch_size 8 
+    --num_train_epochs 3 
+    --learning_rate 3e-5 
+    --max_seq_length 384 
+    --doc_stride 128 
+    --output_dir ./tmp/distilbert_fine_tuned/
+```
+
+Performance on a dev subset is close to the original paper:
+
+```
+Results: 
+{
+    'exact': 64.88976637051661, 
+    'f1': 68.1776176526635, 
+    'total': 6078, 
+    'HasAns_exact': 69.7594501718213, 
+    'HasAns_f1': 76.62665295288285, 
+    'HasAns_total': 2910, 
+    'NoAns_exact': 60.416666666666664, 
+    'NoAns_f1': 60.416666666666664, 
+    'NoAns_total': 3168, 
+    'best_exact': 64.88976637051661, 
+    'best_exact_thresh': 0.0, 
+    'best_f1': 68.17761765266337, 
+    'best_f1_thresh': 0.0
+}
+```
+
+We are hopeful this might save you time, energy, and compute. Cheers!

model_cards/twmkn9/distilroberta-base-squad2/README.md

@@ -0,0 +1,44 @@
+This model is [Distilroberta base](https://huggingface.co/distilroberta-base) trained on SQuAD v2 as:
+
+```
+export SQUAD_DIR=../../squad2
+python3 run_squad.py 
+    --model_type robberta 
+    --model_name_or_path distilroberta-base 
+    --do_train 
+    --do_eval 
+    --overwrite_cache 
+    --do_lower_case 
+    --version_2_with_negative 
+    --save_steps 100000 
+    --train_file $SQUAD_DIR/train-v2.0.json 
+    --predict_file $SQUAD_DIR/dev-v2.0.json 
+    --per_gpu_train_batch_size 8 
+    --num_train_epochs 3 
+    --learning_rate 3e-5 
+    --max_seq_length 384 
+    --doc_stride 128 
+    --output_dir ./tmp/distilroberta_fine_tuned/
+```
+
+Performance on a dev subset is close to the original paper:
+
+```
+Results: 
+{
+    'exact': 70.9279368213228, 
+    'f1': 74.60439802429168, 
+    'total': 6078, 
+    'HasAns_exact': 67.62886597938144, 
+    'HasAns_f1': 75.30774267754136, 
+    'HasAns_total': 2910, 
+    'NoAns_exact': 73.95833333333333, 
+    'NoAns_f1': 73.95833333333333, 'NoAns_total': 3168, 
+    'best_exact': 70.94438960184272, 
+    'best_exact_thresh': 0.0, 
+    'best_f1': 74.62085080481161, 
+    'best_f1_thresh': 0.0
+}
+```
+
+We are hopeful this might save you time, energy, and compute. Cheers!

setup.py

@@ -64,7 +64,7 @@ if stale_egg_info.exists():
 extras = {}
 
 extras["mecab"] = ["mecab-python3"]
-extras["sklearn"] = ["scikit-learn==0.22.1"]
+extras["sklearn"] = ["scikit-learn"]
 extras["tf"] = ["tensorflow"]
 extras["tf-cpu"] = ["tensorflow-cpu"]
 extras["torch"] = ["torch"]
@@ -83,7 +83,7 @@ extras["dev"] = extras["testing"] + extras["quality"] + ["mecab-python3", "sciki
 
 setup(
     name="transformers",
-    version="2.5.1",
+    version="2.6.0",
     author="Thomas Wolf, Lysandre Debut, Victor Sanh, Julien Chaumond, Sam Shleifer, Google AI Language Team Authors, Open AI team Authors, Facebook AI Authors, Carnegie Mellon University Authors",
     author_email="thomas@huggingface.co",
     description="State-of-the-art Natural Language Processing for TensorFlow 2.0 and PyTorch",
@@ -97,6 +97,8 @@ setup(
     install_requires=[
         "numpy",
         "tokenizers == 0.5.2",
+        # dataclasses for Python versions that don't have it
+        "dataclasses;python_version<'3.7'",
         # accessing files from S3 directly
         "boto3",
         # filesystem locks e.g. to prevent parallel downloads

src/transformers/__init__.py

@@ -2,7 +2,7 @@
 # There's no way to ignore "F401 '...' imported but unused" warnings in this
 # module, but to preserve other warnings. So, don't check this module at all.
 
-__version__ = "2.5.1"
+__version__ = "2.6.0"
 
 # Work around to update TensorFlow's absl.logging threshold which alters the
 # default Python logging output behavior when present.
@@ -32,7 +32,7 @@ from .benchmark_utils import (
     stop_memory_tracing,
 )
 from .configuration_albert import ALBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, AlbertConfig
-from .configuration_auto import ALL_PRETRAINED_CONFIG_ARCHIVE_MAP, AutoConfig
+from .configuration_auto import ALL_PRETRAINED_CONFIG_ARCHIVE_MAP, CONFIG_MAPPING, AutoConfig
 from .configuration_bart import BartConfig
 from .configuration_bert import BERT_PRETRAINED_CONFIG_ARCHIVE_MAP, BertConfig
 from .configuration_camembert import CAMEMBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, CamembertConfig
@@ -116,10 +116,11 @@ from .pipelines import (
     SummarizationPipeline,
     TextClassificationPipeline,
     TokenClassificationPipeline,
+    TranslationPipeline,
     pipeline,
 )
 from .tokenization_albert import AlbertTokenizer
-from .tokenization_auto import AutoTokenizer
+from .tokenization_auto import TOKENIZER_MAPPING, AutoTokenizer
 from .tokenization_bart import BartTokenizer
 from .tokenization_bert import BasicTokenizer, BertTokenizer, BertTokenizerFast, WordpieceTokenizer
 from .tokenization_bert_japanese import BertJapaneseTokenizer, CharacterTokenizer, MecabTokenizer
@@ -158,6 +159,12 @@ if is_torch_available():
         AutoModelWithLMHead,
         AutoModelForTokenClassification,
         ALL_PRETRAINED_MODEL_ARCHIVE_MAP,
+        MODEL_MAPPING,
+        MODEL_FOR_PRETRAINING_MAPPING,
+        MODEL_WITH_LM_HEAD_MAPPING,
+        MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING,
+        MODEL_FOR_QUESTION_ANSWERING_MAPPING,
+        MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING,
     )
 
     from .modeling_bert import (
@@ -215,6 +222,7 @@ if is_torch_available():
         XLMModel,
         XLMWithLMHeadModel,
         XLMForSequenceClassification,
+        XLMForTokenClassification,
         XLMForQuestionAnswering,
         XLMForQuestionAnsweringSimple,
         XLM_PRETRAINED_MODEL_ARCHIVE_MAP,
@@ -317,6 +325,12 @@ if is_tf_available():
         TFAutoModelWithLMHead,
         TFAutoModelForTokenClassification,
         TF_ALL_PRETRAINED_MODEL_ARCHIVE_MAP,
+        TF_MODEL_MAPPING,
+        TF_MODEL_FOR_PRETRAINING_MAPPING,
+        TF_MODEL_WITH_LM_HEAD_MAPPING,
+        TF_MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING,
+        TF_MODEL_FOR_QUESTION_ANSWERING_MAPPING,
+        TF_MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING,
     )
 
     from .modeling_tf_bert import (

src/transformers/bart_mem_prof.py

@@ -1,67 +0,0 @@
-from transformers import *
-import torch
-DEFAULT_DEVICE = 'cuda' if torch.cuda.is_available() else 'cpu'
-def runner(source_path, out_file, batch_size=8, device=DEFAULT_DEVICE, prof_generate=False):
-
-    tokenizer = BartTokenizer.from_pretrained('bart-large')
-    lns = [" " + x.rstrip() for x in open(source_path).readlines()][:batch_size]
-
-    dct = tokenizer.batch_encode_plus(lns, max_length=1024, return_tensors="pt", pad_to_max_length=True)
-    ids = dct['input_ids'].to(DEFAULT_DEVICE)
-    msk = dct['attention_mask'].to(DEFAULT_DEVICE)
-    model = BartForConditionalGeneration.from_pretrained('bart-large-cnn', output_past=prof_generate).to(DEFAULT_DEVICE)
-    model.log_mem('starting')
-    if prof_generate:
-
-        summaries = model.generate(
-            input_ids=ids,
-            attention_mask=msk,
-            num_beams=4,
-            length_penalty=2.0,
-            max_length=140 + 2,  # +2 from original because we start at step=1 and stop before max_length
-            min_length=55 + 1,  # +1 from original because we start at step=1
-            no_repeat_ngram_size=3,
-            early_stopping=True,
-            do_sample=False,
-            decoder_start_token_id=model.config.eos_token_ids[0],
-        )
-        model.log_mem('done')
-        dec = [tokenizer.decode(s) for s in summaries]
-        print(dec[0])
-    else:
-        #model.decoder.generation_mode = Fals
-        with torch.no_grad():
-            model(
-                input_ids=ids,
-                attention_mask=msk,
-            )
-
-    log_df = model.combine_logs()
-    log_df.to_csv(out_file)
-
-
-import argparse
-if __name__ == '__main__':
-    parser = argparse.ArgumentParser()
-
-    parser.add_argument(
-        "output_path", type=str, help="where to save summaries",
-    )
-    parser.add_argument(
-        "--source_path", type=str, default="/home/shleifer/transformers_fork/notebooks/test.source",
-        help="like cnn_dm/test.source", required=False
-    )
-    parser.add_argument(
-        "--device", type=str, required=False, default=DEFAULT_DEVICE, help="cuda, cuda:1, cpu etc.",
-    )
-    parser.add_argument(
-        "--bs", type=int, default=8, required=False, help="batch size: how many to summarize at a time",
-    )
-    parser.add_argument(
-        "--do-generate", action='store_true', required=False, help="batch size: how many to summarize at a time",
-    )
-    args = parser.parse_args()
-    runner(args.source_path, args.output_path, batch_size=args.bs, device=args.device, prof_generate=args.do_generate)
-
-
-

src/transformers/configuration_albert.py

@@ -19,10 +19,10 @@ from .configuration_utils import PretrainedConfig
 
 
 ALBERT_PRETRAINED_CONFIG_ARCHIVE_MAP = {
-    "albert-base-v1": "https://s3.amazonaws.com/models.huggingface.co/bert/albert-base-config.json",
-    "albert-large-v1": "https://s3.amazonaws.com/models.huggingface.co/bert/albert-large-config.json",
-    "albert-xlarge-v1": "https://s3.amazonaws.com/models.huggingface.co/bert/albert-xlarge-config.json",
-    "albert-xxlarge-v1": "https://s3.amazonaws.com/models.huggingface.co/bert/albert-xxlarge-config.json",
+    "albert-base-v1": "https://s3.amazonaws.com/models.huggingface.co/bert/albert-base-v1-config.json",
+    "albert-large-v1": "https://s3.amazonaws.com/models.huggingface.co/bert/albert-large-v1-config.json",
+    "albert-xlarge-v1": "https://s3.amazonaws.com/models.huggingface.co/bert/albert-xlarge-v1-config.json",
+    "albert-xxlarge-v1": "https://s3.amazonaws.com/models.huggingface.co/bert/albert-xxlarge-v1-config.json",
     "albert-base-v2": "https://s3.amazonaws.com/models.huggingface.co/bert/albert-base-v2-config.json",
     "albert-large-v2": "https://s3.amazonaws.com/models.huggingface.co/bert/albert-large-v2-config.json",
     "albert-xlarge-v2": "https://s3.amazonaws.com/models.huggingface.co/bert/albert-xlarge-v2-config.json",

src/transformers/configuration_utils.py

@@ -78,9 +78,6 @@ class PretrainedConfig(object):
         self.top_k = kwargs.pop("top_k", 50)
         self.top_p = kwargs.pop("top_p", 1.0)
         self.repetition_penalty = kwargs.pop("repetition_penalty", 1.0)
-        self.bos_token_id = kwargs.pop("bos_token_id", None)
-        self.pad_token_id = kwargs.pop("pad_token_id", None)
-        self.eos_token_id = kwargs.pop("eos_token_id", None)
         self.length_penalty = kwargs.pop("length_penalty", 1.0)
         self.no_repeat_ngram_size = kwargs.pop("no_repeat_ngram_size", 0)
         self.num_return_sequences = kwargs.pop("num_return_sequences", 1)
@@ -94,6 +91,16 @@ class PretrainedConfig(object):
         self.label2id = kwargs.pop("label2id", dict(zip(self.id2label.values(), self.id2label.keys())))
         self.label2id = dict((key, int(value)) for key, value in self.label2id.items())
 
+        # Tokenizer arguments TODO: eventually tokenizer and models should share the same config
+        self.prefix = kwargs.pop("prefix", None)
+        self.bos_token_id = kwargs.pop("bos_token_id", None)
+        self.pad_token_id = kwargs.pop("pad_token_id", None)
+        self.eos_token_id = kwargs.pop("eos_token_id", None)
+        self.decoder_start_token_id = kwargs.pop("decoder_start_token_id", None)
+
+        # task specific arguments
+        self.task_specific_params = kwargs.pop("task_specific_params", None)
+
         # Additional attributes without default values
         for key, value in kwargs.items():
             try:
@@ -373,3 +380,14 @@ class PretrainedConfig(object):
         """
         with open(json_file_path, "w", encoding="utf-8") as writer:
             writer.write(self.to_json_string())
+
+    def update(self, config_dict: Dict):
+        """
+        Updates attributes of this class
+        with attributes from `config_dict`.
+
+        Args:
+            :obj:`Dict[str, any]`: Dictionary of attributes that shall be updated for this class.
+        """
+        for key, value in config_dict.items():
+            setattr(self, key, value)

src/transformers/data/processors/squad.py

@@ -139,6 +139,7 @@ def squad_convert_example_to_features(example, max_seq_length, doc_stride, max_q
             pad_to_max_length=True,
             stride=max_seq_length - doc_stride - len(truncated_query) - sequence_pair_added_tokens,
             truncation_strategy="only_second" if tokenizer.padding_side == "right" else "only_first",
+            return_token_type_ids=True,
         )
 
         paragraph_len = min(

src/transformers/data/processors/utils.py

@@ -16,8 +16,11 @@
 
 import copy
 import csv
+import dataclasses
 import json
 import logging
+from dataclasses import dataclass
+from typing import Optional
 
 from ...file_utils import is_tf_available, is_torch_available
 
@@ -25,7 +28,8 @@ from ...file_utils import is_tf_available, is_torch_available
 logger = logging.getLogger(__name__)
 
 
-class InputExample(object):
+@dataclass(frozen=True)
+class InputExample:
     """
     A single training/test example for simple sequence classification.
 
@@ -39,23 +43,14 @@ class InputExample(object):
             specified for train and dev examples, but not for test examples.
     """
 
-    def __init__(self, guid, text_a, text_b=None, label=None):
-        self.guid = guid
-        self.text_a = text_a
-        self.text_b = text_b
-        self.label = label
-
-    def __repr__(self):
-        return str(self.to_json_string())
-
-    def to_dict(self):
-        """Serializes this instance to a Python dictionary."""
-        output = copy.deepcopy(self.__dict__)
-        return output
+    guid: str
+    text_a: str
+    text_b: Optional[str] = None
+    label: Optional[str] = None
 
     def to_json_string(self):
         """Serializes this instance to a JSON string."""
-        return json.dumps(self.to_dict(), indent=2, sort_keys=True) + "\n"
+        return json.dumps(dataclasses.asdict(self), indent=2, sort_keys=True) + "\n"
 
 
 class InputFeatures(object):

src/transformers/modeling_albert.py

@@ -33,10 +33,10 @@ logger = logging.getLogger(__name__)
 
 
 ALBERT_PRETRAINED_MODEL_ARCHIVE_MAP = {
-    "albert-base-v1": "https://s3.amazonaws.com/models.huggingface.co/bert/albert-base-pytorch_model.bin",
-    "albert-large-v1": "https://s3.amazonaws.com/models.huggingface.co/bert/albert-large-pytorch_model.bin",
-    "albert-xlarge-v1": "https://s3.amazonaws.com/models.huggingface.co/bert/albert-xlarge-pytorch_model.bin",
-    "albert-xxlarge-v1": "https://s3.amazonaws.com/models.huggingface.co/bert/albert-xxlarge-pytorch_model.bin",
+    "albert-base-v1": "https://s3.amazonaws.com/models.huggingface.co/bert/albert-base-v1-pytorch_model.bin",
+    "albert-large-v1": "https://s3.amazonaws.com/models.huggingface.co/bert/albert-large-v1-pytorch_model.bin",
+    "albert-xlarge-v1": "https://s3.amazonaws.com/models.huggingface.co/bert/albert-xlarge-v1-pytorch_model.bin",
+    "albert-xxlarge-v1": "https://s3.amazonaws.com/models.huggingface.co/bert/albert-xxlarge-v1-pytorch_model.bin",
     "albert-base-v2": "https://s3.amazonaws.com/models.huggingface.co/bert/albert-base-v2-pytorch_model.bin",
     "albert-large-v2": "https://s3.amazonaws.com/models.huggingface.co/bert/albert-large-v2-pytorch_model.bin",
     "albert-xlarge-v2": "https://s3.amazonaws.com/models.huggingface.co/bert/albert-xlarge-v2-pytorch_model.bin",

src/transformers/modeling_auto.py

@@ -99,6 +99,7 @@ from .modeling_xlm import (
     XLM_PRETRAINED_MODEL_ARCHIVE_MAP,
     XLMForQuestionAnsweringSimple,
     XLMForSequenceClassification,
+    XLMForTokenClassification,
     XLMModel,
     XLMWithLMHeadModel,
 )
@@ -235,6 +236,7 @@ MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING = OrderedDict(
     [
         (DistilBertConfig, DistilBertForTokenClassification),
         (CamembertConfig, CamembertForTokenClassification),
+        (XLMConfig, XLMForTokenClassification),
         (XLMRobertaConfig, XLMRobertaForTokenClassification),
         (RobertaConfig, RobertaForTokenClassification),
         (BertConfig, BertForTokenClassification),
@@ -418,12 +420,12 @@ class AutoModelForPreTraining(object):
             config (:class:`~transformers.PretrainedConfig`):
                 The model class to instantiate is selected based on the configuration class:
 
-                - isInstance of `distilbert` configuration class: :class:`~transformers.DistilBertModelForMaskedLM` (DistilBERT model)
-                - isInstance of `roberta` configuration class: :class:`~transformers.RobertaModelForMaskedLM` (RoBERTa model)
+                - isInstance of `distilbert` configuration class: :class:`~transformers.DistilBertForMaskedLM` (DistilBERT model)
+                - isInstance of `roberta` configuration class: :class:`~transformers.RobertaForMaskedLM` (RoBERTa model)
                 - isInstance of `bert` configuration class: :class:`~transformers.BertForPreTraining` (Bert model)
                 - isInstance of `openai-gpt` configuration class: :class:`~transformers.OpenAIGPTLMHeadModel` (OpenAI GPT model)
-                - isInstance of `gpt2` configuration class: :class:`~transformers.GPT2ModelLMHeadModel` (OpenAI GPT-2 model)
-                - isInstance of `ctrl` configuration class: :class:`~transformers.CTRLModelLMHeadModel` (Salesforce CTRL  model)
+                - isInstance of `gpt2` configuration class: :class:`~transformers.GPT2LMHeadModel` (OpenAI GPT-2 model)
+                - isInstance of `ctrl` configuration class: :class:`~transformers.CTRLLMHeadModel` (Salesforce CTRL  model)
                 - isInstance of `transfo-xl` configuration class: :class:`~transformers.TransfoXLLMHeadModel` (Transformer-XL model)
                 - isInstance of `xlnet` configuration class: :class:`~transformers.XLNetLMHeadModel` (XLNet model)
                 - isInstance of `xlm` configuration class: :class:`~transformers.XLMWithLMHeadModel` (XLM model)
@@ -559,12 +561,12 @@ class AutoModelWithLMHead(object):
             config (:class:`~transformers.PretrainedConfig`):
                 The model class to instantiate is selected based on the configuration class:
 
-                - isInstance of `distilbert` configuration class: :class:`~transformers.DistilBertModelForMaskedLM` (DistilBERT model)
-                - isInstance of `roberta` configuration class: :class:`~transformers.RobertaModelForMaskedLM` (RoBERTa model)
-                - isInstance of `bert` configuration class: :class:`~transformers.BertModelForMaskedLM` (Bert model)
+                - isInstance of `distilbert` configuration class: :class:`~transformers.DistilBertForMaskedLM` (DistilBERT model)
+                - isInstance of `roberta` configuration class: :class:`~transformers.RobertaForMaskedLM` (RoBERTa model)
+                - isInstance of `bert` configuration class: :class:`~transformers.BertForMaskedLM` (Bert model)
                 - isInstance of `openai-gpt` configuration class: :class:`~transformers.OpenAIGPTLMHeadModel` (OpenAI GPT model)
-                - isInstance of `gpt2` configuration class: :class:`~transformers.GPT2ModelLMHeadModel` (OpenAI GPT-2 model)
-                - isInstance of `ctrl` configuration class: :class:`~transformers.CTRLModelLMHeadModel` (Salesforce CTRL  model)
+                - isInstance of `gpt2` configuration class: :class:`~transformers.GPT2LMHeadModel` (OpenAI GPT-2 model)
+                - isInstance of `ctrl` configuration class: :class:`~transformers.CTRLLMHeadModel` (Salesforce CTRL  model)
                 - isInstance of `transfo-xl` configuration class: :class:`~transformers.TransfoXLLMHeadModel` (Transformer-XL model)
                 - isInstance of `xlnet` configuration class: :class:`~transformers.XLNetLMHeadModel` (XLNet model)
                 - isInstance of `xlm` configuration class: :class:`~transformers.XLMWithLMHeadModel` (XLM model)
@@ -701,14 +703,14 @@ class AutoModelForSequenceClassification(object):
             config (:class:`~transformers.PretrainedConfig`):
                 The model class to instantiate is selected based on the configuration class:
 
-                - isInstance of `distilbert` configuration class: :class:`~transformers.DistilBertModelForSequenceClassification` (DistilBERT model)
-                - isInstance of `albert` configuration class: :class:`~transformers.AlbertModelForSequenceClassification` (ALBERT model)
-                - isInstance of `camembert` configuration class: :class:`~transformers.CamembertModelForSequenceClassification` (CamemBERT model)
-                - isInstance of `xlm roberta` configuration class: :class:`~transformers.XLMRobertaModelForSequenceClassification` (XLM-RoBERTa model)
-                - isInstance of `roberta` configuration class: :class:`~transformers.RobertaModelForSequenceClassification` (RoBERTa model)
-                - isInstance of `bert` configuration class: :class:`~transformers.BertModelForSequenceClassification` (Bert model)
-                - isInstance of `xlnet` configuration class: :class:`~transformers.XLNetModelForSequenceClassification` (XLNet model)
-                - isInstance of `xlm` configuration class: :class:`~transformers.XLMModelForSequenceClassification` (XLM model)
+                - isInstance of `distilbert` configuration class: :class:`~transformers.DistilBertForSequenceClassification` (DistilBERT model)
+                - isInstance of `albert` configuration class: :class:`~transformers.AlbertForSequenceClassification` (ALBERT model)
+                - isInstance of `camembert` configuration class: :class:`~transformers.CamembertForSequenceClassification` (CamemBERT model)
+                - isInstance of `xlm roberta` configuration class: :class:`~transformers.XLMRobertaForSequenceClassification` (XLM-RoBERTa model)
+                - isInstance of `roberta` configuration class: :class:`~transformers.RobertaForSequenceClassification` (RoBERTa model)
+                - isInstance of `bert` configuration class: :class:`~transformers.BertForSequenceClassification` (Bert model)
+                - isInstance of `xlnet` configuration class: :class:`~transformers.XLNetForSequenceClassification` (XLNet model)
+                - isInstance of `xlm` configuration class: :class:`~transformers.XLMForSequenceClassification` (XLM model)
                 - isInstance of `flaubert` configuration class: :class:`~transformers.FlaubertForSequenceClassification` (Flaubert model)
 
 
@@ -848,11 +850,11 @@ class AutoModelForQuestionAnswering(object):
             config (:class:`~transformers.PretrainedConfig`):
                 The model class to instantiate is selected based on the configuration class:
 
-                - isInstance of `distilbert` configuration class: :class:`~transformers.DistilBertModelForQuestionAnswering` (DistilBERT model)
-                - isInstance of `albert` configuration class: :class:`~transformers.AlbertModelForQuestionAnswering` (ALBERT model)
+                - isInstance of `distilbert` configuration class: :class:`~transformers.DistilBertForQuestionAnswering` (DistilBERT model)
+                - isInstance of `albert` configuration class: :class:`~transformers.AlbertForQuestionAnswering` (ALBERT model)
                 - isInstance of `bert` configuration class: :class:`~transformers.BertModelForQuestionAnswering` (Bert model)
-                - isInstance of `xlnet` configuration class: :class:`~transformers.XLNetModelForQuestionAnswering` (XLNet model)
-                - isInstance of `xlm` configuration class: :class:`~transformers.XLMModelForQuestionAnswering` (XLM model)
+                - isInstance of `xlnet` configuration class: :class:`~transformers.XLNetForQuestionAnswering` (XLNet model)
+                - isInstance of `xlm` configuration class: :class:`~transformers.XLMForQuestionAnswering` (XLM model)
                 - isInstance of `flaubert` configuration class: :class:`~transformers.FlaubertForQuestionAnswering` (XLM model)
 
         Examples::
@@ -989,8 +991,10 @@ class AutoModelForTokenClassification:
                 The model class to instantiate is selected based on the configuration class:
 
                 - isInstance of `distilbert` configuration class: :class:`~transformers.DistilBertModelForTokenClassification` (DistilBERT model)
+                - isInstance of `xlm` configuration class: :class:`~transformers.XLMForTokenClassification` (XLM model)
                 - isInstance of `xlm roberta` configuration class: :class:`~transformers.XLMRobertaModelForTokenClassification` (XLMRoberta model)
                 - isInstance of `bert` configuration class: :class:`~transformers.BertModelForTokenClassification` (Bert model)
+                - isInstance of `albert` configuration class: :class:`~transformers.AlbertForTokenClassification` (AlBert model)
                 - isInstance of `xlnet` configuration class: :class:`~transformers.XLNetModelForTokenClassification` (XLNet model)
                 - isInstance of `camembert` configuration class: :class:`~transformers.CamembertModelForTokenClassification` (Camembert model)
                 - isInstance of `roberta` configuration class: :class:`~transformers.RobertaModelForTokenClassification` (Roberta model)
@@ -1025,6 +1029,7 @@ class AutoModelForTokenClassification:
         The model class to instantiate is selected as the first pattern matching
         in the `pretrained_model_name_or_path` string (in the following order):
             - contains `distilbert`: :class:`~transformers.DistilBertForTokenClassification` (DistilBERT model)
+            - contains `xlm`: :class:`~transformers.XLMForTokenClassification` (XLM model)
             - contains `xlm-roberta`: :class:`~transformers.XLMRobertaForTokenClassification` (XLM-RoBERTa?Para model)
             - contains `camembert`: :class:`~transformers.CamembertForTokenClassification` (Camembert model)
             - contains `bert`: :class:`~transformers.BertForTokenClassification` (Bert model)

src/transformers/modeling_bart.py

@@ -25,8 +25,7 @@ from .activations import ACT2FN
 from .configuration_bart import BartConfig
 from .file_utils import add_start_docstrings, add_start_docstrings_to_callable
 from .modeling_utils import PreTrainedModel, create_position_ids_from_input_ids
-from durbango.logging_utils import LoggingMixin
-from durbango.torch_utils import print_tensor_sizes, local_sizeof, get_tensor_shapes_and_pointers
+
 
 logger = logging.getLogger(__name__)
 
@@ -73,47 +72,50 @@ BART_INPUTS_DOCSTRING = r"""
             Mask to avoid performing attention on padding token indices in input_ids.
             Mask values selected in ``[0, 1]``:
             ``1`` for tokens that are NOT MASKED, ``0`` for MASKED tokens.
+        encoder_outputs (tuple(:obj:`tuple(torch.FloatTensor)`, `optional`, defaults to :obj:`None`):
+            Tuple consists of (`last_hidden_state`, `optional`: `hidden_states`, `optional`: `attentions`)
+            `last_hidden_state` of shape :obj:`(batch_size, sequence_length, hidden_size)`, `optional`, defaults to :obj:`None`) is a sequence of hidden-states at the output of the last layer of the encoder.
+            Used in the cross-attention of the decoder.
         decoder_input_ids (:obj:`torch.LongTensor` of shape :obj:`(batch_size, target_sequence_length)`, `optional`, defaults to :obj:`None`):
             Provide for translation and summarization training. By default, the model will create this tensor by shifting the input_ids right, following the paper.
-        decoder_attention_mask (:obj:`torch.Tensor` of shape :obj:`(batch_size, 1, tgt_seq_len, tgt_seq_len)`, `optional`, defaults to :obj:`None`):
-            Default behavior: generate a tensor that ignores pad tokens and future tokens, as in the paper.
+        decoder_attention_mask (:obj:`torch.BoolTensor` of shape :obj:`(batch_size, tgt_seq_len)`, `optional`, defaults to :obj:`None`):
+            Default behavior: generate a tensor that ignores pad tokens in decoder_input_ids. Causal mask will also be used by default.
             If you want to change padding behavior, you should read :func:`~transformers.modeling_bart._prepare_decoder_inputs` and modify.
             See diagram 1 in the paper for more info on the default strategy
 """
-LARGE_NEGATIVE = -1e8
+
+
+def invert_mask(attention_mask):
+    assert attention_mask.dim() == 2
+    return attention_mask.eq(0)
 
 
 def _prepare_bart_decoder_inputs(
-    config, input_ids, decoder_input_ids=None, decoder_attn_mask=None, mask_dtype=None,
+    config, input_ids, decoder_input_ids=None, decoder_padding_mask=None, causal_mask_dtype=torch.float32
 ):
-    """Prepare masks that ignore padding tokens in the decoder and a causal lm mask for the decoder if
+    """Prepare masks that ignore padding tokens in the decoder and a causal mask for the decoder if
     none are provided. This mimics the default behavior in fairseq. To override it pass in masks.
     Note: this is not called during generation
     """
     pad_token_id = config.pad_token_id
-    need_causal_mask = not config.output_past
     if decoder_input_ids is None:
         decoder_input_ids = shift_tokens_right(input_ids, pad_token_id)
-    bsz, tgt_len = decoder_input_ids.size()[:2]
-    if decoder_attn_mask is None:
+    bsz, tgt_len = decoder_input_ids.size()
+    if decoder_padding_mask is None:
         decoder_padding_mask = make_padding_mask(decoder_input_ids, pad_token_id)
-        if need_causal_mask:
-            causal_lm_mask = torch.triu(fill_with_neg_inf(torch.zeros(tgt_len, tgt_len)), 1)
-        else:
-            causal_lm_mask = None
-        new_shape = (bsz, tgt_len, tgt_len)
-        # make it broadcastable so can just be added to the attention coefficients
-        decoder_attn_mask = _combine_masks(decoder_padding_mask, causal_lm_mask, new_shape).to(device=input_ids.device)
-        if mask_dtype is not None:
-            decoder_attn_mask = decoder_attn_mask.to(mask_dtype)
-    assert decoder_attn_mask is None or decoder_attn_mask.shape == (bsz, 1, tgt_len, tgt_len)
-    return decoder_input_ids, decoder_attn_mask
+    else:
+        decoder_padding_mask = invert_mask(decoder_padding_mask)
+    causal_mask = torch.triu(fill_with_neg_inf(torch.zeros(tgt_len, tgt_len)), 1).to(
+        dtype=causal_mask_dtype, device=decoder_input_ids.device
+    )
+    return decoder_input_ids, decoder_padding_mask, causal_mask
 
 
-class PretrainedBartModel(PreTrainedModel, LoggingMixin):
+class PretrainedBartModel(PreTrainedModel):
     config_class = BartConfig
     base_model_prefix = "model"
     pretrained_model_archive_map = BART_PRETRAINED_MODEL_ARCHIVE_MAP
+    encoder_outputs_batch_dim_idx = 1  # outputs shaped (seq_len, bs, ...)
 
     def _init_weights(self, module):
         std = self.config.init_std
@@ -129,13 +131,10 @@ class PretrainedBartModel(PreTrainedModel, LoggingMixin):
     @property
     def dummy_inputs(self):
         pad_token = self.config.pad_token_id
-        input_ids = torch.tensor([[0, 6, 10, 4, 2], [0, 8, 12, 2, pad_token]])
-        decoder_input_ids, decoder_attn_mask = _prepare_bart_decoder_inputs(self.config, input_ids,)
+        input_ids = torch.tensor([[0, 6, 10, 4, 2], [0, 8, 12, 2, pad_token]], device=self.device)
         dummy_inputs = {
-            "decoder_input_ids": decoder_input_ids,
             "attention_mask": input_ids.ne(pad_token),
             "input_ids": input_ids,
-            "decoder_attention_mask": decoder_attn_mask,
         }
         return dummy_inputs
 
@@ -153,21 +152,6 @@ def _check_shapes(shape_1, shape2):
         raise AssertionError("shape mismatch: {} != {}".format(shape_1, shape2))
 
 
-def _combine_masks(key_padding_mask, causal_lm_mask, targ_size):
-    """Make one mask of shape (bsz, 1, tgt_len, src_len) """
-    a = torch.zeros(targ_size)  # targ_size is(bsz, tgt_len, src_len)
-    b = torch.zeros(targ_size)
-    if key_padding_mask is not None:  # (bsz, tgt_len) -> targ_size
-        _check_shapes(key_padding_mask.shape, targ_size[:2])
-        reshaped = key_padding_mask.unsqueeze(2).expand(*targ_size)
-        a[reshaped] = LARGE_NEGATIVE
-
-    if causal_lm_mask is not None:  # (tgt_len, src_len) -> targ_size
-        _check_shapes(causal_lm_mask.shape, targ_size[-2:])
-        b = causal_lm_mask.unsqueeze(0).expand(*targ_size)
-    return (a + b).unsqueeze(1).clamp(LARGE_NEGATIVE,)
-
-
 def shift_tokens_right(input_ids, pad_token_id):
     """Shift input ids one token to the right, and wrap the last non pad token (usually <eos>)."""
     prev_output_tokens = input_ids.clone()
@@ -186,10 +170,9 @@ def make_padding_mask(input_ids, padding_idx=1):
 
 
 # Helper Modules
-from durbango.torch_utils import get_shapes
 
 
-class EncoderLayer(nn.Module, LoggingMixin):
+class EncoderLayer(nn.Module):
     def __init__(self, config: BartConfig):
         super().__init__()
         self.embed_dim = config.d_model
@@ -218,8 +201,9 @@ class EncoderLayer(nn.Module, LoggingMixin):
             encoded output of shape `(seq_len, batch, embed_dim)`
         """
         residual = x
-        x, attn_weights = self.self_attn(query=x, key=x, key_padding_mask=encoder_padding_mask, update_layer_state=False,)
-
+        x, attn_weights = self.self_attn(
+            query=x, key=x, key_padding_mask=encoder_padding_mask, need_weights=self.output_attentions
+        )
         x = F.dropout(x, p=self.dropout, training=self.training)
         x = residual + x
         x = self.self_attn_layer_norm(x)
@@ -233,8 +217,8 @@ class EncoderLayer(nn.Module, LoggingMixin):
         x = self.final_layer_norm(x)
         return x, attn_weights
 
-import gc
-class BartEncoder(nn.Module, LoggingMixin):
+
+class BartEncoder(nn.Module):
     """
     Transformer encoder consisting of *config.encoder_layers* self attention layers. Each layer
     is a :class:`EncoderLayer`.
@@ -281,23 +265,19 @@ class BartEncoder(nn.Module, LoggingMixin):
         """
         # check attention mask and invert
         if attention_mask is not None:
-            assert attention_mask.dim() == 2
-            attention_mask = attention_mask.eq(0)
+            attention_mask = invert_mask(attention_mask)
 
         inputs_embeds = self.embed_tokens(input_ids)
-        x = inputs_embeds + self.embed_positions(input_ids)
+        embed_pos = self.embed_positions(input_ids)
+        x = inputs_embeds + embed_pos
         x = self.layernorm_embedding(x)
         x = F.dropout(x, p=self.dropout, training=self.training)
-        assert not (self.output_attentions or self.output_hidden_states)
 
         # B x T x C -> T x B x C
         x = x.transpose(0, 1)
-        self.log_mem('encoder: starting_loop')
-        encoder_states, all_attentions = [], []
-        #rdd_start = print_tensor_sizes()
-        #rdd_start.to_csv(f'rdd_start.csv')
-        for i, encoder_layer in enumerate(self.layers):
 
+        encoder_states, all_attentions = [], []
+        for encoder_layer in self.layers:
             if self.output_hidden_states:
                 encoder_states.append(x)
             # add LayerDrop (see https://arxiv.org/abs/1909.11556 for description)
@@ -305,19 +285,23 @@ class BartEncoder(nn.Module, LoggingMixin):
             if self.training and (dropout_probability < self.layerdrop):  # skip the layer
                 attn = None
             else:
-                x, _ = encoder_layer(x, attention_mask)
-            assert len(encoder_states) == 0
-            assert len(all_attentions) == 0
-            #self.log_mem(f'x: {x.shape}, attn: {attn.shape}')
-            self.log_mem(f'Encoder: called layer {i}')
+                x, attn = encoder_layer(x, attention_mask)
+
+            if self.output_attentions:
+                all_attentions.append(attn)
+
+        if self.output_hidden_states:
+            encoder_states.append(x)
+
         encoder_states = [hidden_state.transpose(0, 1) for hidden_state in encoder_states]
         return x, encoder_states, all_attentions
 
 
-class DecoderLayer(nn.Module, LoggingMixin):
+class DecoderLayer(nn.Module):
     def __init__(self, config: BartConfig):
         super().__init__()
         self.embed_dim = config.d_model
+        self.output_attentions = config.output_attentions
         self.self_attn = SelfAttention(
             embed_dim=self.embed_dim, num_heads=config.decoder_attention_heads, dropout=config.attention_dropout,
         )
@@ -338,21 +322,34 @@ class DecoderLayer(nn.Module, LoggingMixin):
         self.final_layer_norm = LayerNorm(self.embed_dim)
 
     def forward(
-        self, x, encoder_hidden_states, encoder_attn_mask=None, layer_state=None, attention_mask=None,
+        self,
+        x,
+        encoder_hidden_states,
+        encoder_attn_mask=None,
+        layer_state=None,
+        causal_mask=None,
+        decoder_padding_mask=None,
     ):
         residual = x
 
         if layer_state is None:
             layer_state = {}
         # next line mutates layer state
-        x, self_attn_weights = self.self_attn(query=x, key=x, layer_state=layer_state, attn_mask=attention_mask,)
+        x, self_attn_weights = self.self_attn(
+            query=x,
+            key=x,
+            layer_state=layer_state,
+            key_padding_mask=decoder_padding_mask,
+            attn_mask=causal_mask,
+            need_weights=self.output_attentions,
+        )
         x = F.dropout(x, p=self.dropout, training=self.training)
         x = residual + x
         x = self.self_attn_layer_norm(x)
         residual = x
         assert self.encoder_attn.cache_key != self.self_attn.cache_key
 
-        x, encoder_attn_weights = self.encoder_attn(
+        x, _ = self.encoder_attn(
             query=x,
             key=encoder_hidden_states,
             key_padding_mask=encoder_attn_mask,
@@ -377,7 +374,7 @@ class DecoderLayer(nn.Module, LoggingMixin):
         )  # just self_attn weights for now, following t5, layer_state = cache for decoding
 
 
-class BartDecoder(nn.Module, LoggingMixin):
+class BartDecoder(nn.Module):
     """
     Transformer decoder consisting of *config.decoder_layers* layers. Each layer
     is a :class:`DecoderLayer`.
@@ -409,7 +406,8 @@ class BartDecoder(nn.Module, LoggingMixin):
         input_ids,
         encoder_hidden_states,
         encoder_padding_mask,
-        combined_mask,
+        decoder_padding_mask,
+        decoder_causal_mask,
         decoder_cached_states=None,
         generation_mode=False,
         **unused
@@ -434,12 +432,9 @@ class BartDecoder(nn.Module, LoggingMixin):
         """
         # check attention mask and invert
         if encoder_padding_mask is not None:
-            assert encoder_padding_mask.dim() == 2
-            encoder_padding_mask = encoder_padding_mask.eq(0)
+            encoder_padding_mask = invert_mask(encoder_padding_mask)
 
         # embed positions
-
-        self.log_mem('decoder: embedded positions')
         positions = self.embed_positions(input_ids, generation_mode=generation_mode)
 
         if generation_mode:
@@ -448,7 +443,6 @@ class BartDecoder(nn.Module, LoggingMixin):
             assert input_ids.ne(self.padding_idx).any()
 
         x = self.embed_tokens(input_ids)
-        self.log_mem('decoder: embedded tokens')
         x += positions
 
         x = self.layernorm_embedding(x)
@@ -458,7 +452,6 @@ class BartDecoder(nn.Module, LoggingMixin):
         all_hidden_states = ()
         all_self_attns = ()
         next_decoder_cache = []
-
         for i, decoder_layer in enumerate(self.layers):
             decoder_layer  # type: DecoderLayer
             # add LayerDrop (see https://arxiv.org/abs/1909.11556 for description)
@@ -468,9 +461,13 @@ class BartDecoder(nn.Module, LoggingMixin):
 
             layer_state = decoder_cached_states[i] if decoder_cached_states is not None else None
             x, layer_self_attn, layer_past = decoder_layer(
-                x, encoder_hidden_states, encoder_padding_mask, layer_state=layer_state, attention_mask=combined_mask,
+                x,
+                encoder_hidden_states,
+                encoder_attn_mask=encoder_padding_mask,
+                decoder_padding_mask=decoder_padding_mask,
+                layer_state=layer_state,
+                causal_mask=decoder_causal_mask,
             )
-            self.log_mem(f'decoder: called attn {i}')
 
             if self.output_past:
                 next_decoder_cache.append(layer_past.copy())
@@ -490,7 +487,6 @@ class BartDecoder(nn.Module, LoggingMixin):
         return x, next_cache, all_hidden_states, list(all_self_attns)
 
 
-
 def _reorder_buffer(attn_cache, new_order):
     for k, input_buffer_k in attn_cache.items():
         if input_buffer_k is not None:
@@ -498,8 +494,8 @@ def _reorder_buffer(attn_cache, new_order):
     return attn_cache
 
 
-class SelfAttention(nn.Module, LoggingMixin):
-    """Multi-headed attention from "Attention Is All You Need"""
+class SelfAttention(nn.Module):
+    """Multi-headed attention from 'Attention Is All You Need' paper"""
 
     def __init__(
         self,
@@ -527,18 +523,14 @@ class SelfAttention(nn.Module, LoggingMixin):
     def _shape(self, tensor, dim_0, bsz):
         return tensor.contiguous().view(dim_0, bsz * self.num_heads, self.head_dim).transpose(0, 1)
 
-
-    def log_mem(self, msg='', verbose=False):
-        super().log_mem(msg=f'{self.cache_key}_attn:{msg}', verbose=verbose)
-
     def forward(
         self,
         query,
         key: Optional[Tensor],
         key_padding_mask: Optional[Tensor] = None,
-        update_layer_state=True,
         layer_state: Optional[Dict[str, Optional[Tensor]]] = None,
         attn_mask: Optional[Tensor] = None,
+        need_weights=False,
     ) -> Tuple[Tensor, Optional[Tensor]]:
         """Input shape: Time(SeqLen) x Batch x Channel"""
         static_kv = self.encoder_decoder_attention  # type: bool
@@ -557,7 +549,6 @@ class SelfAttention(nn.Module, LoggingMixin):
             layer_state = {}
 
         q = self.q_proj(query) * self.scaling
-        self.log_mem('\tq_proj')
         if static_kv:
             if key is None:
                 k = v = None
@@ -568,39 +559,29 @@ class SelfAttention(nn.Module, LoggingMixin):
             k = self.k_proj(query)
             v = self.v_proj(query)
 
-
         q = self._shape(q, tgt_len, bsz)
-        self.log_mem(f'\tq_reshape -> {q.shape}')
         if k is not None:
             k = self._shape(k, -1, bsz)
-            self.log_mem(f'\t done reshaping k,v ->, {k.shape}')
         if v is not None:
             v = self._shape(v, -1, bsz)
 
-
         if saved_state is not None:
-            self.log_mem('\t about to use saved_state')
             k, v, key_padding_mask = self._use_saved_state(k, v, saved_state, key_padding_mask, static_kv, bsz)
 
         # Update cache
-        if update_layer_state:
-            layer_state[self.cache_key] = {
-                "prev_key": k.view(bsz, self.num_heads, -1, self.head_dim),
-                "prev_value": v.view(bsz, self.num_heads, -1, self.head_dim),
-                "prev_key_padding_mask": key_padding_mask if not static_kv else None,
-            }
-            self.log_mem('\t attn: done layer_state')
+        layer_state[self.cache_key] = {
+            "prev_key": k.view(bsz, self.num_heads, -1, self.head_dim),
+            "prev_value": v.view(bsz, self.num_heads, -1, self.head_dim),
+            "prev_key_padding_mask": key_padding_mask if not static_kv else None,
+        }
 
         assert k is not None
         src_len = k.size(1)
-        self.log_mem('\t attn: before BMM(q,k)')
         attn_weights = torch.bmm(q, k.transpose(1, 2))
-        self.log_mem('\t attn: done BMM(q,k)')
         assert attn_weights.size() == (bsz * self.num_heads, tgt_len, src_len)
 
         if attn_mask is not None:
             attn_weights = attn_weights.view(bsz, self.num_heads, tgt_len, src_len) + attn_mask
-            self.log_mem('\t attn: done causal mask')
             attn_weights = attn_weights.view(bsz * self.num_heads, tgt_len, src_len)
 
         # This is part of a workaround to get around fork/join parallelism not supporting Optional types.
@@ -608,26 +589,24 @@ class SelfAttention(nn.Module, LoggingMixin):
             key_padding_mask = None
         assert key_padding_mask is None or key_padding_mask.size()[:2] == (bsz, src_len,)
 
-        if key_padding_mask is not None:  # shape (bsz, src_len)
+        if key_padding_mask is not None:  # don't attend to padding symbols
             attn_weights = attn_weights.view(bsz, self.num_heads, tgt_len, src_len)
-            attn_weights = attn_weights.masked_fill(key_padding_mask.unsqueeze(1).unsqueeze(2), float("-inf"))
-            self.log_mem('\t attn: done masked_fill')
+            reshaped = key_padding_mask.unsqueeze(1).unsqueeze(2)
+            attn_weights = attn_weights.masked_fill(reshaped, float("-inf"))
             attn_weights = attn_weights.view(bsz * self.num_heads, tgt_len, src_len)
         attn_weights = F.softmax(attn_weights, dim=-1)
-        self.log_mem('\t attn: done softmax')
         attn_probs = F.dropout(attn_weights, p=self.dropout, training=self.training,)
 
-
         assert v is not None
         attn_output = torch.bmm(attn_probs, v)
-        self.log_mem('\t attn: done BMM(probs, v)')
         assert attn_output.size() == (bsz * self.num_heads, tgt_len, self.head_dim)
         attn_output = attn_output.transpose(0, 1).contiguous().view(tgt_len, bsz, embed_dim)
-        self.log_mem('\t attn: done view(output)')
         attn_output = self.out_proj(attn_output)
-        self.log_mem('\t attn: done out_proj')
-        #attn_weights = attn_weights.view(bsz, self.num_heads, tgt_len, src_len)
-        return attn_output, None
+        if need_weights:
+            attn_weights = attn_weights.view(bsz, self.num_heads, tgt_len, src_len)
+        else:
+            attn_weights = None
+        return attn_output, attn_weights
 
     def _use_saved_state(self, k, v, saved_state, key_padding_mask, static_kv, bsz):
         # saved states are stored with shape (bsz, num_heads, seq_len, head_dim)
@@ -684,7 +663,7 @@ class SelfAttention(nn.Module, LoggingMixin):
         return new_key_padding_mask
 
 
-class BartClassificationHead(nn.Module, LoggingMixin):
+class BartClassificationHead(nn.Module):
     """Head for sentence-level classification tasks."""
 
     # This can trivially be shared with RobertaClassificationHead
@@ -755,9 +734,9 @@ def _filter_out_falsey_values(tup) -> Tuple:
 
 
 # Public API
+def _get_shape(t):
+    return getattr(t, "shape", None)
 
-import time
-import pandas as pd
 
 @add_start_docstrings(
     "The bare BART Model outputting raw hidden-states without any specific head on top.", BART_START_DOCSTRING,
@@ -774,7 +753,6 @@ class BartModel(PretrainedBartModel):
         self.encoder = BartEncoder(config, self.shared)
         self.decoder = BartDecoder(config, self.shared)
 
-
         self.init_weights()
 
     @add_start_docstrings_to_callable(BART_INPUTS_DOCSTRING)
@@ -790,24 +768,28 @@ class BartModel(PretrainedBartModel):
     ):
 
         # make masks if user doesn't supply
-        if encoder_outputs is None:
-            encoder_outputs = self.encoder(input_ids=input_ids, attention_mask=attention_mask)
-        assert isinstance(encoder_outputs, tuple)
         if not generation_mode:
-            decoder_input_ids, decoder_attention_mask = _prepare_bart_decoder_inputs(
+            decoder_input_ids, decoder_padding_mask, causal_mask = _prepare_bart_decoder_inputs(
                 self.config,
                 input_ids,
                 decoder_input_ids=decoder_input_ids,
-                decoder_attn_mask=decoder_attention_mask,
-                mask_dtype=self.shared.weight.dtype,
+                decoder_padding_mask=decoder_attention_mask,
+                causal_mask_dtype=self.shared.weight.dtype,
             )
+        else:
+            decoder_padding_mask, causal_mask = None, None
+
         assert decoder_input_ids is not None
+        if encoder_outputs is None:
+            encoder_outputs = self.encoder(input_ids=input_ids, attention_mask=attention_mask)
+        assert isinstance(encoder_outputs, tuple)
         # decoder outputs consists of (dec_features, layer_state, dec_hidden, dec_attn)
         decoder_outputs = self.decoder(
             decoder_input_ids,
             encoder_outputs[0],
             attention_mask,
-            decoder_attention_mask,
+            decoder_padding_mask,
+            decoder_causal_mask=causal_mask,
             decoder_cached_states=decoder_cached_states,
             generation_mode=generation_mode,
         )
@@ -836,13 +818,8 @@ class BartForConditionalGeneration(PretrainedBartModel):
 
     def __init__(self, config: BartConfig):
         super().__init__(config)
-        # if base_model is Nones:
-        #self.log_mem('pre-init')
-        self.model = BartModel(config)
-        #self.lm_head = _make_linear_from_emb(self.model.shared)
-
-    def tie_weights(self):
-        pass  # hack to prevent changing lm_head.out_features. The input and output embeddings are still the same.
+        base_model = BartModel(config)
+        self.model = base_model
 
     @add_start_docstrings_to_callable(BART_INPUTS_DOCSTRING)
     def forward(
@@ -898,7 +875,6 @@ class BartForConditionalGeneration(PretrainedBartModel):
             tokenizer.decode(predictions).split()
             # ['good', 'great', 'all', 'really', 'very']
         """
-        self.model.log_mem('before BartModel.forward')
         outputs = self.model(
             input_ids,
             attention_mask=attention_mask,
@@ -908,10 +884,7 @@ class BartForConditionalGeneration(PretrainedBartModel):
             decoder_cached_states=decoder_cached_states,
             generation_mode=generation_mode,
         )
-        self.model.log_mem('after call, before lm_head')
         lm_logits = F.linear(outputs[0], self.model.shared.weight)
-        #lm_logits = self.lm_head(outputs[0])
-        self.model.log_mem('after lm_head')
         outputs = (lm_logits,) + outputs[1:]  # Add hidden states and attention if they are here
         if lm_labels is not None:
             loss_fct = nn.CrossEntropyLoss()
@@ -921,7 +894,6 @@ class BartForConditionalGeneration(PretrainedBartModel):
 
         return outputs
 
-
     def prepare_inputs_for_generation(self, decoder_input_ids, past, attention_mask, **kwargs):
         assert past is not None, "past has to be defined for encoder_outputs"
 
@@ -930,8 +902,6 @@ class BartForConditionalGeneration(PretrainedBartModel):
             encoder_outputs, decoder_cached_states = past, None
         else:
             encoder_outputs, decoder_cached_states = past
-        self.log_mem(f'encoder_outputs.shape: {encoder_outputs[0].shape}')
-        self.log_mem(f'decoder_input_ids.shape: {decoder_input_ids.shape}')
         return {
             "input_ids": None,  # encoder_outputs is defined. input_ids not needed
             "encoder_outputs": encoder_outputs,
@@ -1006,7 +976,7 @@ class BartForSequenceClassification(PretrainedBartModel):
     Returns:
         :obj:`tuple(torch.FloatTensor)` comprising various elements depending on the configuration (:class:`~transformers.BartConfig`) and inputs:
             loss (:obj:`torch.FloatTensor` of shape :obj:`(1,)`, `optional`, returned when :obj:`label` is provided):
-                Classification  loss (cross entropy)
+                Classification loss (cross entropy)
             logits (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, config.num_labels)`):
                 Classification (or regression if config.num_labels==1) scores (before SoftMax).
             hidden_states (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``config.output_hidden_states=True``):

src/transformers/modeling_t5.py

@@ -27,7 +27,7 @@ from torch import nn
 from torch.nn import CrossEntropyLoss
 
 from .configuration_t5 import T5Config
-from .file_utils import DUMMY_INPUTS, DUMMY_MASK, add_start_docstrings
+from .file_utils import DUMMY_INPUTS, DUMMY_MASK, add_start_docstrings, add_start_docstrings_to_callable
 from .modeling_utils import PreTrainedModel, prune_linear_layer
 
 
@@ -457,6 +457,7 @@ class T5PreTrainedModel(PreTrainedModel):
     pretrained_model_archive_map = T5_PRETRAINED_MODEL_ARCHIVE_MAP
     load_tf_weights = load_tf_weights_in_t5
     base_model_prefix = "transformer"
+    encoder_outputs_batch_dim_idx = 0  # outputs shaped (bs, ...)
 
     @property
     def dummy_inputs(self):
@@ -695,8 +696,8 @@ T5_START_DOCSTRING = r"""    The T5 model was proposed in
 """
 
 T5_INPUTS_DOCSTRING = r"""
-    Inputs:
-        **input_ids**: ``torch.LongTensor`` of shape ``(batch_size, sequence_length)``:
+    Args:
+        input_ids (:obj:`torch.LongTensor` of shape :obj:`(batch_size, sequence_length)`):
             Indices of input sequence tokens in the vocabulary.
             To match pre-training, T5 input sequence should be formatted with [CLS] and [SEP] tokens as follows:
 
@@ -714,11 +715,27 @@ T5_INPUTS_DOCSTRING = r"""
             Indices can be obtained using :class:`transformers.T5Tokenizer`.
             See :func:`transformers.PreTrainedTokenizer.encode` and
             :func:`transformers.PreTrainedTokenizer.convert_tokens_to_ids` for details.
-        **attention_mask**: (`optional`) ``torch.FloatTensor`` of shape ``(batch_size, sequence_length)``:
+        attention_mask (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, sequence_length)`, `optional`, defaults to :obj:`None`):
             Mask to avoid performing attention on padding token indices.
             Mask values selected in ``[0, 1]``:
             ``1`` for tokens that are NOT MASKED, ``0`` for MASKED tokens.
-        **head_mask**: (`optional`) ``torch.FloatTensor`` of shape ``(num_heads,)`` or ``(num_layers, num_heads)``:
+        encoder_outputs (tuple(:obj:`tuple(torch.FloatTensor)`, `optional`, defaults to :obj:`None`):
+            Tuple consists of (`last_hidden_state`, `optional`: `hidden_states`, `optional`: `attentions`)
+            `last_hidden_state` of shape :obj:`(batch_size, sequence_length, hidden_size)`, `optional`, defaults to :obj:`None`) is a sequence of hidden-states at the output of the last layer of the encoder.
+            Used in the cross-attention of the decoder.
+        decoder_input_ids (:obj:`torch.LongTensor` of shape :obj:`(batch_size, target_sequence_length)`, `optional`, defaults to :obj:`None`):
+            Provide for sequence to sequence training. T5 uses the pad_token_id as the starting token for decoder_input_ids generation
+        decoder_attention_mask (:obj:`torch.BoolTensor` of shape :obj:`(batch_size, tgt_seq_len)`, `optional`, defaults to :obj:`None`):
+            Default behavior: generate a tensor that ignores pad tokens in decoder_input_ids. Causal mask will also be used by default.
+        inputs_embeds (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, sequence_length, hidden_size)`, `optional`, defaults to :obj:`None`):
+            Optionally, instead of passing :obj:`input_ids` you can choose to directly pass an embedded representation.
+            This is useful if you want more control over how to convert `input_ids` indices into associated vectors
+            than the model's internal embedding lookup matrix.
+        decoder_inputs_embeds (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, target_sequence_length, hidden_size)`, `optional`, defaults to :obj:`None`):
+            Optionally, instead of passing :obj:`decoder_input_ids` you can choose to directly pass an embedded representation.
+            This is useful if you want more control over how to convert `decoder_input_ids` indices into associated vectors
+            than the model's internal embedding lookup matrix.
+        head_mask: (:obj:`torch.FloatTensor` of shape :obj:`(num_heads,)` or :obj:`(num_layers, num_heads)`, `optional`, defaults to :obj:`None`):
             Mask to nullify selected heads of the self-attention modules.
             Mask values selected in ``[0, 1]``:
             ``1`` indicates the head is **not masked**, ``0`` indicates the head is **masked**.
@@ -728,31 +745,8 @@ T5_INPUTS_DOCSTRING = r"""
 @add_start_docstrings(
     "The bare T5 Model transformer outputting raw hidden-states" "without any specific head on top.",
     T5_START_DOCSTRING,
-    T5_INPUTS_DOCSTRING,
 )
 class T5Model(T5PreTrainedModel):
-    r"""
-    Outputs: `Tuple` comprising various elements depending on the configuration (config) and inputs:
-        **last_hidden_state**: ``torch.FloatTensor`` of shape ``(batch_size, sequence_length, hidden_size)``
-            Sequence of hidden-states at the output of the last layer of the model.
-        **hidden_states**: (`optional`, returned when ``config.output_hidden_states=True``)
-            list of ``torch.FloatTensor`` (one for the output of each layer + the output of the embeddings)
-            of shape ``(batch_size, sequence_length, hidden_size)``:
-            Hidden-states of the model at the output of each layer plus the initial embedding outputs.
-        **attentions**: (`optional`, returned when ``config.output_attentions=True``)
-            list of ``torch.FloatTensor`` (one for each layer) of shape ``(batch_size, num_heads, sequence_length, sequence_length)``:
-            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention heads.
-
-    Examples::
-
-        tokenizer = T5Tokenizer.from_pretrained('t5-small')
-        model = T5Model.from_pretrained('t5-small')
-        input_ids = torch.tensor(tokenizer.encode("Hello, my dog is cute")).unsqueeze(0)  # Batch size 1
-        outputs = model(input_ids=input_ids)
-        last_hidden_states = outputs[0]  # The last hidden-state is the first element of the output tuple
-
-    """
-
     def __init__(self, config):
         super().__init__(config)
         self.shared = nn.Embedding(config.vocab_size, config.d_model)
@@ -782,6 +776,7 @@ class T5Model(T5PreTrainedModel):
         for layer, heads in heads_to_prune.items():
             self.encoder.layer[layer].attention.prune_heads(heads)
 
+    @add_start_docstrings_to_callable(T5_INPUTS_DOCSTRING)
     def forward(
         self,
         input_ids=None,
@@ -793,6 +788,32 @@ class T5Model(T5PreTrainedModel):
         decoder_inputs_embeds=None,
         head_mask=None,
     ):
+        r"""
+        Return:
+            :obj:`tuple(torch.FloatTensor)` comprising various elements depending on the configuration (:class:`~transformers.T5`) and inputs:
+            last_hidden_state (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, sequence_length, hidden_size)`):
+                Sequence of hidden-states at the output of the last layer of the model.
+            hidden_states (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``config.output_hidden_states=True``):
+                Tuple of :obj:`torch.FloatTensor` (one for the output of the embeddings + one for the output of each layer)
+                of shape :obj:`(batch_size, sequence_length, hidden_size)`.
+
+                Hidden-states of the model at the output of each layer plus the initial embedding outputs.
+            attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``config.output_attentions=True``):
+                Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
+                :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
+
+                Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+                heads.
+
+        Examples::
+
+            tokenizer = T5Tokenizer.from_pretrained('t5-small')
+            model = T5Model.from_pretrained('t5-small')
+            input_ids = tokenizer.encode("Hello, my dog is cute", return_tensor="pt")  # Batch size 1
+            outputs = model(input_ids=input_ids)
+            last_hidden_states = outputs[0]  # The last hidden-state is the first element of the output tuple
+
+        """
 
         # Encode if needed (training, first prediction pass)
         if encoder_outputs is None:
@@ -814,40 +835,9 @@ class T5Model(T5PreTrainedModel):
 
         return decoder_outputs + encoder_outputs
 
-from durbango.logging_utils import LoggingMixin
-
-@add_start_docstrings("""T5 Model with a `language modeling` head on top. """, T5_START_DOCSTRING, T5_INPUTS_DOCSTRING)
-class T5ForConditionalGeneration(T5PreTrainedModel, LoggingMixin):
-    r"""
-        **lm_labels**: (`optional`) ``torch.LongTensor`` of shape ``(batch_size, sequence_length)``:
-            Labels for computing the masked language modeling loss.
-            Indices should either be in ``[0, ..., config.vocab_size]`` or -100 (see ``input_ids`` docstring).
-            Tokens with indices set to ``-100`` are ignored (masked), the loss is only computed for the tokens with labels
-            in ``[0, ..., config.vocab_size]``.
-
-    Outputs: `Tuple` comprising various elements depending on the configuration (config) and inputs:
-        **loss**: (`optional`, returned when ``lm_labels`` is provided) ``torch.FloatTensor`` of shape ``(1,)``:
-            Masked language modeling loss.
-        **prediction_scores**: ``torch.FloatTensor`` of shape ``(batch_size, sequence_length, config.vocab_size)``
-            Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax).
-        **hidden_states**: (`optional`, returned when ``config.output_hidden_states=True``)
-            list of ``torch.FloatTensor`` (one for the output of each layer + the output of the embeddings)
-            of shape ``(batch_size, sequence_length, hidden_size)``:
-            Hidden-states of the model at the output of each layer plus the initial embedding outputs.
-        **attentions**: (`optional`, returned when ``config.output_attentions=True``)
-            list of ``torch.FloatTensor`` (one for each layer) of shape ``(batch_size, num_heads, sequence_length, sequence_length)``:
-            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention heads.
-
-    Examples::
-
-        tokenizer = T5Tokenizer.from_pretrained('t5-small')
-        model = T5ForConditionalGeneration.from_pretrained('t5-small')
-        input_ids = torch.tensor(tokenizer.encode("Hello, my dog is cute")).unsqueeze(0)  # Batch size 1
-        outputs = model(input_ids=input_ids, lm_labels=input_ids)
-        loss, prediction_scores = outputs[:2]
-
-    """
 
+@add_start_docstrings("""T5 Model with a `language modeling` head on top. """, T5_START_DOCSTRING)
+class T5ForConditionalGeneration(T5PreTrainedModel):
     def __init__(self, config):
         super().__init__(config)
         self.model_dim = config.d_model
@@ -879,6 +869,7 @@ class T5ForConditionalGeneration(T5PreTrainedModel, LoggingMixin):
     def get_encoder(self):
         return self.encoder
 
+    @add_start_docstrings_to_callable(T5_INPUTS_DOCSTRING)
     def forward(
         self,
         input_ids=None,
@@ -891,6 +882,37 @@ class T5ForConditionalGeneration(T5PreTrainedModel, LoggingMixin):
         decoder_inputs_embeds=None,
         head_mask=None,
     ):
+        r"""
+        lm_labels (:obj:`torch.LongTensor` of shape :obj:`(batch_size,)`, `optional`, defaults to :obj:`None`):
+            Labels for computing the sequence classification/regression loss.
+            Indices should be in :obj:`[0, ..., config.vocab_size - 1]`.
+            If :obj:`config.num_labels > 1` a classification loss is computed (Cross-Entropy).
+
+        Returns:
+            :obj:`tuple(torch.FloatTensor)` comprising various elements depending on the configuration (:class:`~transformers.T5Config`) and inputs:
+            loss (:obj:`torch.FloatTensor` of shape :obj:`(1,)`, `optional`, returned when :obj:`lm_label` is provided):
+                Classification loss (cross entropy)
+            prediction_scores (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, sequence_length, config.vocab_size)`)
+                Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax).
+            hidden_states (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``config.output_hidden_states=True``):
+                Tuple of :obj:`torch.FloatTensor` (one for the output of the embeddings + one for the output of each layer)
+                of shape :obj:`(batch_size, sequence_length, hidden_size)`.
+
+                Hidden-states of the model at the output of each layer plus the initial embedding outputs.
+            attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``config.output_attentions=True``):
+                Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
+                :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
+
+                Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+
+        Examples::
+
+            tokenizer = T5Tokenizer.from_pretrained('t5-small')
+            model = T5ForConditionalGeneration.from_pretrained('t5-small')
+            input_ids = tokenizer.encode("Hello, my dog is cute", return_tensor="pt")  # Batch size 1
+            outputs = model(input_ids=input_ids, lm_labels=input_ids)
+            loss, prediction_scores = outputs[:2]
+        """
 
         # Encode if needed (training, first prediction pass)
         if encoder_outputs is None:

src/transformers/modeling_tf_t5.py

@@ -24,7 +24,7 @@ import math
 import tensorflow as tf
 
 from .configuration_t5 import T5Config
-from .file_utils import DUMMY_INPUTS, DUMMY_MASK, add_start_docstrings
+from .file_utils import DUMMY_INPUTS, DUMMY_MASK, add_start_docstrings, add_start_docstrings_to_callable
 from .modeling_tf_utils import TFPreTrainedModel, TFSharedEmbeddings, shape_list
 
 
@@ -630,68 +630,60 @@ T5_START_DOCSTRING = r"""    The T5 model was proposed in
 """
 
 T5_INPUTS_DOCSTRING = r"""
-    Inputs:
-        **input_ids**: ``Numpy array`` or ``tf.Tensor`` of shape ``(batch_size, sequence_length)``:
-            Indices of input sequence tokens in the vocabulary.
-            To match pre-training, T5 input sequence should be formatted with [CLS] and [SEP] tokens as follows:
+    Args:
+        decoder_input_ids (:obj: `list`,`tuple`, or `dict`):
+                decoder_input_ids are usually used as a `dict` (see T5 description above for more information) containing all the following:
+            decoder_input_ids (:obj:`tf.Tensor` of shape :obj:`(batch_size, target_sequence_length)`, `optional`, defaults to :obj:`None`):
+                Provide for sequence to sequence training. T5 uses the pad_token_id as the starting token for decoder_input_ids generation
 
-            (a) For sequence pairs:
+            input_ids (:obj:`tf.Tensor` of shape :obj:`(batch_size, sequence_length)`):
+                Indices of input sequence tokens in the vocabulary.
+                To match pre-training, T5 input sequence should be formatted with [CLS] and [SEP] tokens as follows:
 
-                ``tokens:         [CLS] is this jack ##son ##ville ? [SEP] no it is not . [SEP]``
+                (a) For sequence pairs:
 
-            (b) For single sequences:
+                    ``tokens:         [CLS] is this jack ##son ##ville ? [SEP] no it is not . [SEP]``
 
-                ``tokens:         [CLS] the dog is hairy . [SEP]``
+                (b) For single sequences:
 
+                    ``tokens:         [CLS] the dog is hairy . [SEP]``
 
-            T5 is a model with relative position embeddings so you should be able to pad the inputs on
-            the right or the left.
+                T5 is a model with relative position embeddings so you should be able to pad the inputs on
+                the right or the left.
 
-            Indices can be obtained using :class:`transformers.T5Tokenizer`.
-            See :func:`transformers.PreTrainedTokenizer.encode` and
-            :func:`transformers.PreTrainedTokenizer.convert_tokens_to_ids` for details.
-        **attention_mask**: (`optional`) ``Numpy array`` or ``tf.Tensor`` of shape ``(batch_size, sequence_length)``:
-            Mask to avoid performing attention on padding token indices.
-            Mask values selected in ``[0, 1]``:
-            ``1`` for tokens that are NOT MASKED, ``0`` for MASKED tokens.
-        **head_mask**: (`optional`) ``Numpy array`` or ``tf.Tensor`` of shape ``(num_heads,)`` or ``(num_layers, num_heads)``:
-            Mask to nullify selected heads of the self-attention modules.
-            Mask values selected in ``[0, 1]``:
-            ``1`` indicates the head is **not masked**, ``0`` indicates the head is **masked**.
+                Indices can be obtained using :class:`transformers.T5Tokenizer`.
+                See :func:`transformers.PreTrainedTokenizer.encode` and
+                :func:`transformers.PreTrainedTokenizer.convert_tokens_to_ids` for details.
+            attention_mask (:obj:`tf.Tensor` of shape :obj:`(batch_size, sequence_length)`, `optional`, defaults to :obj:`None`):
+                Mask to avoid performing attention on padding token indices.
+                Mask values selected in ``[0, 1]``:
+                ``1`` for tokens that are NOT MASKED, ``0`` for MASKED tokens.
+            encoder_outputs (tuple(:obj:`tuple(tf.FloatTensor)`, `optional`, defaults to :obj:`None`):
+                Tuple consists of (`last_hidden_state`, `optional`: `hidden_states`, `optional`: `attentions`)
+                `last_hidden_state` of shape :obj:`(batch_size, sequence_length, hidden_size)`, `optional`, defaults to :obj:`None`) is a sequence of hidden-states at the output of the last layer of the encoder.
+                Used in the cross-attention of the decoder.
+            decoder_attention_mask (:obj:`tf.Tensor` of shape :obj:`(batch_size, tgt_seq_len)`, `optional`, defaults to :obj:`None`):
+                Default behavior: generate a tensor that ignores pad tokens in decoder_input_ids. Causal mask will also be used by default.
+            inputs_embeds (:obj:`tf.Tensor` of shape :obj:`(batch_size, sequence_length, hidden_size)`, `optional`, defaults to :obj:`None`):
+                Optionally, instead of passing :obj:`input_ids` you can choose to directly pass an embedded representation.
+                This is useful if you want more control over how to convert `input_ids` indices into associated vectors
+                than the model's internal embedding lookup matrix.
+            decoder_inputs_embeds (:obj:`tf.Tensor` of shape :obj:`(batch_size, target_sequence_length, hidden_size)`, `optional`, defaults to :obj:`None`):
+                Optionally, instead of passing :obj:`decoder_input_ids` you can choose to directly pass an embedded representation.
+                This is useful if you want more control over how to convert `decoder_input_ids` indices into associated vectors
+                than the model's internal embedding lookup matrix.
+            head_mask: (:obj:`tf.Tensor` of shape :obj:`(num_heads,)` or :obj:`(num_layers, num_heads)`, `optional`, defaults to :obj:`None`):
+                Mask to nullify selected heads of the self-attention modules.
+                Mask values selected in ``[0, 1]``:
+                ``1`` indicates the head is **not masked**, ``0`` indicates the head is **masked**.
 """
 
 
 @add_start_docstrings(
     "The bare T5 Model transformer outputting raw hidden-states" "without any specific head on top.",
     T5_START_DOCSTRING,
-    T5_INPUTS_DOCSTRING,
 )
 class TFT5Model(TFT5PreTrainedModel):
-    r"""
-    Outputs: `Tuple` comprising various elements depending on the configuration (config) and inputs:
-        **last_hidden_state**: ``tf.Tensor`` of shape ``(batch_size, sequence_length, hidden_size)``
-            Sequence of hidden-states at the output of the last layer of the model.
-        **hidden_states**: (`optional`, returned when ``config.output_hidden_states=True``)
-            list of ``tf.Tensor`` (one for the output of each layer + the output of the embeddings)
-            of shape ``(batch_size, sequence_length, hidden_size)``:
-            Hidden-states of the model at the output of each layer plus the initial embedding outputs.
-        **attentions**: (`optional`, returned when ``config.output_attentions=True``)
-            list of ``tf.Tensor`` (one for each layer) of shape ``(batch_size, num_heads, sequence_length, sequence_length)``:
-            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention heads.
-
-    Examples::
-
-        import tensorflow as tf
-        from transformers import T5Tokenizer, TFT5Model
-
-        tokenizer = T5Tokenizer.from_pretrained('t5-small')
-        model = TFT5Model.from_pretrained('t5-small')
-        input_ids = tf.constant(tokenizer.encode("Hello, my dog is cute"))[None, :]  # Batch size 1
-        outputs = model(input_ids=input_ids)
-        last_hidden_states = outputs[0]  # The last hidden-state is the first element of the output tuple
-
-    """
-
     def __init__(self, config, *inputs, **kwargs):
         super().__init__(config, *inputs, **kwargs)
         self.shared = TFSharedEmbeddings(config.vocab_size, config.d_model, name="shared")
@@ -715,7 +707,34 @@ class TFT5Model(TFT5PreTrainedModel):
     def get_output_embeddings(self):
         return self.shared
 
+    @add_start_docstrings_to_callable(T5_INPUTS_DOCSTRING)
     def call(self, decoder_input_ids, **kwargs):
+        r"""
+        Return:
+            :obj:`tuple(tf.Tensor)` comprising various elements depending on the configuration (:class:`~transformers.T5Config`) and inputs:
+            last_hidden_state (:obj:`tf.Tensor` of shape :obj:`(batch_size, sequence_length, hidden_size)`):
+                Sequence of hidden-states at the output of the last layer of the model.
+            hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``config.output_hidden_states=True``):
+                Tuple of :obj:`tf.Tensor` (one for the output of the embeddings + one for the output of each layer)
+                of shape :obj:`(batch_size, sequence_length, hidden_size)`.
+
+                Hidden-states of the model at the output of each layer plus the initial embedding outputs.
+            attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``config.output_attentions=True``):
+                Tuple of :obj:`tf.Tensor` (one for each layer) of shape
+                :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
+
+                Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+                heads.
+
+        Examples::
+
+            tokenizer = T5Tokenizer.from_pretrained('t5-small')
+            model = TFT5Model.from_pretrained('t5-small')
+            input_ids = tokenizer.encode("Hello, my dog is cute", return_tensor="tf")  # Batch size 1
+            outputs = model(input_ids)
+            last_hidden_states = outputs[0]  # The last hidden-state is the first element of the output tuple
+
+        """
 
         if isinstance(decoder_input_ids, dict):
             kwargs.update(decoder_input_ids)
@@ -753,33 +772,8 @@ class TFT5Model(TFT5PreTrainedModel):
         return decoder_outputs + encoder_outputs
 
 
-@add_start_docstrings("""T5 Model with a `language modeling` head on top. """, T5_START_DOCSTRING, T5_INPUTS_DOCSTRING)
+@add_start_docstrings("""T5 Model with a `language modeling` head on top. """, T5_START_DOCSTRING)
 class TFT5ForConditionalGeneration(TFT5PreTrainedModel):
-    r"""
-    Outputs: `Tuple` comprising various elements depending on the configuration (config) and inputs:
-        **prediction_scores**: ``Numpy array`` or ``tf.Tensor`` of shape ``(batch_size, sequence_length, config.vocab_size)``
-            Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax).
-        **hidden_states**: (`optional`, returned when ``config.output_hidden_states=True``)
-            list of ``Numpy array`` or ``tf.Tensor`` (one for the output of each layer + the output of the embeddings)
-            of shape ``(batch_size, sequence_length, hidden_size)``:
-            Hidden-states of the model at the output of each layer plus the initial embedding outputs.
-        **attentions**: (`optional`, returned when ``config.output_attentions=True``)
-            list of ``Numpy array`` or ``tf.Tensor`` (one for each layer) of shape ``(batch_size, num_heads, sequence_length, sequence_length)``:
-            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention heads.
-
-    Examples::
-
-        import tensorflow as tf
-        from transformers import T5Tokenizer, TFT5ForConditionalGeneration
-
-        tokenizer = T5Tokenizer.from_pretrained('t5-small')
-        model = TFT5ForConditionalGeneration.from_pretrained('t5-small')
-        input_ids = tf.constant(tokenizer.encode("Hello, my dog is cute"))[None, :]  # Batch size 1
-        outputs = model(input_ids=input_ids)
-        prediction_scores = outputs[0]
-
-    """
-
     def __init__(self, config, *inputs, **kwargs):
         super().__init__(config, *inputs, **kwargs)
         self.model_dim = config.d_model
@@ -808,7 +802,39 @@ class TFT5ForConditionalGeneration(TFT5PreTrainedModel):
     def get_encoder(self):
         return self.encoder
 
+    @add_start_docstrings_to_callable(T5_INPUTS_DOCSTRING)
     def call(self, decoder_input_ids, **kwargs):
+        r"""
+        lm_labels (:obj:`tf.Tensor` of shape :obj:`(batch_size,)`, `optional`, defaults to :obj:`None`):
+            Labels for computing the sequence classification/regression loss.
+            Indices should be in :obj:`[0, ..., config.vocab_size - 1]`.
+            If :obj:`config.num_labels > 1` a classification loss is computed (Cross-Entropy).
+
+        Returns:
+            :obj:`tuple(tf.Tensor)` comprising various elements depending on the configuration (:class:`~transformers.T5Config`) and inputs:
+            loss (:obj:`tf.Tensor` of shape :obj:`(1,)`, `optional`, returned when :obj:`lm_label` is provided):
+                Classification loss (cross entropy)
+            prediction_scores (:obj:`tf.Tensor` of shape :obj:`(batch_size, sequence_length, config.vocab_size)`)
+                Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax).
+            hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``config.output_hidden_states=True``):
+                Tuple of :obj:`tf.Tensor` (one for the output of the embeddings + one for the output of each layer)
+                of shape :obj:`(batch_size, sequence_length, hidden_size)`.
+
+                Hidden-states of the model at the output of each layer plus the initial embedding outputs.
+            attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``config.output_attentions=True``):
+                Tuple of :obj:`tf.Tensor` (one for each layer) of shape
+                :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
+
+                Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+
+        Examples::
+
+            tokenizer = T5Tokenizer.from_pretrained('t5-small')
+            model = T5ForConditionalGeneration.from_pretrained('t5-small')
+            input_ids = tokenizer.encode("Hello, my dog is cute", return_tensor="pt")  # Batch size 1
+            outputs = model(input_ids, lm_labels=input_ids)
+            loss, prediction_scores = outputs[:2]
+        """
 
         if isinstance(decoder_input_ids, dict):
             kwargs.update(decoder_input_ids)
@@ -844,6 +870,7 @@ class TFT5ForConditionalGeneration(TFT5PreTrainedModel):
             head_mask=head_mask,
         )
 
+        # TODO (thom / patrick): add lm_labels for loss function
         sequence_output = decoder_outputs[0] * (self.model_dim ** -0.5)
         embed_tokens = self.get_output_embeddings()
         lm_logits = embed_tokens(sequence_output, mode="linear")

src/transformers/modeling_tf_utils.py

@@ -610,7 +610,9 @@ class TFPreTrainedModel(tf.keras.Model, TFModelUtilsMixin):
         num_return_sequences = (
             num_return_sequences if num_return_sequences is not None else self.config.num_return_sequences
         )
-        decoder_start_token_id = decoder_start_token_id if decoder_start_token_id is not None else bos_token_id
+        decoder_start_token_id = (
+            decoder_start_token_id if decoder_start_token_id is not None else self.config.decoder_start_token_id
+        )
 
         if input_ids is not None:
             batch_size = shape_list(input_ids)[0]  # overriden by the input batch_size
@@ -635,9 +637,6 @@ class TFPreTrainedModel(tf.keras.Model, TFModelUtilsMixin):
         assert (eos_token_id is None) or (
             isinstance(eos_token_id, int) and (eos_token_id >= 0)
         ), "`eos_token_id` should be a positive integer."
-        assert (
-            decoder_start_token_id is not None or self.config.is_encoder_decoder is False
-        ), "`decoder_start_token_id` has to be defined if model is encoder-decoder model"
         assert length_penalty > 0, "`length_penalty` should be strictely positive."
         assert (
             isinstance(num_return_sequences, int) and num_return_sequences > 0
@@ -708,8 +707,12 @@ class TFPreTrainedModel(tf.keras.Model, TFModelUtilsMixin):
             )  # shape: (batch_size * num_return_sequences * num_beams, cur_len)
 
         if self.config.is_encoder_decoder:
+            if decoder_start_token_id is None:
+                decoder_start_token_id = bos_token_id
 
-            assert bos_token_id is not None, "Encoder Decoder Models need to have a bos_token_id"
+            assert (
+                decoder_start_token_id is not None
+            ), "decoder_start_token_id or bos_token_id has to be defined for encoder-decoder generation"
             assert hasattr(self, "get_encoder"), "{} should have a 'get_encoder' function defined".format(self)
             assert callable(self.get_encoder), "{} should be a method".format(self.get_encoder)
 
@@ -996,10 +999,12 @@ class TFPreTrainedModel(tf.keras.Model, TFModelUtilsMixin):
             # set eos token prob to zero if min_length is not reached
             if eos_token_id is not None and cur_len < min_length:
                 # create eos_token_id boolean mask
+                num_batch_hypotheses = batch_size * num_beams
+
                 is_token_logit_eos_token = tf.convert_to_tensor(
                     [True if token is eos_token_id else False for token in range(vocab_size)], dtype=tf.bool
                 )
-                eos_token_indices_mask = tf.broadcast_to(is_token_logit_eos_token, [batch_size, vocab_size])
+                eos_token_indices_mask = tf.broadcast_to(is_token_logit_eos_token, [num_batch_hypotheses, vocab_size])
 
                 scores = set_tensor_by_indices_to_value(scores, eos_token_indices_mask, -float("inf"))
 

src/transformers/modeling_utils.py

@@ -108,6 +108,10 @@ class ModuleUtilsMixin:
             module.mem_rss_post_forward = 0
             module.mem_rss_pre_forward = 0
 
+    @property
+    def device(self):
+        return next(self.parameters()).device
+
 
 class PreTrainedModel(nn.Module, ModuleUtilsMixin):
     r""" Base class for all models.
@@ -809,7 +813,9 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin):
         num_return_sequences = (
             num_return_sequences if num_return_sequences is not None else self.config.num_return_sequences
         )
-        decoder_start_token_id = decoder_start_token_id if decoder_start_token_id is not None else bos_token_id
+        decoder_start_token_id = (
+            decoder_start_token_id if decoder_start_token_id is not None else self.config.decoder_start_token_id
+        )
 
         if input_ids is not None:
             batch_size = input_ids.shape[0]  # overriden by the input batch_size
@@ -831,9 +837,6 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin):
         assert pad_token_id is None or (
             isinstance(pad_token_id, int) and (pad_token_id >= 0)
         ), "`pad_token_id` should be a positive integer."
-        assert (
-            decoder_start_token_id is not None or self.config.is_encoder_decoder is False
-        ), "`decoder_start_token_id` has to be defined if model is encoder-decoder model"
         assert (eos_token_id is None) or (
             isinstance(eos_token_id, int) and (eos_token_id >= 0)
         ), "`eos_token_id` should be a positive integer."
@@ -897,7 +900,12 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin):
             effective_batch_mult = 1
 
         if self.config.is_encoder_decoder:
-            assert bos_token_id is not None, "Encoder Decoder Models need to have a bos_token_id"
+            if decoder_start_token_id is None:
+                decoder_start_token_id = bos_token_id
+
+            assert (
+                decoder_start_token_id is not None
+            ), "decoder_start_token_id or bos_token_id has to be defined for encoder-decoder generation"
             assert hasattr(self, "get_encoder"), "{} should have a 'get_encoder' function defined".format(self)
             assert callable(self.get_encoder), "{} should be a method".format(self.get_encoder)
 
@@ -905,7 +913,6 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin):
             encoder = self.get_encoder()
 
             encoder_outputs = encoder(input_ids, attention_mask=attention_mask)
-            assert encoder_outputs[0].shape[1] == input_ids.shape[0]
 
         # Expand input ids if num_beams > 1 or num_return_sequences > 1
         if num_return_sequences > 1 or num_beams > 1:
@@ -922,7 +929,6 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin):
                 effective_batch_size * num_beams, input_ids_len
             )  # shape: (batch_size * num_return_sequences * num_beams, cur_len)
 
-
         if self.config.is_encoder_decoder:
             # create empty decoder_input_ids
             input_ids = torch.full(
@@ -932,16 +938,22 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin):
                 device=next(self.parameters()).device,
             )
             cur_len = 1
-            self.log_mem('about to rearrange')
-            assert batch_size == encoder_outputs[0].shape[1], "NEED MSG"
-            expanded_index = torch.arange(batch_size).view(-1, 1).repeat(1, num_beams* effective_batch_mult).view(-1).to(input_ids.device)
-            encoder_outputs = (encoder_outputs[0].index_select(1, expanded_index), *encoder_outputs[1:])
-            self.log_mem('done rearrange')
+            batch_idx = self.encoder_outputs_batch_dim_idx
+            assert (
+                batch_size == encoder_outputs[0].shape[batch_idx]
+            ), f"expected encoder_outputs[0] to have 1st dimension bs={batch_size}, got {encoder_outputs[0].shape[1]} "
+            expanded_idx = (
+                torch.arange(batch_size)
+                .view(-1, 1)
+                .repeat(1, num_beams * effective_batch_mult)
+                .view(-1)
+                .to(input_ids.device)
+            )
+            encoder_outputs = (encoder_outputs[0].index_select(batch_idx, expanded_idx), *encoder_outputs[1:])
         else:
             encoder_outputs = None
             cur_len = input_ids.shape[-1]
 
-
         if num_beams > 1:
             output = self._generate_beam_search(
                 input_ids,

src/transformers/modeling_xlm.py

@@ -1040,3 +1040,98 @@ class XLMForQuestionAnswering(XLMPreTrainedModel):
         outputs = outputs + transformer_outputs[1:]  # Keep new_mems and attention/hidden states if they are here
 
         return outputs
+
+
+@add_start_docstrings(
+    """XLM Model with a token classification head on top (a linear layer on top of
+    the hidden-states output) e.g. for Named-Entity-Recognition (NER) tasks. """,
+    XLM_START_DOCSTRING,
+)
+class XLMForTokenClassification(XLMPreTrainedModel):
+    def __init__(self, config):
+        super().__init__(config)
+        self.num_labels = config.num_labels
+
+        self.transformer = XLMModel(config)
+        self.dropout = nn.Dropout(config.dropout)
+        self.classifier = nn.Linear(config.hidden_size, config.num_labels)
+
+        self.init_weights()
+
+    @add_start_docstrings_to_callable(XLM_INPUTS_DOCSTRING)
+    def forward(
+        self,
+        input_ids=None,
+        attention_mask=None,
+        langs=None,
+        token_type_ids=None,
+        position_ids=None,
+        head_mask=None,
+        labels=None,
+    ):
+        r"""
+        labels (:obj:`torch.LongTensor` of shape :obj:`(batch_size, sequence_length)`, `optional`, defaults to :obj:`None`):
+            Labels for computing the token classification loss.
+            Indices should be in ``[0, ..., config.num_labels - 1]``.
+
+    Returns:
+        :obj:`tuple(torch.FloatTensor)` comprising various elements depending on the configuration (:class:`~transformers.XLMConfig`) and inputs:
+        loss (:obj:`torch.FloatTensor` of shape :obj:`(1,)`, `optional`, returned when ``labels`` is provided) :
+            Classification loss.
+        scores (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, sequence_length, config.num_labels)`)
+            Classification scores (before SoftMax).
+        hidden_states (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``config.output_hidden_states=True``):
+            Tuple of :obj:`torch.FloatTensor` (one for the output of the embeddings + one for the output of each layer)
+            of shape :obj:`(batch_size, sequence_length, hidden_size)`.
+
+            Hidden-states of the model at the output of each layer plus the initial embedding outputs.
+        attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``config.output_attentions=True``):
+            Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
+            :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
+
+            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+            heads.
+
+    Examples::
+
+        from transformers import XLMTokenizer, XLMForTokenClassification
+        import torch
+
+        tokenizer = XLMTokenizer.from_pretrained('xlm-mlm-100-1280')
+        model = XLMForTokenClassification.from_pretrained('xlm-mlm-100-1280')
+        input_ids = torch.tensor(tokenizer.encode("Hello, my dog is cute")).unsqueeze(0)  # Batch size 1
+        labels = torch.tensor([1] * input_ids.size(1)).unsqueeze(0)  # Batch size 1
+        outputs = model(input_ids, labels=labels)
+        loss, scores = outputs[:2]
+
+        """
+        outputs = self.transformer(
+            input_ids,
+            attention_mask=attention_mask,
+            langs=langs,
+            token_type_ids=token_type_ids,
+            position_ids=position_ids,
+            head_mask=head_mask,
+        )
+
+        sequence_output = outputs[0]
+
+        sequence_output = self.dropout(sequence_output)
+        logits = self.classifier(sequence_output)
+
+        outputs = (logits,) + outputs[2:]  # add hidden states and attention if they are here
+        if labels is not None:
+            loss_fct = CrossEntropyLoss()
+            # Only keep active parts of the loss
+            if attention_mask is not None:
+                active_loss = attention_mask.view(-1) == 1
+                active_logits = logits.view(-1, self.num_labels)
+                active_labels = torch.where(
+                    active_loss, labels.view(-1), torch.tensor(loss_fct.ignore_index).type_as(labels)
+                )
+                loss = loss_fct(active_logits, active_labels)
+            else:
+                loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1))
+            outputs = (loss,) + outputs
+
+        return outputs  # (loss), scores, (hidden_states), (attentions)

src/transformers/pipelines.py

@@ -31,6 +31,7 @@ from .configuration_auto import ALL_PRETRAINED_CONFIG_ARCHIVE_MAP, AutoConfig
 from .configuration_bart import BartConfig
 from .configuration_distilbert import DistilBertConfig
 from .configuration_roberta import RobertaConfig
+from .configuration_t5 import T5Config
 from .configuration_utils import PretrainedConfig
 from .configuration_xlm import XLMConfig
 from .data import SquadExample, squad_convert_examples_to_features
@@ -60,7 +61,6 @@ if is_torch_available():
         AutoModelForTokenClassification,
         AutoModelWithLMHead,
     )
-    from .modeling_bart import BartForConditionalGeneration
 
 
 logger = logging.getLogger(__name__)
@@ -130,7 +130,9 @@ class PipelineDataFormat:
 
     SUPPORTED_FORMATS = ["json", "csv", "pipe"]
 
-    def __init__(self, output_path: Optional[str], input_path: Optional[str], column: Optional[str], overwrite=False):
+    def __init__(
+        self, output_path: Optional[str], input_path: Optional[str], column: Optional[str], overwrite=False,
+    ):
         self.output_path = output_path
         self.input_path = input_path
         self.column = column.split(",") if column is not None else [""]
@@ -176,7 +178,7 @@ class PipelineDataFormat:
 
     @staticmethod
     def from_str(
-        format: str, output_path: Optional[str], input_path: Optional[str], column: Optional[str], overwrite=False
+        format: str, output_path: Optional[str], input_path: Optional[str], column: Optional[str], overwrite=False,
     ):
         if format == "json":
             return JsonPipelineDataFormat(output_path, input_path, column, overwrite=overwrite)
@@ -189,7 +191,9 @@ class PipelineDataFormat:
 
 
 class CsvPipelineDataFormat(PipelineDataFormat):
-    def __init__(self, output_path: Optional[str], input_path: Optional[str], column: Optional[str], overwrite=False):
+    def __init__(
+        self, output_path: Optional[str], input_path: Optional[str], column: Optional[str], overwrite=False,
+    ):
         super().__init__(output_path, input_path, column, overwrite=overwrite)
 
     def __iter__(self):
@@ -210,7 +214,9 @@ class CsvPipelineDataFormat(PipelineDataFormat):
 
 
 class JsonPipelineDataFormat(PipelineDataFormat):
-    def __init__(self, output_path: Optional[str], input_path: Optional[str], column: Optional[str], overwrite=False):
+    def __init__(
+        self, output_path: Optional[str], input_path: Optional[str], column: Optional[str], overwrite=False,
+    ):
         super().__init__(output_path, input_path, column, overwrite=overwrite)
 
         with open(input_path, "r") as f:
@@ -336,6 +342,7 @@ class Pipeline(_ScikitCompat):
         tokenizer: PreTrainedTokenizer,
         modelcard: Optional[ModelCard] = None,
         framework: Optional[str] = None,
+        task: str = "",
         args_parser: ArgumentHandler = None,
         device: int = -1,
         binary_output: bool = False,
@@ -356,6 +363,11 @@ class Pipeline(_ScikitCompat):
         if self.framework == "pt" and self.device.type == "cuda":
             self.model = self.model.to(self.device)
 
+        # Update config with task specific parameters
+        task_specific_params = self.model.config.task_specific_params
+        if task_specific_params is not None and task in task_specific_params:
+            self.model.config.update(task_specific_params.get(task))
+
     def save_pretrained(self, save_directory):
         """
         Save the pipeline's model and tokenizer to the specified save_directory
@@ -420,7 +432,7 @@ class Pipeline(_ScikitCompat):
         """
         args = ["input_ids", "attention_mask"]
 
-        if not isinstance(self.model.config, (DistilBertConfig, XLMConfig, RobertaConfig, BartConfig)):
+        if not isinstance(self.model.config, (DistilBertConfig, XLMConfig, RobertaConfig, BartConfig, T5Config)):
             args += ["token_type_ids"]
 
         # PR #1548 (CLI) There is an issue with attention_mask
@@ -432,14 +444,18 @@ class Pipeline(_ScikitCompat):
         else:
             return {k: [feature[k] for feature in features] for k in args}
 
-    def _parse_and_tokenize(self, *texts, **kwargs):
+    def _parse_and_tokenize(self, *texts, pad_to_max_length=False, **kwargs):
         """
         Parse arguments and tokenize
         """
         # Parse arguments
         inputs = self._args_parser(*texts, **kwargs)
         inputs = self.tokenizer.batch_encode_plus(
-            inputs, add_special_tokens=True, return_tensors=self.framework, max_length=self.tokenizer.max_len
+            inputs,
+            add_special_tokens=True,
+            return_tensors=self.framework,
+            max_length=self.tokenizer.max_len,
+            pad_to_max_length=pad_to_max_length,
         )
 
         # Filter out features not available on specific models
@@ -520,6 +536,7 @@ class FeatureExtractionPipeline(Pipeline):
         framework: Optional[str] = None,
         args_parser: ArgumentHandler = None,
         device: int = -1,
+        task: str = "",
     ):
         super().__init__(
             model=model,
@@ -529,6 +546,7 @@ class FeatureExtractionPipeline(Pipeline):
             args_parser=args_parser,
             device=device,
             binary_output=True,
+            task=task,
         )
 
     def __call__(self, *args, **kwargs):
@@ -625,6 +643,7 @@ class FillMaskPipeline(Pipeline):
         args_parser: ArgumentHandler = None,
         device: int = -1,
         topk=5,
+        task: str = "",
     ):
         super().__init__(
             model=model,
@@ -634,6 +653,7 @@ class FillMaskPipeline(Pipeline):
             args_parser=args_parser,
             device=device,
             binary_output=True,
+            task=task,
         )
 
         self.topk = topk
@@ -725,6 +745,7 @@ class NerPipeline(Pipeline):
         device: int = -1,
         binary_output: bool = False,
         ignore_labels=["O"],
+        task: str = "",
     ):
         super().__init__(
             model=model,
@@ -734,6 +755,7 @@ class NerPipeline(Pipeline):
             args_parser=args_parser,
             device=device,
             binary_output=binary_output,
+            task=task,
         )
 
         self._basic_tokenizer = BasicTokenizer(do_lower_case=False)
@@ -896,6 +918,7 @@ class QuestionAnsweringPipeline(Pipeline):
         modelcard: Optional[ModelCard] = None,
         framework: Optional[str] = None,
         device: int = -1,
+        task: str = "",
         **kwargs
     ):
         super().__init__(
@@ -905,6 +928,7 @@ class QuestionAnsweringPipeline(Pipeline):
             framework=framework,
             args_parser=QuestionAnsweringArgumentHandler(),
             device=device,
+            task=task,
             **kwargs,
         )
 
@@ -1102,7 +1126,11 @@ class QuestionAnsweringPipeline(Pipeline):
             chars_idx += len(word) + 1
 
         # Join text with spaces
-        return {"answer": " ".join(words), "start": max(0, char_start_idx), "end": min(len(text), char_end_idx)}
+        return {
+            "answer": " ".join(words),
+            "start": max(0, char_start_idx),
+            "end": min(len(text), char_end_idx),
+        }
 
 
 class SummarizationPipeline(Pipeline):
@@ -1111,12 +1139,16 @@ class SummarizationPipeline(Pipeline):
 
     Usage::
 
+        # use bart in pytorch
         summarizer = pipeline("summarization")
-        summarizer("Sam Shleifer writes the best docstring examples in the whole world.")
+        summarizer("Sam Shleifer writes the best docstring examples in the whole world.", min_length=5, max_length=20)
+
+        # use t5 in tf
+        summarizer = pipeline("summarization", model="t5-base", tokenizer="t5-base", framework="tf")
+        summarizer("Sam Shleifer writes the best docstring examples in the whole world.", min_length=5, max_length=20)
 
     Supported Models:
-        The models that this pipeline can use are models that have been fine-tuned on a summarization task, which is
-        currently only ``BartForConditionalGeneration.from_pretrained('bart-large-cnn')``
+        The models that this pipeline can use are models that have been fine-tuned on a summarization task, which is currently, '`bart-large-cnn`', '`t5-small`', '`t5-base`', '`t5-large`', '`t5-3b`', '`t5-11b`'.
 
     Arguments:
         model (:obj:`str` or :obj:`~transformers.PreTrainedModel` or :obj:`~transformers.TFPreTrainedModel`, `optional`, defaults to :obj:`None`):
@@ -1147,17 +1179,8 @@ class SummarizationPipeline(Pipeline):
             on the associated CUDA device id.
     """
 
-    task = "summarization"
-
     def __call__(
-        self,
-        *documents,
-        return_tensors=False,
-        return_text=True,
-        max_length=142,
-        min_length=21,
-        clean_up_tokenization_spaces=False,
-        **generate_kwargs
+        self, *documents, return_tensors=False, return_text=True, clean_up_tokenization_spaces=False, **generate_kwargs
     ):
         r"""
         Args:
@@ -1165,10 +1188,6 @@ class SummarizationPipeline(Pipeline):
             return_text: (bool, default=True) whether to add a decoded "summary_text" to each result
             return_tensors: (bool, default=False) whether to return the raw "summary_token_ids" to each result
 
-            max_length: (`optional`) int
-                The max length of the sequence to be generated. Does not include tokens in input_ids.
-            min_len: (`optional`) int
-            no_repeat_ngram_size:  (`optional`) int. ban ngrams of this length from being repeated in the generated text
             clean_up_tokenization_spaces: (`optional`) bool whether to include extra spaces in the output
             **generate_kwargs: extra kwargs passed to `self.model.generate`_
 
@@ -1180,19 +1199,60 @@ class SummarizationPipeline(Pipeline):
 
         """
         assert return_tensors or return_text, "You must specify return_tensors=True or return_text=True"
-        if self.framework == "tf":
-            raise NotImplementedError("Tensorflow not supported")
-        with self.device_placement():
-            inputs = self._parse_and_tokenize(*documents)
-            inputs = self.ensure_tensor_on_device(**inputs)
-            summaries = self.model.generate(
-                inputs["input_ids"],
-                attention_mask=inputs["attention_mask"],
-                max_length=max_length,
-                min_length=min_length,
-                do_sample=False,
-                **generate_kwargs,
+        assert len(documents) > 0, "Please provide a document to summarize"
+
+        if self.framework == "tf" and "BartForConditionalGeneration" in self.model.__class__.__name__:
+            raise NotImplementedError(
+                "Tensorflow is not yet supported for Bart. Please consider using T5, e.g. `t5-base`"
             )
+
+        prefix = self.model.config.prefix if self.model.config.prefix is not None else ""
+
+        if isinstance(documents[0], list):
+            assert (
+                self.tokenizer.pad_token_id is not None
+            ), "Please make sure that the tokenizer has a pad_token_id when using a batch input"
+
+            documents = ([prefix + document for document in documents[0]],)
+            pad_to_max_length = True
+
+        elif isinstance(documents[0], str):
+            documents = (prefix + documents[0],)
+            pad_to_max_length = False
+        else:
+            raise ValueError(
+                " `documents[0]`: {} have the wrong format. The should be either of type `str` or type `list`".format(
+                    documents[0]
+                )
+            )
+
+        with self.device_placement():
+            inputs = self._parse_and_tokenize(*documents, pad_to_max_length=pad_to_max_length)
+
+            if self.framework == "pt":
+                inputs = self.ensure_tensor_on_device(**inputs)
+                input_length = inputs["input_ids"].shape[-1]
+            elif self.framework == "tf":
+                input_length = tf.shape(inputs["input_ids"])[-1].numpy()
+
+            if input_length < self.model.config.min_length // 2:
+                logger.warning(
+                    "Your min_length is set to {}, but you input_length is only {}. You might consider decreasing min_length manually, e.g. summarizer('...', min_length=10)".format(
+                        self.model.config.min_length, input_length
+                    )
+                )
+
+            if input_length < self.model.config.max_length:
+                logger.warning(
+                    "Your max_length is set to {}, but you input_length is only {}. You might consider decreasing max_length manually, e.g. summarizer('...', max_length=50)".format(
+                        self.model.config.max_length, input_length
+                    )
+                )
+
+            summaries = self.model.generate(
+                inputs["input_ids"], attention_mask=inputs["attention_mask"], **generate_kwargs,
+            )
+
             results = []
             for summary in summaries:
                 record = {}
@@ -1200,7 +1260,115 @@ class SummarizationPipeline(Pipeline):
                     record["summary_token_ids"] = summary
                 if return_text:
                     record["summary_text"] = self.tokenizer.decode(
-                        summary, skip_special_tokens=True, clean_up_tokenization_spaces=clean_up_tokenization_spaces
+                        summary, skip_special_tokens=True, clean_up_tokenization_spaces=clean_up_tokenization_spaces,
+                    )
+                results.append(record)
+            return results
+
+
+class TranslationPipeline(Pipeline):
+    """
+    Translates from one language to another.
+
+    Usage::
+        en_fr_translator = pipeline("translation_en_to_fr")
+        en_fr_translator("How old are you?")
+
+    Supported Models: "t5-small", "t5-base", "t5-large", "t5-3b", "t5-11b"
+
+    Arguments:
+        model (:obj:`str` or :obj:`~transformers.PreTrainedModel` or :obj:`~transformers.TFPreTrainedModel`, `optional`, defaults to :obj:`None`):
+            The model that will be used by the pipeline to make predictions. This can be :obj:`None`, a string
+            checkpoint identifier or an actual pre-trained model inheriting from
+            :class:`~transformers.PreTrainedModel` for PyTorch and :class:`~transformers.TFPreTrainedModel` for
+            TensorFlow.
+            If :obj:`None`, the default of the pipeline will be loaded.
+        tokenizer (:obj:`str` or :obj:`~transformers.PreTrainedTokenizer`, `optional`, defaults to :obj:`None`):
+            The tokenizer that will be used by the pipeline to encode data for the model. This can be :obj:`None`,
+            a string checkpoint identifier or an actual pre-trained tokenizer inheriting from
+            :class:`~transformers.PreTrainedTokenizer`.
+            If :obj:`None`, the default of the pipeline will be loaded.
+        modelcard (:obj:`str` or :class:`~transformers.ModelCard`, `optional`, defaults to :obj:`None`):
+            Model card attributed to the model for this pipeline.
+        framework (:obj:`str`, `optional`, defaults to :obj:`None`):
+            The framework to use, either "pt" for PyTorch or "tf" for TensorFlow. The specified framework must be
+            installed.
+            If no framework is specified, will default to the one currently installed. If no framework is specified
+            and both frameworks are installed, will default to PyTorch.
+        args_parser (:class:`~transformers.pipelines.ArgumentHandler`, `optional`, defaults to :obj:`None`):
+            Reference to the object in charge of parsing supplied pipeline parameters.
+        device (:obj:`int`, `optional`, defaults to :obj:`-1`):
+            Device ordinal for CPU/GPU supports. Setting this to -1 will leverage CPU, >=0 will run the model
+            on the associated CUDA device id.
+    """
+
+    def __call__(
+        self, *texts, return_tensors=False, return_text=True, clean_up_tokenization_spaces=False, **generate_kwargs
+    ):
+        r"""
+        Args:
+            *texts: (list of strings) texts to be translated
+            return_text: (bool, default=True) whether to add a decoded "translation_text" to each result
+            return_tensors: (bool, default=False) whether to return the raw "translation_token_ids" to each result
+
+            **generate_kwargs: extra kwargs passed to `self.model.generate`_
+
+        Returns:
+            list of dicts with 'translation_text' and/or 'translation_token_ids' for each text_to_translate
+        .. _`self.model.generate`:
+            https://huggingface.co/transformers/model_doc/bart.html#transformers.BartForConditionalGeneration.generate
+        """
+        assert return_tensors or return_text, "You must specify return_tensors=True or return_text=True"
+
+        prefix = self.model.config.prefix if self.model.config.prefix is not None else ""
+
+        if isinstance(texts[0], list):
+            assert (
+                self.tokenizer.pad_token_id is not None
+            ), "Please make sure that the tokenizer has a pad_token_id when using a batch input"
+            texts = ([prefix + text for text in texts[0]],)
+            pad_to_max_length = True
+
+        elif isinstance(texts[0], str):
+            texts = (prefix + texts[0],)
+            pad_to_max_length = False
+        else:
+            raise ValueError(
+                " `documents[0]`: {} have the wrong format. The should be either of type `str` or type `list`".format(
+                    texts[0]
+                )
+            )
+
+        with self.device_placement():
+            inputs = self._parse_and_tokenize(*texts, pad_to_max_length=pad_to_max_length)
+
+            if self.framework == "pt":
+                inputs = self.ensure_tensor_on_device(**inputs)
+                input_length = inputs["input_ids"].shape[-1]
+
+            elif self.framework == "tf":
+                input_length = tf.shape(inputs["input_ids"])[-1].numpy()
+
+            if input_length > 0.9 * self.model.config.max_length:
+                logger.warning(
+                    "Your input_length: {} is bigger than 0.9 * max_length: {}. You might consider increasing your max_length manually, e.g. translator('...', max_length=400)".format(
+                        input_length, self.model.config.max_length
+                    )
+                )
+
+            translations = self.model.generate(
+                inputs["input_ids"], attention_mask=inputs["attention_mask"], **generate_kwargs,
+            )
+            results = []
+            for translation in translations:
+                record = {}
+                if return_tensors:
+                    record["translation_token_ids"] = translation
+                if return_text:
+                    record["translation_text"] = self.tokenizer.decode(
+                        translation,
+                        skip_special_tokens=True,
+                        clean_up_tokenization_spaces=clean_up_tokenization_spaces,
                     )
                 results.append(record)
             return results
@@ -1266,14 +1434,44 @@ SUPPORTED_TASKS = {
     },
     "summarization": {
         "impl": SummarizationPipeline,
-        "pt": BartForConditionalGeneration if is_torch_available() else None,
-        "tf": None,
+        "tf": TFAutoModelWithLMHead if is_tf_available() else None,
+        "pt": AutoModelWithLMHead if is_torch_available() else None,
         "default": {
             "model": {"pt": "bart-large-cnn", "tf": None},
             "config": None,
             "tokenizer": ("bart-large-cnn", {"use_fast": False}),
         },
     },
+    "translation_en_to_fr": {
+        "impl": TranslationPipeline,
+        "tf": TFAutoModelWithLMHead if is_tf_available() else None,
+        "pt": AutoModelWithLMHead if is_torch_available() else None,
+        "default": {
+            "model": {"pt": "t5-base", "tf": "t5-base"},
+            "config": None,
+            "tokenizer": ("t5-base", {"use_fast": False}),
+        },
+    },
+    "translation_en_to_de": {
+        "impl": TranslationPipeline,
+        "tf": TFAutoModelWithLMHead if is_tf_available() else None,
+        "pt": AutoModelWithLMHead if is_torch_available() else None,
+        "default": {
+            "model": {"pt": "t5-base", "tf": "t5-base"},
+            "config": None,
+            "tokenizer": ("t5-base", {"use_fast": False}),
+        },
+    },
+    "translation_en_to_ro": {
+        "impl": TranslationPipeline,
+        "tf": TFAutoModelWithLMHead if is_tf_available() else None,
+        "pt": AutoModelWithLMHead if is_torch_available() else None,
+        "default": {
+            "model": {"pt": "t5-base", "tf": "t5-base"},
+            "config": None,
+            "tokenizer": ("t5-base", {"use_fast": False}),
+        },
+    },
 }
 
 
@@ -1361,7 +1559,7 @@ def pipeline(
     framework = framework or get_framework(model)
 
     targeted_task = SUPPORTED_TASKS[task]
-    task, model_class = targeted_task["impl"], targeted_task[framework]
+    task_class, model_class = targeted_task["impl"], targeted_task[framework]
 
     # Use default model/config/tokenizer for the task if no model is provided
     if model is None:
@@ -1422,4 +1620,4 @@ def pipeline(
             )
         model = model_class.from_pretrained(model, config=config, **model_kwargs)
 
-    return task(model=model, tokenizer=tokenizer, modelcard=modelcard, framework=framework, **kwargs)
+    return task_class(model=model, tokenizer=tokenizer, modelcard=modelcard, framework=framework, task=task, **kwargs,)

src/transformers/tokenization_albert.py

@@ -29,10 +29,10 @@ VOCAB_FILES_NAMES = {"vocab_file": "spiece.model"}
 
 PRETRAINED_VOCAB_FILES_MAP = {
     "vocab_file": {
-        "albert-base-v1": "https://s3.amazonaws.com/models.huggingface.co/bert/albert-base-spiece.model",
-        "albert-large-v1": "https://s3.amazonaws.com/models.huggingface.co/bert/albert-large-spiece.model",
-        "albert-xlarge-v1": "https://s3.amazonaws.com/models.huggingface.co/bert/albert-xlarge-spiece.model",
-        "albert-xxlarge-v1": "https://s3.amazonaws.com/models.huggingface.co/bert/albert-xxlarge-spiece.model",
+        "albert-base-v1": "https://s3.amazonaws.com/models.huggingface.co/bert/albert-base-v1-spiece.model",
+        "albert-large-v1": "https://s3.amazonaws.com/models.huggingface.co/bert/albert-large-v1-spiece.model",
+        "albert-xlarge-v1": "https://s3.amazonaws.com/models.huggingface.co/bert/albert-xlarge-v1-spiece.model",
+        "albert-xxlarge-v1": "https://s3.amazonaws.com/models.huggingface.co/bert/albert-xxlarge-v1-spiece.model",
         "albert-base-v2": "https://s3.amazonaws.com/models.huggingface.co/bert/albert-base-v2-spiece.model",
         "albert-large-v2": "https://s3.amazonaws.com/models.huggingface.co/bert/albert-large-v2-spiece.model",
         "albert-xlarge-v2": "https://s3.amazonaws.com/models.huggingface.co/bert/albert-xlarge-v2-spiece.model",

src/transformers/tokenization_t5.py

@@ -61,14 +61,35 @@ PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES = {
 
 class T5Tokenizer(PreTrainedTokenizer):
     """
-        SentencePiece based tokenizer. Peculiarities:
+        Constructs an XLNet tokenizer. Based on `SentencePiece <https://github.com/google/sentencepiece>`__
 
-            - requires `SentencePiece <https://github.com/google/sentencepiece>`_
-            - `extra_ids` add a number of extra ids added to the end of the vocabulary for use as sentinels.
+        This tokenizer inherits from :class:`~transformers.PreTrainedTokenizer` which contains most of the methods. Users
+        should refer to the superclass for more information regarding methods.
+
+        Args:
+            vocab_file (:obj:`string`):
+                `SentencePiece <https://github.com/google/sentencepiece>`__ file (generally has a .spm extension) that
+                contains the vocabulary necessary to instantiate a tokenizer.
+            eos_token (:obj:`string`, `optional`, defaults to "</s>"):
+                The end of sequence token.
+
+                .. note::
+
+                    When building a sequence using special tokens, this is not the token that is used for the end
+                    of sequence. The token used is the :obj:`sep_token`.
+            unk_token (:obj:`string`, `optional`, defaults to "<unk>"):
+                The unknown token. A token that is not in the vocabulary cannot be converted to an ID and is set to be this
+                token instead.
+            pad_token (:obj:`string`, `optional`, defaults to "<pad>"):
+                The token used for padding, for example when batching sequences of different lengths.
+            extra_ids (:obj:`List[str]`, `optional`, defaults to :obj:`100`):
+                Add a number of extra ids added to the end of the vocabulary for use as sentinels.
                 These tokens are accessible as `<extra_id_{%d}>` where `{%d}` is a number between 0 and extra_ids-1.
                 Extra tokens are indexed from the end of the vocabulary up to beginnning (<extra_id_0> is the last token in the vocabulary)
                 (like in T5 preprocessing
                 see: https://github.com/google-research/text-to-text-transfer-transformer/blob/9fd7b14a769417be33bc6c850f9598764913c833/t5/data/preprocessors.py#L2117)
+            additional_special_tokens (:obj:`List[str]`, `optional`, defaults to :obj:`None`):
+                Additional special tokens used by the tokenizer.
     """
 
     vocab_files_names = VOCAB_FILES_NAMES

src/transformers/tokenization_utils.py

@@ -1997,3 +1997,14 @@ class PreTrainedTokenizerFast(PreTrainedTokenizer):
             files = self._tokenizer.save(folder, name=file)
 
         return tuple(files)
+
+
+def trim_batch(
+    input_ids, pad_token_id, attention_mask=None,
+):
+    """Remove columns that are populated exclusively by pad_token_id"""
+    keep_column_mask = input_ids.ne(pad_token_id).any(dim=0)
+    if attention_mask is None:
+        return input_ids[:, keep_column_mask]
+    else:
+        return (input_ids[:, keep_column_mask], attention_mask[:, keep_column_mask])

templates/adding_a_new_example_script/run_xxx.py

@@ -28,20 +28,12 @@ from torch.utils.data.distributed import DistributedSampler
 from tqdm import tqdm, trange
 
 from transformers import (
+    MODEL_FOR_QUESTION_ANSWERING_MAPPING,
     WEIGHTS_NAME,
     AdamW,
-    BertConfig,
-    BertForQuestionAnswering,
-    BertTokenizer,
-    DistilBertConfig,
-    DistilBertForQuestionAnswering,
-    DistilBertTokenizer,
-    XLMConfig,
-    XLMForQuestionAnswering,
-    XLMTokenizer,
-    XLNetConfig,
-    XLNetForQuestionAnswering,
-    XLNetTokenizer,
+    AutoConfig,
+    AutoModelForQuestionAnswering,
+    AutoTokenizer,
     get_linear_schedule_with_warmup,
 )
 from utils_squad import (
@@ -68,16 +60,10 @@ except ImportError:
 
 logger = logging.getLogger(__name__)
 
-ALL_MODELS = sum(
-    (tuple(conf.pretrained_config_archive_map.keys()) for conf in (BertConfig, XLNetConfig, XLMConfig)), ()
-)
+MODEL_CONFIG_CLASSES = list(MODEL_FOR_QUESTION_ANSWERING_MAPPING.keys())
+MODEL_TYPES = tuple(conf.model_type for conf in MODEL_CONFIG_CLASSES)
 
-MODEL_CLASSES = {
-    "bert": (BertConfig, BertForQuestionAnswering, BertTokenizer),
-    "xlnet": (XLNetConfig, XLNetForQuestionAnswering, XLNetTokenizer),
-    "xlm": (XLMConfig, XLMForQuestionAnswering, XLMTokenizer),
-    "distilbert": (DistilBertConfig, DistilBertForQuestionAnswering, DistilBertTokenizer),
-}
+ALL_MODELS = sum((tuple(conf.pretrained_config_archive_map.keys()) for conf in MODEL_CONFIG_CLASSES), (),)
 
 
 def set_seed(args):
@@ -418,7 +404,7 @@ def main():
         default=None,
         type=str,
         required=True,
-        help="Model type selected in the list: " + ", ".join(MODEL_CLASSES.keys()),
+        help="Model type selected in the list: " + ", ".join(MODEL_TYPES),
     )
     parser.add_argument(
         "--model_name_or_path",
@@ -626,17 +612,16 @@ def main():
         # download model & vocab
 
     args.model_type = args.model_type.lower()
-    config_class, model_class, tokenizer_class = MODEL_CLASSES[args.model_type]
-    config = config_class.from_pretrained(
+    config = AutoConfig.from_pretrained(
         args.config_name if args.config_name else args.model_name_or_path,
         cache_dir=args.cache_dir if args.cache_dir else None,
     )
-    tokenizer = tokenizer_class.from_pretrained(
+    tokenizer = AutoTokenizer.from_pretrained(
         args.tokenizer_name if args.tokenizer_name else args.model_name_or_path,
         do_lower_case=args.do_lower_case,
         cache_dir=args.cache_dir if args.cache_dir else None,
     )
-    model = model_class.from_pretrained(
+    model = AutoModelForQuestionAnswering.from_pretrained(
         args.model_name_or_path,
         from_tf=bool(".ckpt" in args.model_name_or_path),
         config=config,
@@ -687,8 +672,8 @@ def main():
         torch.save(args, os.path.join(args.output_dir, "training_args.bin"))
 
         # Load a trained model and vocabulary that you have fine-tuned
-        model = model_class.from_pretrained(args.output_dir)
-        tokenizer = tokenizer_class.from_pretrained(args.output_dir, do_lower_case=args.do_lower_case)
+        model = AutoModelForQuestionAnswering.from_pretrained(args.output_dir)
+        tokenizer = AutoTokenizer.from_pretrained(args.output_dir, do_lower_case=args.do_lower_case)
         model.to(args.device)
 
     # Evaluation - we can ask to evaluate all the checkpoints (sub-directories) in a directory
@@ -706,7 +691,7 @@ def main():
         for checkpoint in checkpoints:
             # Reload the model
             global_step = checkpoint.split("-")[-1] if len(checkpoints) > 1 else ""
-            model = model_class.from_pretrained(checkpoint)
+            model = AutoModelForQuestionAnswering.from_pretrained(checkpoint)
             model.to(args.device)
 
             # Evaluate

tests/test_bart_memory.py

@@ -1,52 +0,0 @@
-import unittest
-
-import torch
-
-from tests.utils import require_torch, slow
-from transformers import BartTokenizer, BartModel
-from transformers.modeling_bart import shift_tokens_right
-
-DEFAULT_DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
-
-@require_torch
-class TestHface(unittest.TestCase):
-
-    @classmethod
-    def setUpClass(cls):
-        source_path = "test.source"
-        cls.lns = [" " + x.rstrip() for x in open(source_path).readlines()][:6]
-        tokenizer = BartTokenizer.from_pretrained('bart-large')
-        dct = tokenizer.batch_encode_plus(cls.lns, max_length=1024, return_tensors="pt", pad_to_max_length=True)
-        cls.ids = dct['input_ids'].to(DEFAULT_DEVICE)
-        cls.enc_mask = dct['attention_mask'].to(DEFAULT_DEVICE)
-        cls.prev_output_tokens = shift_tokens_right(cls.ids, 1).to(DEFAULT_DEVICE)
-        cls.model = BartForConditionalGeneration.from_pretrained('bart-large-cnn').half().to(DEFAULT_DEVICE).half()
-        return cls
-
-
-
-    @classmethod
-    def setUpClass(cls):
-        source_path = "test.source"
-        cls.lns = [" " + x.rstrip() for x in open(source_path).readlines()][:6]
-        tokenizer = BartTokenizer.from_pretrained('bart-large')
-        dct = tokenizer.batch_encode_plus(cls.lns, max_length=100, return_tensors="pt", pad_to_max_length=True)
-        cls.ids = dct['input_ids'].to(DEFAULT_DEVICE)
-        cls.prev_output_tokens = shift_tokens_right(cls.ids, 1).to(DEFAULT_DEVICE)
-        cls.model = BartModel.from_pretrained('bart-large').to(DEFAULT_DEVICE)
-        #cls.lns = pickle_load('/Users/shleifer/transformers_fork/lns.pkl')
-        return cls
-
-    def test_hf_fwd_batch(self):
-        bart = self.model
-        bart.reset_logs()
-        with torch.no_grad():
-            bart(self.ids)
-        try:
-            log_df = bart.combine_logs()
-            #log_df.to_csv('hf_batch_fwd_logs.csv')
-            bart.save_logs('hf_batch_fwd_logs.txt')
-            print(bart.summary)
-        except AttributeError as e:
-            print(e)
-

tests/test_modeling_auto.py

@@ -37,6 +37,8 @@ if is_torch_available():
         BertForSequenceClassification,
         AutoModelForQuestionAnswering,
         BertForQuestionAnswering,
+        AutoModelForTokenClassification,
+        BertForTokenClassification,
     )
     from transformers.modeling_bert import BERT_PRETRAINED_MODEL_ARCHIVE_MAP
     from transformers.modeling_auto import (
@@ -109,7 +111,7 @@ class AutoModelTest(unittest.TestCase):
             self.assertIsNotNone(model)
             self.assertIsInstance(model, BertForSequenceClassification)
 
-    # @slow
+    @slow
     def test_question_answering_model_from_pretrained(self):
         logging.basicConfig(level=logging.INFO)
         for model_name in list(BERT_PRETRAINED_MODEL_ARCHIVE_MAP.keys())[:1]:
@@ -122,6 +124,19 @@ class AutoModelTest(unittest.TestCase):
             self.assertIsNotNone(model)
             self.assertIsInstance(model, BertForQuestionAnswering)
 
+    @slow
+    def test_token_classification_model_from_pretrained(self):
+        logging.basicConfig(level=logging.INFO)
+        for model_name in list(BERT_PRETRAINED_MODEL_ARCHIVE_MAP.keys())[:1]:
+            config = AutoConfig.from_pretrained(model_name)
+            self.assertIsNotNone(config)
+            self.assertIsInstance(config, BertConfig)
+
+            model = AutoModelForTokenClassification.from_pretrained(model_name)
+            model, loading_info = AutoModelForTokenClassification.from_pretrained(model_name, output_loading_info=True)
+            self.assertIsNotNone(model)
+            self.assertIsInstance(model, BertForTokenClassification)
+
     def test_from_pretrained_identifier(self):
         logging.basicConfig(level=logging.INFO)
         model = AutoModelWithLMHead.from_pretrained(SMALL_MODEL_IDENTIFIER)

tests/test_modeling_bart.py

@@ -36,8 +36,8 @@ if is_torch_available():
     from transformers.modeling_bart import (
         BART_PRETRAINED_MODEL_ARCHIVE_MAP,
         shift_tokens_right,
+        invert_mask,
         _prepare_bart_decoder_inputs,
-        LARGE_NEGATIVE,
     )
     from transformers.tokenization_bart import BartTokenizer
 
@@ -113,7 +113,8 @@ class BARTModelTest(ModelTesterMixin, unittest.TestCase):
     test_pruning = False
     test_torchscript = False
     test_head_masking = False
-    test_resize_embeddings = False  # This requires inputs_dict['input_ids']
+    test_resize_embeddings = True  # This requires inputs_dict['input_ids']
+    test_missing_keys = False  # because BartForConditionalGeneration and BartModel now have identical state_dict
 
     def setUp(self):
         self.model_tester = ModelTester(self)
@@ -122,10 +123,9 @@ class BARTModelTest(ModelTesterMixin, unittest.TestCase):
     def test_config(self):
         self.config_tester.run_common_tests()
 
-    def test_advanced_inputs(self):
+    def test_initialization_more(self):
         # (config, input_ids, token_type_ids, input_mask, *unused) = \
         config, inputs_dict = self.model_tester.prepare_config_and_inputs_for_common()
-        decoder_input_ids, decoder_attn_mask = _prepare_bart_decoder_inputs(config, inputs_dict["input_ids"])
         model = BartModel(config)
         model.to(torch_device)
         model.eval()
@@ -141,9 +141,17 @@ class BARTModelTest(ModelTesterMixin, unittest.TestCase):
         _check_var(model.encoder.layers[0].fc1)
         _check_var(model.encoder.embed_positions)
 
+    def test_advanced_inputs(self):
+        config, inputs_dict = self.model_tester.prepare_config_and_inputs_for_common()
+        inputs_dict["input_ids"][:, -2:] = config.pad_token_id
+        decoder_input_ids, decoder_attn_mask, causal_mask = _prepare_bart_decoder_inputs(
+            config, inputs_dict["input_ids"]
+        )
+        model = BartModel(config).to(torch_device).eval()
+
         decoder_features_with_created_mask = model(**inputs_dict)[0]
         decoder_features_with_passed_mask = model(
-            decoder_attention_mask=decoder_attn_mask, decoder_input_ids=decoder_input_ids, **inputs_dict
+            decoder_attention_mask=invert_mask(decoder_attn_mask), decoder_input_ids=decoder_input_ids, **inputs_dict
         )[0]
         _assert_tensors_equal(decoder_features_with_passed_mask, decoder_features_with_created_mask)
         useless_mask = torch.zeros_like(decoder_attn_mask)
@@ -237,7 +245,7 @@ class BartHeadTests(unittest.TestCase):
         lm_labels = ids_tensor([batch_size, input_ids.shape[1]], self.vocab_size).to(torch_device)
         lm_model = BartForConditionalGeneration(config)
         lm_model.to(torch_device)
-        loss, logits, enc_features = lm_model(input_ids=input_ids, lm_labels=lm_labels, decoder_input_ids=input_ids)
+        loss, logits, enc_features = lm_model(input_ids=input_ids, lm_labels=lm_labels)
         expected_shape = (batch_size, input_ids.shape[1], config.vocab_size)
         self.assertEqual(logits.shape, expected_shape)
         self.assertIsInstance(loss.item(), float)
@@ -335,41 +343,39 @@ class BartHeadTests(unittest.TestCase):
         model.generate(num_beams=4, do_sample=True, early_stopping=False, num_return_sequences=3)
 
     def test_dummy_inputs(self):
-        config, *_ = self._get_config_and_data(output_past=True)
+        config, *_ = self._get_config_and_data()
         model = BartForConditionalGeneration(config).eval().to(torch_device)
         model(**model.dummy_inputs)
 
     def test_prepare_bart_decoder_inputs(self):
         config, *_ = self._get_config_and_data(output_past=False)
-        input_ids = _long_tensor(([4, 4, 2]))  # only used for .device if decoder_input_ids is passed
+        input_ids = _long_tensor(([4, 4, 2]))
         decoder_input_ids = _long_tensor([[26388, 2, config.pad_token_id]])
-        ignore = LARGE_NEGATIVE
-        decoder_input_ids, decoder_attn_mask = _prepare_bart_decoder_inputs(config, input_ids, decoder_input_ids)
-        expected_mask = torch.tensor(
-            [
-                [0, ignore, ignore],
-                [0, 0, ignore],
-                [ignore, ignore, ignore],  # never attend to the final token, because its pad
-            ]
-        ).to(input_ids.device)
-        self.assertEqual(decoder_attn_mask.size(), (1, 1, 3, 3))
-        self.assertTrue(torch.eq(expected_mask, decoder_attn_mask).all())
-
-        # Test no causal mask
-        config, *_ = self._get_config_and_data(output_past=True)
-        expected_just_padding_mask = torch.tensor(
-            [[0, 0, 0], [0, 0, 0], [ignore, ignore, ignore]]  # never attend to the final token, because its pad
-        ).to(input_ids.device)
-        _, decoder_attn_mask_no_causal_mask = _prepare_bart_decoder_inputs(config, input_ids, decoder_input_ids)
-        self.assertEqual(decoder_attn_mask_no_causal_mask.size(), (1, 1, 3, 3))
-        self.assertTrue(torch.eq(expected_just_padding_mask, decoder_attn_mask_no_causal_mask).all())
-
-        decoder_input_ids = _long_tensor([[0, 26388, 4133, 2]])
-        # Attend to everything if no pad tokens and no causal mask
-        _, decoder_attn_mask_no_padding_no_causal_mask = _prepare_bart_decoder_inputs(
+        ignore = float("-inf")
+        decoder_input_ids, decoder_attn_mask, causal_mask = _prepare_bart_decoder_inputs(
             config, input_ids, decoder_input_ids
         )
-        self.assertTrue(torch.eq(decoder_attn_mask_no_padding_no_causal_mask, 0).all())
+        expected_causal_mask = torch.tensor(
+            [[0, ignore, ignore], [0, 0, ignore], [0, 0, 0]]  # never attend to the final token, because its pad
+        ).to(input_ids.device)
+        self.assertEqual(decoder_attn_mask.size(), decoder_input_ids.size())
+        self.assertTrue(torch.eq(expected_causal_mask, causal_mask).all())
+
+    def test_resize_tokens_embeddings_more(self):
+        config, input_ids, _ = self._get_config_and_data()
+
+        def _get_embs(m):
+            return (m.get_input_embeddings().weight.data.clone(), m.get_output_embeddings().weight.data.clone())
+
+        model = BartForConditionalGeneration(config).eval().to(torch_device)
+        input, output = _get_embs(model)
+        self.assertTrue(torch.eq(input, output).all())
+        new_vocab_size = 45
+        model.resize_token_embeddings(new_vocab_size)
+        input_new, output_new = _get_embs(model)
+        self.assertEqual(input_new.shape, (new_vocab_size, config.d_model))
+        self.assertEqual(output_new.shape, (new_vocab_size, config.d_model))
+        self.assertTrue(torch.eq(input_new, output_new).all())
 
 
 def _assert_tensors_equal(a, b, atol=1e-12, prefix=""):

tests/test_modeling_camembert.py

@@ -0,0 +1,50 @@
+# coding=utf-8
+# Copyright 2018 The Google AI Language Team Authors.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import unittest
+
+from transformers import is_torch_available
+
+from .utils import require_torch, slow, torch_device
+
+
+if is_torch_available():
+    import torch
+    from transformers import CamembertModel
+
+
+@require_torch
+class CamembertModelIntegrationTest(unittest.TestCase):
+    @slow
+    def test_output_embeds_base_model(self):
+        model = CamembertModel.from_pretrained("camembert-base")
+
+        input_ids = torch.tensor(
+            [[5, 121, 11, 660, 16, 730, 25543, 110, 83, 6]], device=torch_device, dtype=torch.long,
+        )  # J'aime le camembert !
+        output = model(input_ids)[0]
+        expected_shape = torch.Size((1, 10, 768))
+        self.assertEqual(output.shape, expected_shape)
+        # compare the actual values for a slice.
+        expected_slice = torch.tensor(
+            [[[-0.0254, 0.0235, 0.1027], [0.0606, -0.1811, -0.0418], [-0.1561, -0.1127, 0.2687]]],
+            device=torch_device,
+            dtype=torch.float,
+        )
+        # camembert = torch.hub.load('pytorch/fairseq', 'camembert.v0')
+        # camembert.eval()
+        # expected_slice = roberta.model.forward(input_ids)[0][:, :3, :3].detach()
+
+        self.assertTrue(torch.allclose(output[:, :3, :3], expected_slice, atol=1e-4))

tests/test_modeling_common.py

@@ -58,6 +58,7 @@ class ModelTesterMixin:
     test_pruning = True
     test_resize_embeddings = True
     test_head_masking = True
+    test_missing_keys = True
     is_encoder_decoder = False
 
     def test_save_load(self):
@@ -527,6 +528,8 @@ class ModelTesterMixin:
             self.assertTrue(x is None or isinstance(x, torch.nn.Linear))
 
     def test_correct_missing_keys(self):
+        if not self.test_missing_keys:
+            return
         config, _ = self.model_tester.prepare_config_and_inputs_for_common()
 
         for model_class in self.all_model_classes:

tests/test_modeling_tf_camembert.py

@@ -0,0 +1,50 @@
+# coding=utf-8
+# Copyright 2018 The Google AI Language Team Authors.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import unittest
+
+from transformers import is_tf_available
+
+from .utils import require_tf, slow
+
+
+if is_tf_available():
+    import tensorflow as tf
+    import numpy as np
+    from transformers import TFCamembertModel
+
+
+@require_tf
+class TFCamembertModelIntegrationTest(unittest.TestCase):
+    @slow
+    def test_output_embeds_base_model(self):
+        model = TFCamembertModel.from_pretrained("jplu/tf-camembert-base")
+
+        input_ids = tf.convert_to_tensor(
+            [[5, 121, 11, 660, 16, 730, 25543, 110, 83, 6]], dtype=tf.int32,
+        )  # J'aime le camembert !"
+
+        output = model(input_ids)[0]
+        expected_shape = tf.TensorShape((1, 10, 768))
+        self.assertEqual(output.shape, expected_shape)
+        # compare the actual values for a slice.
+        expected_slice = tf.convert_to_tensor(
+            [[[-0.0254, 0.0235, 0.1027], [0.0606, -0.1811, -0.0418], [-0.1561, -0.1127, 0.2687]]], dtype=tf.float32,
+        )
+        # camembert = torch.hub.load('pytorch/fairseq', 'camembert.v0')
+        # camembert.eval()
+        # expected_slice = roberta.model.forward(input_ids)[0][:, :3, :3].detach()
+
+        self.assertTrue(np.allclose(output[:, :3, :3].numpy(), expected_slice.numpy(), atol=1e-4))

tests/test_modeling_xlm.py

@@ -29,6 +29,7 @@ if is_torch_available():
         XLMConfig,
         XLMModel,
         XLMWithLMHeadModel,
+        XLMForTokenClassification,
         XLMForQuestionAnswering,
         XLMForSequenceClassification,
         XLMForQuestionAnsweringSimple,
@@ -350,6 +351,32 @@ class XLMModelTest(ModelTesterMixin, unittest.TestCase):
                 list(result["logits"].size()), [self.batch_size, self.type_sequence_label_size]
             )
 
+        def create_and_check_xlm_for_token_classification(
+            self,
+            config,
+            input_ids,
+            token_type_ids,
+            input_lengths,
+            sequence_labels,
+            token_labels,
+            is_impossible_labels,
+            input_mask,
+        ):
+            config.num_labels = self.num_labels
+            model = XLMForTokenClassification(config)
+            model.to(torch_device)
+            model.eval()
+
+            loss, logits = model(input_ids, attention_mask=input_mask, labels=token_labels)
+            result = {
+                "loss": loss,
+                "logits": logits,
+            }
+            self.parent.assertListEqual(
+                list(result["logits"].size()), [self.batch_size, self.seq_length, self.num_labels]
+            )
+            self.check_loss_output(result)
+
         def prepare_config_and_inputs_for_common(self):
             config_and_inputs = self.prepare_config_and_inputs()
             (
@@ -392,6 +419,10 @@ class XLMModelTest(ModelTesterMixin, unittest.TestCase):
         config_and_inputs = self.model_tester.prepare_config_and_inputs()
         self.model_tester.create_and_check_xlm_sequence_classif(*config_and_inputs)
 
+    def test_xlm_for_token_classification(self):
+        config_and_inputs = self.model_tester.prepare_config_and_inputs()
+        self.model_tester.create_and_check_xlm_for_token_classification(*config_and_inputs)
+
     @slow
     def test_model_from_pretrained(self):
         for model_name in list(XLM_PRETRAINED_MODEL_ARCHIVE_MAP.keys())[:1]:

tests/test_pipelines.py

@@ -78,6 +78,15 @@ TF_FILL_MASK_FINETUNED_MODELS = [
     (("distilroberta-base", {"use_fast": False}), "distilroberta-base", None),
 ]
 
+SUMMARIZATION_FINETUNED_MODELS = {("bart-large-cnn", "bart-large-cnn"), ("t5-small", "t5-small")}
+TF_SUMMARIZATION_FINETUNED_MODELS = {("t5-small", "t5-small")}
+
+TRANSLATION_FINETUNED_MODELS = {
+    ("t5-small", "t5-small", "translation_en_to_de"),
+    ("t5-small", "t5-small", "translation_en_to_ro"),
+}
+TF_TRANSLATION_FINETUNED_MODELS = {("t5-small", "t5-small", "translation_en_to_fr")}
+
 
 class MonoColumnInputTestCase(unittest.TestCase):
     def _test_mono_column_pipeline(
@@ -252,10 +261,44 @@ class MonoColumnInputTestCase(unittest.TestCase):
         valid_inputs = ["A string like this", ["list of strings entry 1", "list of strings v2"]]
         invalid_inputs = [4, "<mask>"]
         mandatory_keys = ["summary_text"]
-        nlp = pipeline(task="summarization")
-        self._test_mono_column_pipeline(
-            nlp, valid_inputs, invalid_inputs, mandatory_keys,
-        )
+        for model, tokenizer in SUMMARIZATION_FINETUNED_MODELS:
+            nlp = pipeline(task="summarization", model=model, tokenizer=tokenizer)
+            self._test_mono_column_pipeline(
+                nlp, valid_inputs, invalid_inputs, mandatory_keys,
+            )
+
+    @require_tf
+    def test_tf_summarization(self):
+        valid_inputs = ["A string like this", ["list of strings entry 1", "list of strings v2"]]
+        invalid_inputs = [4, "<mask>"]
+        mandatory_keys = ["summary_text"]
+        for model, tokenizer in TF_SUMMARIZATION_FINETUNED_MODELS:
+            nlp = pipeline(task="summarization", model=model, tokenizer=tokenizer, framework="tf")
+            self._test_mono_column_pipeline(
+                nlp, valid_inputs, invalid_inputs, mandatory_keys,
+            )
+
+    @require_torch
+    def test_translation(self):
+        valid_inputs = ["A string like this", ["list of strings entry 1", "list of strings v2"]]
+        invalid_inputs = [4, "<mask>"]
+        mandatory_keys = ["translation_text"]
+        for model, tokenizer, task in TRANSLATION_FINETUNED_MODELS:
+            nlp = pipeline(task=task, model=model, tokenizer=tokenizer)
+            self._test_mono_column_pipeline(
+                nlp, valid_inputs, invalid_inputs, mandatory_keys,
+            )
+
+    @require_tf
+    def test_tf_translation(self):
+        valid_inputs = ["A string like this", ["list of strings entry 1", "list of strings v2"]]
+        invalid_inputs = [4, "<mask>"]
+        mandatory_keys = ["translation_text"]
+        for model, tokenizer, task in TF_TRANSLATION_FINETUNED_MODELS:
+            nlp = pipeline(task=task, model=model, tokenizer=tokenizer, framework="tf")
+            self._test_mono_column_pipeline(
+                nlp, valid_inputs, invalid_inputs, mandatory_keys,
+            )
 
 
 class MultiColumnInputTestCase(unittest.TestCase):