diff --git a/examples/benchmarks.py b/examples/benchmarks.py
index 4ef0640e..07de19d4 100644
--- a/examples/benchmarks.py
+++ b/examples/benchmarks.py
@@ -36,215 +36,215 @@ if is_torch_available():
from transformers import AutoModel
-input_text = """Bent over their instruments, three hundred Fertilizers were plunged, as
-the Director of Hatcheries and Conditioning entered the room, in the
+input_text = """Bent over their instruments, three hundred Fertilizers were plunged, as
+the Director of Hatcheries and Conditioning entered the room, in the
-scarcely breathing silence, the absent-minded, soliloquizing hum or
-whistle, of absorbed concentration. A troop of newly arrived students,
-very young, pink and callow, followed nervously, rather abjectly, at the
-Director's heels. Each of them carried a notebook, in which, whenever
-the great man spoke, he desperately scribbled. Straight from the
-horse's mouth. It was a rare privilege. The D. H. C. for Central London
-always made a point of personally conducting his new students round
-the various departments.
+scarcely breathing silence, the absent-minded, soliloquizing hum or
+whistle, of absorbed concentration. A troop of newly arrived students,
+very young, pink and callow, followed nervously, rather abjectly, at the
+Director's heels. Each of them carried a notebook, in which, whenever
+the great man spoke, he desperately scribbled. Straight from the
+horse's mouth. It was a rare privilege. The D. H. C. for Central London
+always made a point of personally conducting his new students round
+the various departments.
-"Just to give you a general idea," he would explain to them. For of
-course some sort of general idea they must have, if they were to do
-their work intelligently-though as little of one, if they were to be good
-and happy members of society, as possible. For particulars, as every
-one knows, make for virtue and happiness; generalities are intellectu-
-ally necessary evils. Not philosophers but fret-sawyers and stamp col-
-lectors compose the backbone of society.
+"Just to give you a general idea," he would explain to them. For of
+course some sort of general idea they must have, if they were to do
+their work intelligently-though as little of one, if they were to be good
+and happy members of society, as possible. For particulars, as every
+one knows, make for virtue and happiness; generalities are intellectu-
+ally necessary evils. Not philosophers but fret-sawyers and stamp col-
+lectors compose the backbone of society.
-"To-morrow," he would add, smiling at them with a slightly menacing
-geniality, "you'll be settling down to serious work. You won't have time
-for generalities. Meanwhile ..."
+"To-morrow," he would add, smiling at them with a slightly menacing
+geniality, "you'll be settling down to serious work. You won't have time
+for generalities. Meanwhile ..."
-Meanwhile, it was a privilege. Straight from the horse's mouth into the
-notebook. The boys scribbled like mad.
+Meanwhile, it was a privilege. Straight from the horse's mouth into the
+notebook. The boys scribbled like mad.
-Tall and rather thin but upright, the Director advanced into the room.
-He had a long chin and big rather prominent teeth, just covered, when
-he was not talking, by his full, floridly curved lips. Old, young? Thirty?
-Fifty? Fifty-five? It was hard to say. And anyhow the question didn't
-arise; in this year of stability, A. F. 632, it didn't occur to you to ask it.
+Tall and rather thin but upright, the Director advanced into the room.
+He had a long chin and big rather prominent teeth, just covered, when
+he was not talking, by his full, floridly curved lips. Old, young? Thirty?
+Fifty? Fifty-five? It was hard to say. And anyhow the question didn't
+arise; in this year of stability, A. F. 632, it didn't occur to you to ask it.
-"I shall begin at the beginning," said the D.H.C. and the more zealous
-students recorded his intention in their notebooks: Begin at the begin-
-ning. "These," he waved his hand, "are the incubators." And opening
-an insulated door he showed them racks upon racks of numbered test-
-tubes. "The week's supply of ova. Kept," he explained, "at blood heat;
-whereas the male gametes," and here he opened another door, "they
-have to be kept at thirty-five instead of thirty-seven. Full blood heat
-sterilizes." Rams wrapped in theremogene beget no lambs.
+"I shall begin at the beginning," said the D.H.C. and the more zealous
+students recorded his intention in their notebooks: Begin at the begin-
+ning. "These," he waved his hand, "are the incubators." And opening
+an insulated door he showed them racks upon racks of numbered test-
+tubes. "The week's supply of ova. Kept," he explained, "at blood heat;
+whereas the male gametes," and here he opened another door, "they
+have to be kept at thirty-five instead of thirty-seven. Full blood heat
+sterilizes." Rams wrapped in theremogene beget no lambs.
-Still leaning against the incubators he gave them, while the pencils
-scurried illegibly across the pages, a brief description of the modern
+Still leaning against the incubators he gave them, while the pencils
+scurried illegibly across the pages, a brief description of the modern
-fertilizing process; spoke first, of course, of its surgical introduc-
-tion-"the operation undergone voluntarily for the good of Society, not
-to mention the fact that it carries a bonus amounting to six months'
-salary"; continued with some account of the technique for preserving
-the excised ovary alive and actively developing; passed on to a consid-
-eration of optimum temperature, salinity, viscosity; referred to the liq-
-uor in which the detached and ripened eggs were kept; and, leading
-his charges to the work tables, actually showed them how this liquor
-was drawn off from the test-tubes; how it was let out drop by drop
-onto the specially warmed slides of the microscopes; how the eggs
-which it contained were inspected for abnormalities, counted and
-transferred to a porous receptacle; how (and he now took them to
-watch the operation) this receptacle was immersed in a warm bouillon
-containing free-swimming spermatozoa-at a minimum concentration
-of one hundred thousand per cubic centimetre, he insisted; and how,
-after ten minutes, the container was lifted out of the liquor and its
-contents re-examined; how, if any of the eggs remained unfertilized, it
-was again immersed, and, if necessary, yet again; how the fertilized
-ova went back to the incubators; where the Alphas and Betas re-
-mained until definitely bottled; while the Gammas, Deltas and Epsilons
-were brought out again, after only thirty-six hours, to undergo Bo-
-kanovsky's Process.
+fertilizing process; spoke first, of course, of its surgical introduc-
+tion-"the operation undergone voluntarily for the good of Society, not
+to mention the fact that it carries a bonus amounting to six months'
+salary"; continued with some account of the technique for preserving
+the excised ovary alive and actively developing; passed on to a consid-
+eration of optimum temperature, salinity, viscosity; referred to the liq-
+uor in which the detached and ripened eggs were kept; and, leading
+his charges to the work tables, actually showed them how this liquor
+was drawn off from the test-tubes; how it was let out drop by drop
+onto the specially warmed slides of the microscopes; how the eggs
+which it contained were inspected for abnormalities, counted and
+transferred to a porous receptacle; how (and he now took them to
+watch the operation) this receptacle was immersed in a warm bouillon
+containing free-swimming spermatozoa-at a minimum concentration
+of one hundred thousand per cubic centimetre, he insisted; and how,
+after ten minutes, the container was lifted out of the liquor and its
+contents re-examined; how, if any of the eggs remained unfertilized, it
+was again immersed, and, if necessary, yet again; how the fertilized
+ova went back to the incubators; where the Alphas and Betas re-
+mained until definitely bottled; while the Gammas, Deltas and Epsilons
+were brought out again, after only thirty-six hours, to undergo Bo-
+kanovsky's Process.
-"Bokanovsky's Process," repeated the Director, and the students un-
-derlined the words in their little notebooks.
+"Bokanovsky's Process," repeated the Director, and the students un-
+derlined the words in their little notebooks.
-One egg, one embryo, one adult-normality. But a bokanovskified egg
-will bud, will proliferate, will divide. From eight to ninety-six buds, and
-every bud will grow into a perfectly formed embryo, and every embryo
-into a full-sized adult. Making ninety-six human beings grow where
-only one grew before. Progress.
+One egg, one embryo, one adult-normality. But a bokanovskified egg
+will bud, will proliferate, will divide. From eight to ninety-six buds, and
+every bud will grow into a perfectly formed embryo, and every embryo
+into a full-sized adult. Making ninety-six human beings grow where
+only one grew before. Progress.
-"Essentially," the D.H.C. concluded, "bokanovskification consists of a
-series of arrests of development. We check the normal growth and,
-paradoxically enough, the egg responds by budding."
+"Essentially," the D.H.C. concluded, "bokanovskification consists of a
+series of arrests of development. We check the normal growth and,
+paradoxically enough, the egg responds by budding."
-Responds by budding. The pencils were busy.
+Responds by budding. The pencils were busy.
-He pointed. On a very slowly moving band a rack-full of test-tubes was
-entering a large metal box, another, rack-full was emerging. Machinery
-faintly purred. It took eight minutes for the tubes to go through, he
+He pointed. On a very slowly moving band a rack-full of test-tubes was
+entering a large metal box, another, rack-full was emerging. Machinery
+faintly purred. It took eight minutes for the tubes to go through, he
-told them. Eight minutes of hard X-rays being about as much as an
-egg can stand. A few died; of the rest, the least susceptible divided
-into two; most put out four buds; some eight; all were returned to the
-incubators, where the buds began to develop; then, after two days,
-were suddenly chilled, chilled and checked. Two, four, eight, the buds
-in their turn budded; and having budded were dosed almost to death
-with alcohol; consequently burgeoned again and having budded-bud
-out of bud out of bud-were thereafter-further arrest being generally
-fatal-left to develop in peace. By which time the original egg was in a
-fair way to becoming anything from eight to ninety-six embryos- a
-prodigious improvement, you will agree, on nature. Identical twins-but
-not in piddling twos and threes as in the old viviparous days, when an
-egg would sometimes accidentally divide; actually by dozens, by
-scores at a time.
+told them. Eight minutes of hard X-rays being about as much as an
+egg can stand. A few died; of the rest, the least susceptible divided
+into two; most put out four buds; some eight; all were returned to the
+incubators, where the buds began to develop; then, after two days,
+were suddenly chilled, chilled and checked. Two, four, eight, the buds
+in their turn budded; and having budded were dosed almost to death
+with alcohol; consequently burgeoned again and having budded-bud
+out of bud out of bud-were thereafter-further arrest being generally
+fatal-left to develop in peace. By which time the original egg was in a
+fair way to becoming anything from eight to ninety-six embryos- a
+prodigious improvement, you will agree, on nature. Identical twins-but
+not in piddling twos and threes as in the old viviparous days, when an
+egg would sometimes accidentally divide; actually by dozens, by
+scores at a time.
-"Scores," the Director repeated and flung out his arms, as though he
-were distributing largesse. "Scores."
+"Scores," the Director repeated and flung out his arms, as though he
+were distributing largesse. "Scores."
-But one of the students was fool enough to ask where the advantage
-lay.
+But one of the students was fool enough to ask where the advantage
+lay.
-"My good boy!" The Director wheeled sharply round on him. "Can't you
-see? Can't you see?" He raised a hand; his expression was solemn.
-"Bokanovsky's Process is one of the major instruments of social stabil-
-ity!"
+"My good boy!" The Director wheeled sharply round on him. "Can't you
+see? Can't you see?" He raised a hand; his expression was solemn.
+"Bokanovsky's Process is one of the major instruments of social stabil-
+ity!"
-Major instruments of social stability.
+Major instruments of social stability.
-Standard men and women; in uniform batches. The whole of a small
-factory staffed with the products of a single bokanovskified egg.
+Standard men and women; in uniform batches. The whole of a small
+factory staffed with the products of a single bokanovskified egg.
-"Ninety-six identical twins working ninety-six identical machines!" The
-voice was almost tremulous with enthusiasm. "You really know where
-you are. For the first time in history." He quoted the planetary motto.
-"Community, Identity, Stability." Grand words. "If we could bo-
-kanovskify indefinitely the whole problem would be solved."
+"Ninety-six identical twins working ninety-six identical machines!" The
+voice was almost tremulous with enthusiasm. "You really know where
+you are. For the first time in history." He quoted the planetary motto.
+"Community, Identity, Stability." Grand words. "If we could bo-
+kanovskify indefinitely the whole problem would be solved."
-Solved by standard Gammas, unvarying Deltas, uniform Epsilons. Mil-
-lions of identical twins. The principle of mass production at last applied
-to biology.
+Solved by standard Gammas, unvarying Deltas, uniform Epsilons. Mil-
+lions of identical twins. The principle of mass production at last applied
+to biology.
-"But, alas," the Director shook his head, "we can't bokanovskify indefi-
-nitely."
+"But, alas," the Director shook his head, "we can't bokanovskify indefi-
+nitely."
-Ninety-six seemed to be the limit; seventy-two a good average. From
-the same ovary and with gametes of the same male to manufacture as
-many batches of identical twins as possible-that was the best (sadly a
-second best) that they could do. And even that was difficult.
+Ninety-six seemed to be the limit; seventy-two a good average. From
+the same ovary and with gametes of the same male to manufacture as
+many batches of identical twins as possible-that was the best (sadly a
+second best) that they could do. And even that was difficult.
-"For in nature it takes thirty years for two hundred eggs to reach ma-
-turity. But our business is to stabilize the population at this moment,
-here and now. Dribbling out twins over a quarter of a century-what
-would be the use of that?"
+"For in nature it takes thirty years for two hundred eggs to reach ma-
+turity. But our business is to stabilize the population at this moment,
+here and now. Dribbling out twins over a quarter of a century-what
+would be the use of that?"
-Obviously, no use at all. But Podsnap's Technique had immensely ac-
-celerated the process of ripening. They could make sure of at least a
-hundred and fifty mature eggs within two years. Fertilize and bo-
-kanovskify-in other words, multiply by seventy-two-and you get an
-average of nearly eleven thousand brothers and sisters in a hundred
-and fifty batches of identical twins, all within two years of the same
-age.
+Obviously, no use at all. But Podsnap's Technique had immensely ac-
+celerated the process of ripening. They could make sure of at least a
+hundred and fifty mature eggs within two years. Fertilize and bo-
+kanovskify-in other words, multiply by seventy-two-and you get an
+average of nearly eleven thousand brothers and sisters in a hundred
+and fifty batches of identical twins, all within two years of the same
+age.
-"And in exceptional cases we can make one ovary yield us over fifteen
-thousand adult individuals."
+"And in exceptional cases we can make one ovary yield us over fifteen
+thousand adult individuals."
-Beckoning to a fair-haired, ruddy young man who happened to be
-passing at the moment. "Mr. Foster," he called. The ruddy young man
-approached. "Can you tell us the record for a single ovary, Mr. Foster?"
+Beckoning to a fair-haired, ruddy young man who happened to be
+passing at the moment. "Mr. Foster," he called. The ruddy young man
+approached. "Can you tell us the record for a single ovary, Mr. Foster?"
-"Sixteen thousand and twelve in this Centre," Mr. Foster replied with-
-out hesitation. He spoke very quickly, had a vivacious blue eye, and
-took an evident pleasure in quoting figures. "Sixteen thousand and
-twelve; in one hundred and eighty-nine batches of identicals. But of
-course they've done much better," he rattled on, "in some of the tropi-
-cal Centres. Singapore has often produced over sixteen thousand five
-hundred; and Mombasa has actually touched the seventeen thousand
-mark. But then they have unfair advantages. You should see the way a
-negro ovary responds to pituitary! It's quite astonishing, when you're
-used to working with European material. Still," he added, with a laugh
-(but the light of combat was in his eyes and the lift of his chin was
-challenging), "still, we mean to beat them if we can. I'm working on a
-wonderful Delta-Minus ovary at this moment. Only just eighteen
+"Sixteen thousand and twelve in this Centre," Mr. Foster replied with-
+out hesitation. He spoke very quickly, had a vivacious blue eye, and
+took an evident pleasure in quoting figures. "Sixteen thousand and
+twelve; in one hundred and eighty-nine batches of identicals. But of
+course they've done much better," he rattled on, "in some of the tropi-
+cal Centres. Singapore has often produced over sixteen thousand five
+hundred; and Mombasa has actually touched the seventeen thousand
+mark. But then they have unfair advantages. You should see the way a
+negro ovary responds to pituitary! It's quite astonishing, when you're
+used to working with European material. Still," he added, with a laugh
+(but the light of combat was in his eyes and the lift of his chin was
+challenging), "still, we mean to beat them if we can. I'm working on a
+wonderful Delta-Minus ovary at this moment. Only just eighteen
-months old. Over twelve thousand seven hundred children already, ei-
-ther decanted or in embryo. And still going strong. We'll beat them
-yet."
+months old. Over twelve thousand seven hundred children already, ei-
+ther decanted or in embryo. And still going strong. We'll beat them
+yet."
-"That's the spirit I like!" cried the Director, and clapped Mr. Foster on
-the shoulder. "Come along with us, and give these boys the benefit of
-your expert knowledge."
+"That's the spirit I like!" cried the Director, and clapped Mr. Foster on
+the shoulder. "Come along with us, and give these boys the benefit of
+your expert knowledge."
-Mr. Foster smiled modestly. "With pleasure." They went.
-In the Bottling Room all was harmonious bustle and ordered activity.
-Flaps of fresh sow's peritoneum ready cut to the proper size came
-shooting up in little lifts from the Organ Store in the sub-basement.
-Whizz and then, click! the lift-hatches hew open; the bottle-liner had
-only to reach out a hand, take the flap, insert, smooth-down, and be-
-fore the lined bottle had had time to travel out of reach along the end-
-less band, whizz, click! another flap of peritoneum had shot up from
-the depths, ready to be slipped into yet another bottle, the next of that
-slow interminable procession on the band.
+Mr. Foster smiled modestly. "With pleasure." They went.
+In the Bottling Room all was harmonious bustle and ordered activity.
+Flaps of fresh sow's peritoneum ready cut to the proper size came
+shooting up in little lifts from the Organ Store in the sub-basement.
+Whizz and then, click! the lift-hatches hew open; the bottle-liner had
+only to reach out a hand, take the flap, insert, smooth-down, and be-
+fore the lined bottle had had time to travel out of reach along the end-
+less band, whizz, click! another flap of peritoneum had shot up from
+the depths, ready to be slipped into yet another bottle, the next of that
+slow interminable procession on the band.
-Next to the Liners stood the Matriculators. The procession advanced;
-one by one the eggs were transferred from their test-tubes to the
-larger containers; deftly the peritoneal lining was slit, the morula
-dropped into place, the saline solution poured in ... and already the
-bottle had passed, and it was the turn of the labellers. Heredity, date
-of fertilization, membership of Bokanovsky Group-details were trans-
-ferred from test-tube to bottle. No longer anonymous, but named,
-identified, the procession marched slowly on; on through an opening in
-the wall, slowly on into the Social Predestination Room.
-"Eighty-eight cubic metres of card-index," said Mr. Foster with relish,
+Next to the Liners stood the Matriculators. The procession advanced;
+one by one the eggs were transferred from their test-tubes to the
+larger containers; deftly the peritoneal lining was slit, the morula
+dropped into place, the saline solution poured in ... and already the
+bottle had passed, and it was the turn of the labellers. Heredity, date
+of fertilization, membership of Bokanovsky Group-details were trans-
+ferred from test-tube to bottle. No longer anonymous, but named,
+identified, the procession marched slowly on; on through an opening in
+the wall, slowly on into the Social Predestination Room.
+"Eighty-eight cubic metres of card-index," said Mr. Foster with relish,
as they entered."""
diff --git a/hubconf.py b/hubconf.py
index f8d0d1a8..4e5c1b4b 100644
--- a/hubconf.py
+++ b/hubconf.py
@@ -14,7 +14,7 @@ dependencies = ["torch", "tqdm", "boto3", "requests", "regex", "sentencepiece",
@add_start_docstrings(AutoConfig.__doc__)
def config(*args, **kwargs):
- r"""
+ r"""
# Using torch.hub !
import torch
@@ -34,7 +34,7 @@ def config(*args, **kwargs):
@add_start_docstrings(AutoTokenizer.__doc__)
def tokenizer(*args, **kwargs):
- r"""
+ r"""
# Using torch.hub !
import torch
diff --git a/templates/adding_a_new_model/modeling_tf_xxx.py b/templates/adding_a_new_model/modeling_tf_xxx.py
index 3e8f51bf..d5f6c1e7 100644
--- a/templates/adding_a_new_model/modeling_tf_xxx.py
+++ b/templates/adding_a_new_model/modeling_tf_xxx.py
@@ -216,7 +216,7 @@ XXX_START_DOCSTRING = r""" The XXX model was proposed in
`model({'input_ids': input_ids, 'token_type_ids': token_type_ids})`
Parameters:
- config (:class:`~transformers.XxxConfig`): Model configuration class with all the parameters of the model.
+ config (:class:`~transformers.XxxConfig`): Model configuration class with all the parameters of the model.
Initializing with a config file does not load the weights associated with the model, only the configuration.
Check out the :meth:`~transformers.PreTrainedModel.from_pretrained` method to load the model weights.
"""
@@ -230,13 +230,13 @@ XXX_INPUTS_DOCSTRING = r"""
(a) For sequence pairs:
``tokens: [CLS] is this jack ##son ##ville ? [SEP] no it is not . [SEP]``
-
+
``token_type_ids: 0 0 0 0 0 0 0 0 1 1 1 1 1 1``
(b) For single sequences:
``tokens: [CLS] the dog is hairy . [SEP]``
-
+
``token_type_ids: 0 0 0 0 0 0 0``
Xxx is a model with absolute position embeddings so it's usually advised to pad the inputs on
diff --git a/templates/adding_a_new_model/modeling_xxx.py b/templates/adding_a_new_model/modeling_xxx.py
index 4ea3cca8..e4e72c97 100644
--- a/templates/adding_a_new_model/modeling_xxx.py
+++ b/templates/adding_a_new_model/modeling_xxx.py
@@ -198,7 +198,7 @@ XXX_START_DOCSTRING = r""" The XXX model was proposed in
https://pytorch.org/docs/stable/nn.html#module
Parameters:
- config (:class:`~transformers.XxxConfig`): Model configuration class with all the parameters of the model.
+ config (:class:`~transformers.XxxConfig`): Model configuration class with all the parameters of the model.
Initializing with a config file does not load the weights associated with the model, only the configuration.
Check out the :meth:`~transformers.PreTrainedModel.from_pretrained` method to load the model weights.
"""
@@ -212,13 +212,13 @@ XXX_INPUTS_DOCSTRING = r"""
(a) For sequence pairs:
``tokens: [CLS] is this jack ##son ##ville ? [SEP] no it is not . [SEP]``
-
+
``token_type_ids: 0 0 0 0 0 0 0 0 1 1 1 1 1 1``
(b) For single sequences:
``tokens: [CLS] the dog is hairy . [SEP]``
-
+
``token_type_ids: 0 0 0 0 0 0 0``
Xxx is a model with absolute position embeddings so it's usually advised to pad the inputs on
@@ -670,9 +670,9 @@ class XxxForQuestionAnswering(XxxPreTrainedModel):
question, text = "Who was Jim Henson?", "Jim Henson was a nice puppet"
input_text = "[CLS] " + question + " [SEP] " + text + " [SEP]"
input_ids = tokenizer.encode(input_text)
- token_type_ids = [0 if i <= input_ids.index(102) else 1 for i in range(len(input_ids))]
+ token_type_ids = [0 if i <= input_ids.index(102) else 1 for i in range(len(input_ids))]
start_scores, end_scores = model(torch.tensor([input_ids]), token_type_ids=torch.tensor([token_type_ids]))
- all_tokens = tokenizer.convert_ids_to_tokens(input_ids)
+ all_tokens = tokenizer.convert_ids_to_tokens(input_ids)
print(' '.join(all_tokens[torch.argmax(start_scores) : torch.argmax(end_scores)+1]))
# a nice puppet
diff --git a/transformers/commands/user.py b/transformers/commands/user.py
index c6edda28..6800920c 100644
--- a/transformers/commands/user.py
+++ b/transformers/commands/user.py
@@ -49,11 +49,11 @@ class LoginCommand(BaseUserCommand):
def run(self):
print(
"""
- _| _| _| _| _|_|_| _|_|_| _|_|_| _| _| _|_|_| _|_|_|_| _|_| _|_|_| _|_|_|_|
- _| _| _| _| _| _| _| _|_| _| _| _| _| _| _| _|
- _|_|_|_| _| _| _| _|_| _| _|_| _| _| _| _| _| _|_| _|_|_| _|_|_|_| _| _|_|_|
- _| _| _| _| _| _| _| _| _| _| _|_| _| _| _| _| _| _| _|
- _| _| _|_| _|_|_| _|_|_| _|_|_| _| _| _|_|_| _| _| _| _|_|_| _|_|_|_|
+ _| _| _| _| _|_|_| _|_|_| _|_|_| _| _| _|_|_| _|_|_|_| _|_| _|_|_| _|_|_|_|
+ _| _| _| _| _| _| _| _|_| _| _| _| _| _| _| _|
+ _|_|_|_| _| _| _| _|_| _| _|_| _| _| _| _| _| _|_| _|_|_| _|_|_|_| _| _|_|_|
+ _| _| _| _| _| _| _| _| _| _| _|_| _| _| _| _| _| _| _|
+ _| _| _|_| _|_|_| _|_|_| _|_|_| _| _| _|_|_| _| _| _| _|_|_| _|_|_|_|
"""
)
diff --git a/transformers/data/processors/squad.py b/transformers/data/processors/squad.py
index d47211a0..c6581374 100644
--- a/transformers/data/processors/squad.py
+++ b/transformers/data/processors/squad.py
@@ -281,7 +281,7 @@ def squad_convert_examples_to_features(
processor = SquadV2Processor()
examples = processor.get_dev_examples(data_dir)
- features = squad_convert_examples_to_features(
+ features = squad_convert_examples_to_features(
examples=examples,
tokenizer=tokenizer,
max_seq_length=args.max_seq_length,
@@ -640,8 +640,8 @@ class SquadFeatures(object):
has more information related to that token and should be prioritized over this feature for that token.
tokens: list of tokens corresponding to the input ids
token_to_orig_map: mapping between the tokens and the original text, needed in order to identify the answer.
- start_position: start of the answer token index
- end_position: end of the answer token index
+ start_position: start of the answer token index
+ end_position: end of the answer token index
"""
def __init__(
diff --git a/transformers/modeling_albert.py b/transformers/modeling_albert.py
index 7ff20008..669d8fc3 100644
--- a/transformers/modeling_albert.py
+++ b/transformers/modeling_albert.py
@@ -396,7 +396,7 @@ ALBERT_START_DOCSTRING = r""" The ALBERT model was proposed in
https://pytorch.org/docs/stable/nn.html#module
Parameters:
- config (:class:`~transformers.AlbertConfig`): Model configuration class with all the parameters of the model.
+ config (:class:`~transformers.AlbertConfig`): Model configuration class with all the parameters of the model.
Initializing with a config file does not load the weights associated with the model, only the configuration.
Check out the :meth:`~transformers.PreTrainedModel.from_pretrained` method to load the model weights.
"""
@@ -410,13 +410,13 @@ ALBERT_INPUTS_DOCSTRING = r"""
(a) For sequence pairs:
``tokens: [CLS] is this jack ##son ##ville ? [SEP] no it is not . [SEP]``
-
+
``token_type_ids: 0 0 0 0 0 0 0 0 1 1 1 1 1 1``
(b) For single sequences:
``tokens: [CLS] the dog is hairy . [SEP]``
-
+
``token_type_ids: 0 0 0 0 0 0 0``
Albert is a model with absolute position embeddings so it's usually advised to pad the inputs on
@@ -796,9 +796,9 @@ class AlbertForQuestionAnswering(AlbertPreTrainedModel):
question, text = "Who was Jim Henson?", "Jim Henson was a nice puppet"
input_text = "[CLS] " + question + " [SEP] " + text + " [SEP]"
input_ids = tokenizer.encode(input_text)
- token_type_ids = [0 if i <= input_ids.index(102) else 1 for i in range(len(input_ids))]
+ token_type_ids = [0 if i <= input_ids.index(102) else 1 for i in range(len(input_ids))]
start_scores, end_scores = model(torch.tensor([input_ids]), token_type_ids=torch.tensor([token_type_ids]))
- all_tokens = tokenizer.convert_ids_to_tokens(input_ids)
+ all_tokens = tokenizer.convert_ids_to_tokens(input_ids)
print(' '.join(all_tokens[torch.argmax(start_scores) : torch.argmax(end_scores)+1]))
# a nice puppet
diff --git a/transformers/modeling_auto.py b/transformers/modeling_auto.py
index bcdde45b..2df74130 100644
--- a/transformers/modeling_auto.py
+++ b/transformers/modeling_auto.py
@@ -864,7 +864,7 @@ class AutoModelForTokenClassification:
def from_config(cls, config):
r""" Instantiates one of the base model classes of the library
from a configuration.
-
+
config: (`optional`) instance of a class derived from :class:`~transformers.PretrainedConfig`:
The model class to instantiate is selected based on the configuration class:
- isInstance of `distilbert` configuration class: DistilBertModel (DistilBERT model)
@@ -874,7 +874,7 @@ class AutoModelForTokenClassification:
- isInstance of `roberta` configuration class: RobertaModel (Roberta model)
Examples::
-
+
config = BertConfig.from_pretrained('bert-base-uncased') # Download configuration from S3 and cache.
model = AutoModelForTokenClassification.from_config(config) # E.g. model was saved using `save_pretrained('./test/saved_model/')`
"""
diff --git a/transformers/modeling_camembert.py b/transformers/modeling_camembert.py
index 363399ee..d38751a1 100644
--- a/transformers/modeling_camembert.py
+++ b/transformers/modeling_camembert.py
@@ -40,9 +40,9 @@ CAMEMBERT_PRETRAINED_MODEL_ARCHIVE_MAP = {
CAMEMBERT_START_DOCSTRING = r""" The CamemBERT model was proposed in
`CamemBERT: a Tasty French Language Model`_
by Louis Martin, Benjamin Muller, Pedro Javier Ortiz Suárez, Yoann Dupont, Laurent Romary, Éric Villemonte de la Clergerie, Djamé Seddah, and Benoît Sagot. It is based on Facebook's RoBERTa model released in 2019.
-
+
It is a model trained on 138GB of French text.
-
+
This implementation is the same as RoBERTa.
This model is a PyTorch `torch.nn.Module`_ sub-class. Use it as a regular PyTorch Module and
@@ -55,7 +55,7 @@ CAMEMBERT_START_DOCSTRING = r""" The CamemBERT model was proposed in
https://pytorch.org/docs/stable/nn.html#module
Parameters:
- config (:class:`~transformers.CamembertConfig`): Model configuration class with all the parameters of the
+ config (:class:`~transformers.CamembertConfig`): Model configuration class with all the parameters of the
model. Initializing with a config file does not load the weights associated with the model, only the configuration.
Check out the :meth:`~transformers.PreTrainedModel.from_pretrained` method to load the model weights.
"""
@@ -74,7 +74,7 @@ CAMEMBERT_INPUTS_DOCSTRING = r"""
``tokens: the dog is hairy . ``
- Fully encoded sequences or sequence pairs can be obtained using the CamembertTokenizer.encode function with
+ Fully encoded sequences or sequence pairs can be obtained using the CamembertTokenizer.encode function with
the ``add_special_tokens`` parameter set to ``True``.
CamemBERT is a model with absolute position embeddings so it's usually advised to pad the inputs on
@@ -199,7 +199,7 @@ class CamembertForMaskedLM(RobertaForMaskedLM):
@add_start_docstrings(
- """CamemBERT Model transformer with a sequence classification/regression head on top (a linear layer
+ """CamemBERT Model transformer with a sequence classification/regression head on top (a linear layer
on top of the pooled output) e.g. for GLUE tasks. """,
CAMEMBERT_START_DOCSTRING,
CAMEMBERT_INPUTS_DOCSTRING,
diff --git a/transformers/modeling_ctrl.py b/transformers/modeling_ctrl.py
index f3418917..193a3d7d 100644
--- a/transformers/modeling_ctrl.py
+++ b/transformers/modeling_ctrl.py
@@ -192,7 +192,7 @@ class CTRLPreTrainedModel(PreTrainedModel):
module.weight.data.fill_(1.0)
-CTRL_START_DOCSTRING = r""" CTRL model was proposed in
+CTRL_START_DOCSTRING = r""" CTRL model was proposed in
`CTRL: A Conditional Transformer Language Model for Controllable Generation`_
by Nitish Shirish Keskar*, Bryan McCann*, Lav R. Varshney, Caiming Xiong and Richard Socher.
It's a causal (unidirectional) transformer pre-trained using language modeling on a very large
@@ -224,7 +224,7 @@ CTRL_INPUTS_DOCSTRING = r""" Inputs:
**past**:
list of ``torch.FloatTensor`` (one for each layer):
that contains pre-computed hidden-states (key and values in the attention blocks) as computed by the model
- (see `past` output below). Can be used to speed up sequential decoding. The token ids which have their past given to this model
+ (see `past` output below). Can be used to speed up sequential decoding. The token ids which have their past given to this model
should not be passed as input ids as they have already been computed.
**attention_mask**: (`optional`) ``torch.FloatTensor`` of shape ``(batch_size, sequence_length)``:
Mask to avoid performing attention on padding token indices.
@@ -261,7 +261,7 @@ class CTRLModel(CTRLPreTrainedModel):
**past**:
list of ``torch.FloatTensor`` (one for each layer) of shape ``(2, batch_size, num_heads, sequence_length, embed_size_per_head)``:
that contains pre-computed hidden-states (key and values in the attention blocks).
- Can be used (see `past` input) to speed up sequential decoding. The token ids which have their past given to this model
+ Can be used (see `past` input) to speed up sequential decoding. The token ids which have their past given to this model
should not be passed as input ids as they have already been computed.
**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)
@@ -464,7 +464,7 @@ class CTRLLMHeadModel(CTRLPreTrainedModel):
**past**:
list of ``torch.FloatTensor`` (one for each layer) of shape ``(2, batch_size, num_heads, sequence_length, embed_size_per_head)``:
that contains pre-computed hidden-states (key and values in the attention blocks).
- Can be used (see `past` input) to speed up sequential decoding. The token ids which have their past given to this model
+ Can be used (see `past` input) to speed up sequential decoding. The token ids which have their past given to this model
should not be passed as input ids as they have already been computed.
**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)
diff --git a/transformers/modeling_distilbert.py b/transformers/modeling_distilbert.py
index 5fef4438..aa732b31 100644
--- a/transformers/modeling_distilbert.py
+++ b/transformers/modeling_distilbert.py
@@ -366,12 +366,12 @@ DISTILBERT_START_DOCSTRING = r"""
For more information on DistilBERT, please refer to our
`detailed blog post`_
-
+
.. _`detailed blog post`:
https://medium.com/huggingface/distilbert-8cf3380435b5
Parameters:
- config (:class:`~transformers.DistilBertConfig`): Model configuration class with all the parameters of the model.
+ config (:class:`~transformers.DistilBertConfig`): Model configuration class with all the parameters of the model.
Initializing with a config file does not load the weights associated with the model, only the configuration.
Check out the :meth:`~transformers.PreTrainedModel.from_pretrained` method to load the model weights.
"""
@@ -381,7 +381,7 @@ DISTILBERT_INPUTS_DOCSTRING = r"""
**input_ids** ``torch.LongTensor`` of shape ``(batch_size, sequence_length)``:
Indices of input sequence tokens in the vocabulary.
The input sequences should start with `[CLS]` and end with `[SEP]` tokens.
-
+
For now, ONLY BertTokenizer(`bert-base-uncased`) is supported and you should use this tokenizer when using DistilBERT.
**attention_mask**: (`optional`) ``torch.LongTensor`` of shape ``(batch_size, sequence_length)``:
Mask to avoid performing attention on padding token indices.
diff --git a/transformers/modeling_gpt2.py b/transformers/modeling_gpt2.py
index 94e977e6..1e5c54c9 100644
--- a/transformers/modeling_gpt2.py
+++ b/transformers/modeling_gpt2.py
@@ -304,7 +304,7 @@ GPT2_INPUTS_DOCSTRING = r""" Inputs:
**past**:
list of ``torch.FloatTensor`` (one for each layer):
that contains pre-computed hidden-states (key and values in the attention blocks) as computed by the model
- (see `past` output below). Can be used to speed up sequential decoding. The token ids which have their past given to this model
+ (see `past` output below). Can be used to speed up sequential decoding. The token ids which have their past given to this model
should not be passed as input ids as they have already been computed.
**attention_mask**: (`optional`) ``torch.FloatTensor`` of shape ``(batch_size, sequence_length)``:
Mask to avoid performing attention on padding token indices.
@@ -341,7 +341,7 @@ class GPT2Model(GPT2PreTrainedModel):
**past**:
list of ``torch.FloatTensor`` (one for each layer) of shape ``(2, batch_size, num_heads, sequence_length, embed_size_per_head)``:
that contains pre-computed hidden-states (key and values in the attention blocks).
- Can be used (see `past` input) to speed up sequential decoding. The token ids which have their past given to this model
+ Can be used (see `past` input) to speed up sequential decoding. The token ids which have their past given to this model
should not be passed as input ids as they have already been computed.
**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)
@@ -532,7 +532,7 @@ class GPT2LMHeadModel(GPT2PreTrainedModel):
**past**:
list of ``torch.FloatTensor`` (one for each layer) of shape ``(2, batch_size, num_heads, sequence_length, embed_size_per_head)``:
that contains pre-computed hidden-states (key and values in the attention blocks).
- Can be used (see `past` input) to speed up sequential decoding. The token ids which have their past given to this model
+ Can be used (see `past` input) to speed up sequential decoding. The token ids which have their past given to this model
should not be passed as input ids as they have already been computed.
**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)
@@ -640,7 +640,7 @@ class GPT2DoubleHeadsModel(GPT2PreTrainedModel):
**past**:
list of ``torch.FloatTensor`` (one for each layer) of shape ``(2, batch_size, num_heads, sequence_length, embed_size_per_head)``:
that contains pre-computed hidden-states (key and values in the attention blocks).
- Can be used (see `past` input) to speed up sequential decoding. The token ids which have their past given to this model
+ Can be used (see `past` input) to speed up sequential decoding. The token ids which have their past given to this model
should not be passed as input ids as they have already been computed.
**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)
@@ -654,15 +654,15 @@ class GPT2DoubleHeadsModel(GPT2PreTrainedModel):
import torch
from transformers import GPT2Tokenizer, GPT2DoubleHeadsModel
-
+
tokenizer = GPT2Tokenizer.from_pretrained('gpt2')
model = GPT2DoubleHeadsModel.from_pretrained('gpt2')
-
+
# Add a [CLS] to the vocabulary (we should train it also!)
tokenizer.add_special_tokens({'cls_token': '[CLS]'})
model.resize_token_embeddings(len(tokenizer)) # Update the model embeddings with the new vocabulary size
print(tokenizer.cls_token_id, len(tokenizer)) # The newly token the last token of the vocabulary
-
+
choices = ["Hello, my dog is cute [CLS]", "Hello, my cat is cute [CLS]"]
encoded_choices = [tokenizer.encode(s) for s in choices]
cls_token_location = [tokens.index(tokenizer.cls_token_id) for tokens in encoded_choices]
diff --git a/transformers/modeling_mmbt.py b/transformers/modeling_mmbt.py
index 2c22a409..490969fc 100644
--- a/transformers/modeling_mmbt.py
+++ b/transformers/modeling_mmbt.py
@@ -75,10 +75,10 @@ class ModalEmbeddings(nn.Module):
return embeddings
-MMBT_START_DOCSTRING = r""" MMBT model was proposed in
+MMBT_START_DOCSTRING = r""" MMBT model was proposed in
`Supervised Multimodal Bitransformers for Classifying Images and Text`_
by Douwe Kiela, Suvrat Bhooshan, Hamed Firooz, Davide Testuggine.
- It's a supervised multimodal bitransformer model that fuses information from text and other image encoders,
+ It's a supervised multimodal bitransformer model that fuses information from text and other image encoders,
and obtain state-of-the-art performance on various multimodal classification benchmark tasks.
This model is a PyTorch `torch.nn.Module`_ sub-class. Use it as a regular PyTorch Module and
@@ -93,15 +93,15 @@ MMBT_START_DOCSTRING = r""" MMBT model was proposed in
Parameters:
config (:class:`~transformers.MMBTConfig`): Model configuration class with all the parameters of the model.
Initializing with a config file does not load the weights associated with the model, only the configuration.
- transformer (:class: `~nn.Module`): A text transformer that is used by MMBT.
+ transformer (:class: `~nn.Module`): A text transformer that is used by MMBT.
It should have embeddings, encoder, and pooler attributes.
- encoder (:class: `~nn.Module`): Encoder for the second modality.
+ encoder (:class: `~nn.Module`): Encoder for the second modality.
It should take in a batch of modal inputs and return k, n dimension embeddings.
"""
MMBT_INPUTS_DOCSTRING = r""" Inputs:
**input_modal**: ``torch.FloatTensor`` of shape ``(batch_size, ***)``:
- The other modality data. It will be the shape that the encoder for that type expects.
+ The other modality data. It will be the shape that the encoder for that type expects.
e.g. With an Image Encoder, the shape would be (batch_size, channels, height, width)
**input_ids**: ``torch.LongTensor`` of shape ``(batch_size, sequence_length)``:
Indices of input sequence tokens in the vocabulary.
@@ -119,7 +119,7 @@ MMBT_INPUTS_DOCSTRING = r""" Inputs:
**token_type_ids**: (`optional`) ``torch.LongTensor`` of shape ``(batch_size, sequence_length)``:
Segment token indices to indicate different portions of the inputs.
**modal_token_type_ids**: (`optional`) ``torch.LongTensor`` of shape ``(batch_size, modal_sequence_length)``:
- Segment token indices to indicate different portions of the non-text modality.
+ Segment token indices to indicate different portions of the non-text modality.
The embeddings from these tokens will be summed with the respective token embeddings for the non-text modality.
**position_ids**: (`optional`) ``torch.LongTensor`` of shape ``(batch_size, sequence_length)``:
Indices of positions of each input sequence tokens in the position embeddings.
diff --git a/transformers/modeling_roberta.py b/transformers/modeling_roberta.py
index f6233061..d4cad0d0 100644
--- a/transformers/modeling_roberta.py
+++ b/transformers/modeling_roberta.py
@@ -97,11 +97,11 @@ ROBERTA_START_DOCSTRING = r""" The RoBERTa model was proposed in
`RoBERTa: A Robustly Optimized BERT Pretraining Approach`_
by Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer,
Veselin Stoyanov. It is based on Google's BERT model released in 2018.
-
+
It builds on BERT and modifies key hyperparameters, removing the next-sentence pretraining
objective and training with much larger mini-batches and learning rates.
-
- This implementation is the same as BertModel with a tiny embeddings tweak as well as a setup for Roberta pretrained
+
+ This implementation is the same as BertModel with a tiny embeddings tweak as well as a setup for Roberta pretrained
models.
This model is a PyTorch `torch.nn.Module`_ sub-class. Use it as a regular PyTorch Module and
@@ -114,7 +114,7 @@ ROBERTA_START_DOCSTRING = r""" The RoBERTa model was proposed in
https://pytorch.org/docs/stable/nn.html#module
Parameters:
- config (:class:`~transformers.RobertaConfig`): Model configuration class with all the parameters of the
+ config (:class:`~transformers.RobertaConfig`): Model configuration class with all the parameters of the
model. Initializing with a config file does not load the weights associated with the model, only the configuration.
Check out the :meth:`~transformers.PreTrainedModel.from_pretrained` method to load the model weights.
"""
@@ -133,7 +133,7 @@ ROBERTA_INPUTS_DOCSTRING = r"""
``tokens: the dog is hairy . ``
- Fully encoded sequences or sequence pairs can be obtained using the RobertaTokenizer.encode function with
+ Fully encoded sequences or sequence pairs can be obtained using the RobertaTokenizer.encode function with
the ``add_special_tokens`` parameter set to ``True``.
RoBERTa is a model with absolute position embeddings so it's usually advised to pad the inputs on
@@ -319,7 +319,7 @@ class RobertaLMHead(nn.Module):
@add_start_docstrings(
- """RoBERTa Model transformer with a sequence classification/regression head on top (a linear layer
+ """RoBERTa Model transformer with a sequence classification/regression head on top (a linear layer
on top of the pooled output) e.g. for GLUE tasks. """,
ROBERTA_START_DOCSTRING,
ROBERTA_INPUTS_DOCSTRING,
diff --git a/transformers/modeling_t5.py b/transformers/modeling_t5.py
index 1467a0cd..b67f0122 100644
--- a/transformers/modeling_t5.py
+++ b/transformers/modeling_t5.py
@@ -661,7 +661,7 @@ T5_START_DOCSTRING = r""" The T5 model was proposed in
https://pytorch.org/docs/stable/nn.html#module
Parameters:
- config (:class:`~transformers.T5Config`): Model configuration class with all the parameters of the model.
+ config (:class:`~transformers.T5Config`): Model configuration class with all the parameters of the model.
Initializing with a config file does not load the weights associated with the model, only the configuration.
Check out the :meth:`~transformers.PreTrainedModel.from_pretrained` method to load the model weights.
"""
diff --git a/transformers/modeling_tf_albert.py b/transformers/modeling_tf_albert.py
index de6ef405..e921dc8c 100644
--- a/transformers/modeling_tf_albert.py
+++ b/transformers/modeling_tf_albert.py
@@ -510,7 +510,7 @@ ALBERT_START_DOCSTRING = r""" The ALBERT model was proposed in
`model({'input_ids': input_ids, 'token_type_ids': token_type_ids})`
Parameters:
- config (:class:`~transformers.AlbertConfig`): Model configuration class with all the parameters of the model.
+ config (:class:`~transformers.AlbertConfig`): Model configuration class with all the parameters of the model.
Initializing with a config file does not load the weights associated with the model, only the configuration.
Check out the :meth:`~transformers.PreTrainedModel.from_pretrained` method to load the model weights.
"""
@@ -524,13 +524,13 @@ ALBERT_INPUTS_DOCSTRING = r"""
(a) For sequence pairs:
``tokens: [CLS] is this jack ##son ##ville ? [SEP] no it is not . [SEP]``
-
+
``token_type_ids: 0 0 0 0 0 0 0 0 1 1 1 1 1 1``
(b) For single sequences:
``tokens: [CLS] the dog is hairy . [SEP]``
-
+
``token_type_ids: 0 0 0 0 0 0 0``
Albert is a model with absolute position embeddings so it's usually advised to pad the inputs on
diff --git a/transformers/modeling_tf_ctrl.py b/transformers/modeling_tf_ctrl.py
index a3a22040..6f878c54 100644
--- a/transformers/modeling_tf_ctrl.py
+++ b/transformers/modeling_tf_ctrl.py
@@ -356,7 +356,7 @@ class TFCTRLPreTrainedModel(TFPreTrainedModel):
base_model_prefix = "transformer"
-CTRL_START_DOCSTRING = r""" CTRL model was proposed in
+CTRL_START_DOCSTRING = r""" CTRL model was proposed in
`CTRL: A Conditional Transformer Language Model for Controllable Generation`_
by Nitish Shirish Keskar*, Bryan McCann*, Lav R. Varshney, Caiming Xiong and Richard Socher.
It's a causal (unidirectional) transformer pre-trained using language modeling on a very large
diff --git a/transformers/modeling_tf_distilbert.py b/transformers/modeling_tf_distilbert.py
index 98317488..297d7edb 100644
--- a/transformers/modeling_tf_distilbert.py
+++ b/transformers/modeling_tf_distilbert.py
@@ -109,7 +109,7 @@ class TFEmbeddings(tf.keras.layers.Layer):
linear tensor, float32 with shape [batch_size, length, vocab_size].
Raises:
ValueError: if mode is not valid.
-
+
Shared weights logic adapted from
https://github.com/tensorflow/models/blob/a009f4fb9d2fc4949e32192a944688925ef78659/official/transformer/v2/embedding_layer.py#L24
"""
@@ -487,7 +487,7 @@ DISTILBERT_START_DOCSTRING = r"""
For more information on DistilBERT, please refer to our
`detailed blog post`_
-
+
This model is a tf.keras.Model `tf.keras.Model`_ sub-class. Use it as a regular TF 2.0 Keras Model and
refer to the TF 2.0 documentation for all matter related to general usage and behavior.
@@ -514,7 +514,7 @@ DISTILBERT_START_DOCSTRING = r"""
`model({'input_ids': input_ids, 'token_type_ids': token_type_ids})`
Parameters:
- config (:class:`~transformers.DistilBertConfig`): Model configuration class with all the parameters of the model.
+ config (:class:`~transformers.DistilBertConfig`): Model configuration class with all the parameters of the model.
Initializing with a config file does not load the weights associated with the model, only the configuration.
Check out the :meth:`~transformers.PreTrainedModel.from_pretrained` method to load the model weights.
"""
@@ -524,7 +524,7 @@ DISTILBERT_INPUTS_DOCSTRING = r"""
**input_ids** ``Numpy array`` or ``tf.Tensor`` of shape ``(batch_size, sequence_length)``:
Indices of input sequence tokens in the vocabulary.
The input sequences should start with `[CLS]` and end with `[SEP]` tokens.
-
+
For now, ONLY BertTokenizer(`bert-base-uncased`) is supported and you should use this tokenizer when using DistilBERT.
**attention_mask**: (`optional`) ``Numpy array`` or ``tf.Tensor`` of shape ``(batch_size, sequence_length)``:
Mask to avoid performing attention on padding token indices.
diff --git a/transformers/modeling_tf_gpt2.py b/transformers/modeling_tf_gpt2.py
index bf551991..9ad049e9 100644
--- a/transformers/modeling_tf_gpt2.py
+++ b/transformers/modeling_tf_gpt2.py
@@ -584,14 +584,14 @@ class TFGPT2DoubleHeadsModel(TFGPT2PreTrainedModel):
tokenizer = GPT2Tokenizer.from_pretrained('gpt2')
model = TFGPT2DoubleHeadsModel.from_pretrained('gpt2')
-
+
# Add a [CLS] to the vocabulary (we should train it also!)
# This option is currently not implemented in TF 2.0
raise NotImplementedError
tokenizer.add_special_tokens({'cls_token': '[CLS]'})
model.resize_token_embeddings(len(tokenizer)) # Update the model embeddings with the new vocabulary size
print(tokenizer.cls_token_id, len(tokenizer)) # The newly token the last token of the vocabulary
-
+
choices = ["Hello, my dog is cute [CLS]", "Hello, my cat is cute [CLS]"]
encoded_choices = [tokenizer.encode(s) for s in choices]
cls_token_location = [tokens.index(tokenizer.cls_token_id) for tokens in encoded_choices]
diff --git a/transformers/modeling_tf_openai.py b/transformers/modeling_tf_openai.py
index 44924b4f..1fd2e961 100644
--- a/transformers/modeling_tf_openai.py
+++ b/transformers/modeling_tf_openai.py
@@ -553,7 +553,7 @@ class TFOpenAIGPTDoubleHeadsModel(TFOpenAIGPTPreTrainedModel):
tokenizer = OpenAIGPTTokenizer.from_pretrained('openai-gpt')
model = TFOpenAIGPTDoubleHeadsModel.from_pretrained('openai-gpt')
-
+
# Add a [CLS] to the vocabulary (we should train it also!)
# This option is currently not implemented in TF 2.0
raise NotImplementedError
diff --git a/transformers/modeling_tf_roberta.py b/transformers/modeling_tf_roberta.py
index 5c40682c..7aca11f5 100644
--- a/transformers/modeling_tf_roberta.py
+++ b/transformers/modeling_tf_roberta.py
@@ -111,11 +111,11 @@ ROBERTA_START_DOCSTRING = r""" The RoBERTa model was proposed in
`RoBERTa: A Robustly Optimized BERT Pretraining Approach`_
by Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer,
Veselin Stoyanov. It is based on Google's BERT model released in 2018.
-
+
It builds on BERT and modifies key hyperparameters, removing the next-sentence pretraining
objective and training with much larger mini-batches and learning rates.
-
- This implementation is the same as BertModel with a tiny embeddings tweak as well as a setup for Roberta pretrained
+
+ This implementation is the same as BertModel with a tiny embeddings tweak as well as a setup for Roberta pretrained
models.
This model is a tf.keras.Model `tf.keras.Model`_ sub-class. Use it as a regular TF 2.0 Keras Model and
@@ -144,7 +144,7 @@ ROBERTA_START_DOCSTRING = r""" The RoBERTa model was proposed in
`model({'input_ids': input_ids, 'token_type_ids': token_type_ids})`
Parameters:
- config (:class:`~transformers.RobertaConfig`): Model configuration class with all the parameters of the
+ config (:class:`~transformers.RobertaConfig`): Model configuration class with all the parameters of the
model. Initializing with a config file does not load the weights associated with the model, only the configuration.
Check out the :meth:`~transformers.PreTrainedModel.from_pretrained` method to load the model weights.
"""
@@ -163,7 +163,7 @@ ROBERTA_INPUTS_DOCSTRING = r"""
``tokens: the dog is hairy . ``
- Fully encoded sequences or sequence pairs can be obtained using the RobertaTokenizer.encode function with
+ Fully encoded sequences or sequence pairs can be obtained using the RobertaTokenizer.encode function with
the ``add_special_tokens`` parameter set to ``True``.
RoBERTa is a model with absolute position embeddings so it's usually advised to pad the inputs on
@@ -351,7 +351,7 @@ class TFRobertaClassificationHead(tf.keras.layers.Layer):
@add_start_docstrings(
- """RoBERTa Model transformer with a sequence classification/regression head on top (a linear layer
+ """RoBERTa Model transformer with a sequence classification/regression head on top (a linear layer
on top of the pooled output) e.g. for GLUE tasks. """,
ROBERTA_START_DOCSTRING,
ROBERTA_INPUTS_DOCSTRING,
diff --git a/transformers/modeling_tf_t5.py b/transformers/modeling_tf_t5.py
index e995bc3c..5ab16ea4 100644
--- a/transformers/modeling_tf_t5.py
+++ b/transformers/modeling_tf_t5.py
@@ -565,7 +565,7 @@ T5_START_DOCSTRING = r""" The T5 model was proposed in
`model({'input_ids': input_ids, 'token_type_ids': token_type_ids})`
Parameters:
- config (:class:`~transformers.T5Config`): Model configuration class with all the parameters of the model.
+ config (:class:`~transformers.T5Config`): Model configuration class with all the parameters of the model.
Initializing with a config file does not load the weights associated with the model, only the configuration.
Check out the :meth:`~transformers.PreTrainedModel.from_pretrained` method to load the model weights.
"""
diff --git a/transformers/modeling_tf_utils.py b/transformers/modeling_tf_utils.py
index 637013b3..69ceaaf5 100644
--- a/transformers/modeling_tf_utils.py
+++ b/transformers/modeling_tf_utils.py
@@ -139,7 +139,7 @@ class TFPreTrainedModel(tf.keras.Model):
Arguments:
new_num_tokens: (`optional`) int:
- New number of tokens in the embedding matrix. Increasing the size will add newly initialized vectors at the end. Reducing the size will remove vectors from the end.
+ New number of tokens in the embedding matrix. Increasing the size will add newly initialized vectors at the end. Reducing the size will remove vectors from the end.
If not provided or None: does nothing and just returns a pointer to the input tokens ``tf.Variable`` Module of the model.
Return: ``tf.Variable``
@@ -431,7 +431,7 @@ class TFSharedEmbeddings(tf.keras.layers.Layer):
linear tensor, float32 with shape [batch_size, length, vocab_size].
Raises:
ValueError: if mode is not valid.
-
+
Shared weights logic adapted from
https://github.com/tensorflow/models/blob/a009f4fb9d2fc4949e32192a944688925ef78659/official/transformer/v2/embedding_layer.py#L24
"""
diff --git a/transformers/modeling_xlm.py b/transformers/modeling_xlm.py
index 2127bbad..ac45e3e2 100644
--- a/transformers/modeling_xlm.py
+++ b/transformers/modeling_xlm.py
@@ -825,7 +825,7 @@ class XLMForQuestionAnsweringSimple(XLMPreTrainedModel):
**cls_index**: (`optional`) ``torch.LongTensor`` of shape ``(batch_size,)``:
Labels for position (index) of the classification token to use as input for computing plausibility of the answer.
**p_mask**: (`optional`) ``torch.LongTensor`` of shape ``(batch_size, sequence_length)``:
- Optional mask of tokens which can't be in answers (e.g. [CLS], [PAD], ...)
+ Optional mask of tokens which can't be in answers (e.g. [CLS], [PAD], ...)
Outputs: `Tuple` comprising various elements depending on the configuration (config) and inputs:
**loss**: (`optional`, returned when ``labels`` is provided) ``torch.FloatTensor`` of shape ``(1,)``:
@@ -942,7 +942,7 @@ class XLMForQuestionAnswering(XLMPreTrainedModel):
**cls_index**: (`optional`) ``torch.LongTensor`` of shape ``(batch_size,)``:
Labels for position (index) of the classification token to use as input for computing plausibility of the answer.
**p_mask**: (`optional`) ``torch.LongTensor`` of shape ``(batch_size, sequence_length)``:
- Optional mask of tokens which can't be in answers (e.g. [CLS], [PAD], ...)
+ Optional mask of tokens which can't be in answers (e.g. [CLS], [PAD], ...)
Outputs: `Tuple` comprising various elements depending on the configuration (config) and inputs:
**loss**: (`optional`, returned when ``labels`` is provided) ``torch.FloatTensor`` of shape ``(1,)``:
diff --git a/transformers/modeling_xlm_roberta.py b/transformers/modeling_xlm_roberta.py
index adf7f233..f20e8e30 100644
--- a/transformers/modeling_xlm_roberta.py
+++ b/transformers/modeling_xlm_roberta.py
@@ -45,7 +45,7 @@ XLM_ROBERTA_PRETRAINED_MODEL_ARCHIVE_MAP = {
XLM_ROBERTA_START_DOCSTRING = r""" The XLM-RoBERTa model was proposed in
`Unsupervised Cross-lingual Representation Learning at Scale`_
by Alexis Conneau, Kartikay Khandelwal, Naman Goyal, Vishrav Chaudhary, Guillaume Wenzek, Francisco Guzmán, Edouard Grave, Myle Ott, Luke Zettlemoyer and Veselin Stoyanov. It is based on Facebook's RoBERTa model released in 2019.
-
+
It is a large multi-lingual language model, trained on 2.5TB of filtered CommonCrawl data.
This implementation is the same as RoBERTa.
@@ -60,7 +60,7 @@ XLM_ROBERTA_START_DOCSTRING = r""" The XLM-RoBERTa model was proposed in
https://pytorch.org/docs/stable/nn.html#module
Parameters:
- config (:class:`~transformers.XLMRobertaConfig`): Model configuration class with all the parameters of the
+ config (:class:`~transformers.XLMRobertaConfig`): Model configuration class with all the parameters of the
model. Initializing with a config file does not load the weights associated with the model, only the configuration.
Check out the :meth:`~transformers.PreTrainedModel.from_pretrained` method to load the model weights.
"""
@@ -79,7 +79,7 @@ XLM_ROBERTA_INPUTS_DOCSTRING = r"""
``tokens: the dog is hairy . ``
- Fully encoded sequences or sequence pairs can be obtained using the XLMRobertaTokenizer.encode function with
+ Fully encoded sequences or sequence pairs can be obtained using the XLMRobertaTokenizer.encode function with
the ``add_special_tokens`` parameter set to ``True``.
XLM-RoBERTa is a model with absolute position embeddings so it's usually advised to pad the inputs on
@@ -204,7 +204,7 @@ class XLMRobertaForMaskedLM(RobertaForMaskedLM):
@add_start_docstrings(
- """XLM-RoBERTa Model transformer with a sequence classification/regression head on top (a linear layer
+ """XLM-RoBERTa Model transformer with a sequence classification/regression head on top (a linear layer
on top of the pooled output) e.g. for GLUE tasks. """,
XLM_ROBERTA_START_DOCSTRING,
XLM_ROBERTA_INPUTS_DOCSTRING,
diff --git a/transformers/tokenization_utils.py b/transformers/tokenization_utils.py
index 6cc1bedd..f0df6de6 100644
--- a/transformers/tokenization_utils.py
+++ b/transformers/tokenization_utils.py
@@ -868,7 +868,7 @@ class PreTrainedTokenizer(object):
padding index, up to their max length. If no max length is specified, the padding is done up to the model's max length.
The tokenizer padding sides are handled by the following strings:
- 'left': pads on the left of the sequences
- - 'right': pads on the right of the sequences
+ - 'right': pads on the right of the sequences
Defaults to False: no padding.
return_tensors: (optional) can be set to 'tf' or 'pt' to return respectively TensorFlow tf.constant
or PyTorch torch.Tensor instead of a list of python integers.
@@ -1073,7 +1073,7 @@ class PreTrainedTokenizer(object):
padding index, up to their max length. If no max length is specified, the padding is done up to the model's max length.
The tokenizer padding sides are handled by the following strings:
- 'left': pads on the left of the sequences
- - 'right': pads on the right of the sequences
+ - 'right': pads on the right of the sequences
Defaults to False: no padding.
return_tensors: (optional) can be set to 'tf' or 'pt' to return respectively TensorFlow tf.constant
or PyTorch torch.Tensor instead of a list of python integers.
diff --git a/utils/download_glue_data.py b/utils/download_glue_data.py
index 99eac942..af156293 100644
--- a/utils/download_glue_data.py
+++ b/utils/download_glue_data.py
@@ -2,7 +2,7 @@
Original source: https://gist.github.com/W4ngatang/60c2bdb54d156a41194446737ce03e2e
Note: for legal reasons, we are unable to host MRPC.
-You can either use the version hosted by the SentEval team, which is already tokenized,
+You can either use the version hosted by the SentEval team, which is already tokenized,
or you can download the original data from (https://download.microsoft.com/download/D/4/6/D46FF87A-F6B9-4252-AA8B-3604ED519838/MSRParaphraseCorpus.msi) and extract the data from it manually.
For Windows users, you can run the .msi file. For Mac and Linux users, consider an external library such as 'cabextract' (see below for an example).
You should then rename and place specific files in a folder (see below for an example).