Skip to content

GitLab

"vscode:/vscode.git/clone" did not exist on "c63d91dd1c2314abd79ebc0f835347d720e9c8da"
Unverified Commit d4c2cb40 authored by Julien Chaumond's avatar Julien Chaumond Committed by GitHub
Browse files

Kill model archive maps (#4636) 

* Kill model archive maps

* Fixup

* Also kill model_archive_map for MaskedBertPreTrainedModel

* Unhook config_archive_map

* Tokenizers: align with model id changes

* make style && make quality

* Fix CI
parent 47a551d1
Hide whitespace changes
Inline Side-by-side
Showing with 266 additions and 337 deletions
src/transformers/modeling_bart.py
View file @ d4c2cb40
...@@ -32,13 +32,15 @@ from .modeling_utils import PreTrainedModel, create_position_ids_from_input_ids ...@@ -32,13 +32,15 @@ from .modeling_utils import PreTrainedModel, create_position_ids_from_input_ids
logger = logging.getLogger(__name__) logger = logging.getLogger(__name__)
BART_PRETRAINED_MODEL_ARCHIVE_MAP = { BART_PRETRAINED_MODEL_ARCHIVE_LIST = [
"bart-large": "https://cdn.huggingface.co/facebook/bart-large/pytorch_model.bin", "facebook/bart-large",
"bart-large-mnli": "https://cdn.huggingface.co/facebook/bart-large-mnli/pytorch_model.bin", "facebook/bart-large-mnli",
"bart-large-cnn": "https://cdn.huggingface.co/facebook/bart-large-cnn/pytorch_model.bin", "facebook/bart-large-cnn",
"bart-large-xsum": "https://cdn.huggingface.co/facebook/bart-large-xsum/pytorch_model.bin", "facebook/bart-large-xsum",
"mbart-large-en-ro": "https://cdn.huggingface.co/facebook/mbart-large-en-ro/pytorch_model.bin", "facebook/mbart-large-en-ro",
} # See all BART models at https://huggingface.co/models?filter=bart
]
BART_START_DOCSTRING = r""" BART_START_DOCSTRING = r"""
...@@ -118,7 +120,6 @@ def _prepare_bart_decoder_inputs( ...@@ -118,7 +120,6 @@ def _prepare_bart_decoder_inputs(
class PretrainedBartModel(PreTrainedModel): class PretrainedBartModel(PreTrainedModel):
config_class = BartConfig config_class = BartConfig
base_model_prefix = "model" base_model_prefix = "model"
pretrained_model_archive_map = BART_PRETRAINED_MODEL_ARCHIVE_MAP
def _init_weights(self, module): def _init_weights(self, module):
std = self.config.init_std std = self.config.init_std
......
src/transformers/modeling_bert.py
View file @ d4c2cb40
...@@ -32,30 +32,31 @@ from .modeling_utils import PreTrainedModel, prune_linear_layer ...@@ -32,30 +32,31 @@ from .modeling_utils import PreTrainedModel, prune_linear_layer
logger = logging.getLogger(__name__) logger = logging.getLogger(__name__)
BERT_PRETRAINED_MODEL_ARCHIVE_MAP = { BERT_PRETRAINED_MODEL_ARCHIVE_LIST = [
"bert-base-uncased": "https://cdn.huggingface.co/bert-base-uncased-pytorch_model.bin", "bert-base-uncased",
"bert-large-uncased": "https://cdn.huggingface.co/bert-large-uncased-pytorch_model.bin", "bert-large-uncased",
"bert-base-cased": "https://cdn.huggingface.co/bert-base-cased-pytorch_model.bin", "bert-base-cased",
"bert-large-cased": "https://cdn.huggingface.co/bert-large-cased-pytorch_model.bin", "bert-large-cased",
"bert-base-multilingual-uncased": "https://cdn.huggingface.co/bert-base-multilingual-uncased-pytorch_model.bin", "bert-base-multilingual-uncased",
"bert-base-multilingual-cased": "https://cdn.huggingface.co/bert-base-multilingual-cased-pytorch_model.bin", "bert-base-multilingual-cased",
"bert-base-chinese": "https://cdn.huggingface.co/bert-base-chinese-pytorch_model.bin", "bert-base-chinese",
"bert-base-german-cased": "https://cdn.huggingface.co/bert-base-german-cased-pytorch_model.bin", "bert-base-german-cased",
"bert-large-uncased-whole-word-masking": "https://cdn.huggingface.co/bert-large-uncased-whole-word-masking-pytorch_model.bin", "bert-large-uncased-whole-word-masking",
"bert-large-cased-whole-word-masking": "https://cdn.huggingface.co/bert-large-cased-whole-word-masking-pytorch_model.bin", "bert-large-cased-whole-word-masking",
"bert-large-uncased-whole-word-masking-finetuned-squad": "https://cdn.huggingface.co/bert-large-uncased-whole-word-masking-finetuned-squad-pytorch_model.bin", "bert-large-uncased-whole-word-masking-finetuned-squad",
"bert-large-cased-whole-word-masking-finetuned-squad": "https://cdn.huggingface.co/bert-large-cased-whole-word-masking-finetuned-squad-pytorch_model.bin", "bert-large-cased-whole-word-masking-finetuned-squad",
"bert-base-cased-finetuned-mrpc": "https://cdn.huggingface.co/bert-base-cased-finetuned-mrpc-pytorch_model.bin", "bert-base-cased-finetuned-mrpc",
"bert-base-german-dbmdz-cased": "https://cdn.huggingface.co/bert-base-german-dbmdz-cased-pytorch_model.bin", "bert-base-german-dbmdz-cased",
"bert-base-german-dbmdz-uncased": "https://cdn.huggingface.co/bert-base-german-dbmdz-uncased-pytorch_model.bin", "bert-base-german-dbmdz-uncased",
"bert-base-japanese": "https://cdn.huggingface.co/cl-tohoku/bert-base-japanese/pytorch_model.bin", "cl-tohoku/bert-base-japanese",
"bert-base-japanese-whole-word-masking": "https://cdn.huggingface.co/cl-tohoku/bert-base-japanese-whole-word-masking/pytorch_model.bin", "cl-tohoku/bert-base-japanese-whole-word-masking",
"bert-base-japanese-char": "https://cdn.huggingface.co/cl-tohoku/bert-base-japanese-char/pytorch_model.bin", "cl-tohoku/bert-base-japanese-char",
"bert-base-japanese-char-whole-word-masking": "https://cdn.huggingface.co/cl-tohoku/bert-base-japanese-char-whole-word-masking/pytorch_model.bin", "cl-tohoku/bert-base-japanese-char-whole-word-masking",
"bert-base-finnish-cased-v1": "https://cdn.huggingface.co/TurkuNLP/bert-base-finnish-cased-v1/pytorch_model.bin", "TurkuNLP/bert-base-finnish-cased-v1",
"bert-base-finnish-uncased-v1": "https://cdn.huggingface.co/TurkuNLP/bert-base-finnish-uncased-v1/pytorch_model.bin", "TurkuNLP/bert-base-finnish-uncased-v1",
"bert-base-dutch-cased": "https://cdn.huggingface.co/wietsedv/bert-base-dutch-cased/pytorch_model.bin", "wietsedv/bert-base-dutch-cased",
} # See all BERT models at https://huggingface.co/models?filter=bert
]
def load_tf_weights_in_bert(model, config, tf_checkpoint_path): def load_tf_weights_in_bert(model, config, tf_checkpoint_path):
...@@ -513,7 +514,6 @@ class BertPreTrainedModel(PreTrainedModel): ...@@ -513,7 +514,6 @@ class BertPreTrainedModel(PreTrainedModel):
""" """
config_class = BertConfig config_class = BertConfig
pretrained_model_archive_map = BERT_PRETRAINED_MODEL_ARCHIVE_MAP
load_tf_weights = load_tf_weights_in_bert load_tf_weights = load_tf_weights_in_bert
base_model_prefix = "bert" base_model_prefix = "bert"
......
src/transformers/modeling_camembert.py
View file @ d4c2cb40
...@@ -31,11 +31,12 @@ from .modeling_roberta import ( ...@@ -31,11 +31,12 @@ from .modeling_roberta import (
logger = logging.getLogger(__name__) logger = logging.getLogger(__name__)
CAMEMBERT_PRETRAINED_MODEL_ARCHIVE_MAP = { CAMEMBERT_PRETRAINED_MODEL_ARCHIVE_LIST = [
"camembert-base": "https://cdn.huggingface.co/camembert-base-pytorch_model.bin", "camembert-base",
"umberto-commoncrawl-cased-v1": "https://cdn.huggingface.co/Musixmatch/umberto-commoncrawl-cased-v1/pytorch_model.bin", "Musixmatch/umberto-commoncrawl-cased-v1",
"umberto-wikipedia-uncased-v1": "https://cdn.huggingface.co/Musixmatch/umberto-wikipedia-uncased-v1/pytorch_model.bin", "Musixmatch/umberto-wikipedia-uncased-v1",
} # See all CamemBERT models at https://huggingface.co/models?filter=camembert
]
CAMEMBERT_START_DOCSTRING = r""" CAMEMBERT_START_DOCSTRING = r"""
...@@ -62,7 +63,6 @@ class CamembertModel(RobertaModel): ...@@ -62,7 +63,6 @@ class CamembertModel(RobertaModel):
""" """
config_class = CamembertConfig config_class = CamembertConfig
pretrained_model_archive_map = CAMEMBERT_PRETRAINED_MODEL_ARCHIVE_MAP
@add_start_docstrings( @add_start_docstrings(
...@@ -75,7 +75,6 @@ class CamembertForMaskedLM(RobertaForMaskedLM): ...@@ -75,7 +75,6 @@ class CamembertForMaskedLM(RobertaForMaskedLM):
""" """
config_class = CamembertConfig config_class = CamembertConfig
pretrained_model_archive_map = CAMEMBERT_PRETRAINED_MODEL_ARCHIVE_MAP
@add_start_docstrings( @add_start_docstrings(
...@@ -90,7 +89,6 @@ class CamembertForSequenceClassification(RobertaForSequenceClassification): ...@@ -90,7 +89,6 @@ class CamembertForSequenceClassification(RobertaForSequenceClassification):
""" """
config_class = CamembertConfig config_class = CamembertConfig
pretrained_model_archive_map = CAMEMBERT_PRETRAINED_MODEL_ARCHIVE_MAP
@add_start_docstrings( @add_start_docstrings(
...@@ -105,7 +103,6 @@ class CamembertForMultipleChoice(RobertaForMultipleChoice): ...@@ -105,7 +103,6 @@ class CamembertForMultipleChoice(RobertaForMultipleChoice):
""" """
config_class = CamembertConfig config_class = CamembertConfig
pretrained_model_archive_map = CAMEMBERT_PRETRAINED_MODEL_ARCHIVE_MAP
@add_start_docstrings( @add_start_docstrings(
...@@ -120,7 +117,6 @@ class CamembertForTokenClassification(RobertaForTokenClassification): ...@@ -120,7 +117,6 @@ class CamembertForTokenClassification(RobertaForTokenClassification):
""" """
config_class = CamembertConfig config_class = CamembertConfig
pretrained_model_archive_map = CAMEMBERT_PRETRAINED_MODEL_ARCHIVE_MAP
@add_start_docstrings( @add_start_docstrings(
...@@ -135,4 +131,3 @@ class CamembertForQuestionAnswering(RobertaForQuestionAnswering): ...@@ -135,4 +131,3 @@ class CamembertForQuestionAnswering(RobertaForQuestionAnswering):
""" """
config_class = CamembertConfig config_class = CamembertConfig
pretrained_model_archive_map = CAMEMBERT_PRETRAINED_MODEL_ARCHIVE_MAP
src/transformers/modeling_ctrl.py
View file @ d4c2cb40
...@@ -30,7 +30,10 @@ from .modeling_utils import Conv1D, PreTrainedModel ...@@ -30,7 +30,10 @@ from .modeling_utils import Conv1D, PreTrainedModel
logger = logging.getLogger(__name__) logger = logging.getLogger(__name__)
CTRL_PRETRAINED_MODEL_ARCHIVE_MAP = {"ctrl": "https://storage.googleapis.com/sf-ctrl/pytorch/seqlen256_v1.bin"} CTRL_PRETRAINED_MODEL_ARCHIVE_LIST = [
"ctrl"
# See all CTRL models at https://huggingface.co/models?filter=ctrl
]
def angle_defn(pos, i, d_model_size): def angle_defn(pos, i, d_model_size):
...@@ -178,7 +181,6 @@ class CTRLPreTrainedModel(PreTrainedModel): ...@@ -178,7 +181,6 @@ class CTRLPreTrainedModel(PreTrainedModel):
""" """
config_class = CTRLConfig config_class = CTRLConfig
pretrained_model_archive_map = CTRL_PRETRAINED_MODEL_ARCHIVE_MAP
base_model_prefix = "transformer" base_model_prefix = "transformer"
def _init_weights(self, module): def _init_weights(self, module):
......
src/transformers/modeling_distilbert.py
View file @ d4c2cb40
...@@ -36,15 +36,16 @@ from .modeling_utils import PreTrainedModel, prune_linear_layer ...@@ -36,15 +36,16 @@ from .modeling_utils import PreTrainedModel, prune_linear_layer
logger = logging.getLogger(__name__) logger = logging.getLogger(__name__)
DISTILBERT_PRETRAINED_MODEL_ARCHIVE_MAP = { DISTILBERT_PRETRAINED_MODEL_ARCHIVE_LIST = [
"distilbert-base-uncased": "https://cdn.huggingface.co/distilbert-base-uncased-pytorch_model.bin", "distilbert-base-uncased",
"distilbert-base-uncased-distilled-squad": "https://cdn.huggingface.co/distilbert-base-uncased-distilled-squad-pytorch_model.bin", "distilbert-base-uncased-distilled-squad",
"distilbert-base-cased": "https://cdn.huggingface.co/distilbert-base-cased-pytorch_model.bin", "distilbert-base-cased",
"distilbert-base-cased-distilled-squad": "https://cdn.huggingface.co/distilbert-base-cased-distilled-squad-pytorch_model.bin", "distilbert-base-cased-distilled-squad",
"distilbert-base-german-cased": "https://cdn.huggingface.co/distilbert-base-german-cased-pytorch_model.bin", "distilbert-base-german-cased",
"distilbert-base-multilingual-cased": "https://cdn.huggingface.co/distilbert-base-multilingual-cased-pytorch_model.bin", "distilbert-base-multilingual-cased",
"distilbert-base-uncased-finetuned-sst-2-english": "https://cdn.huggingface.co/distilbert-base-uncased-finetuned-sst-2-english-pytorch_model.bin", "distilbert-base-uncased-finetuned-sst-2-english",
} # See all DistilBERT models at https://huggingface.co/models?filter=distilbert
]
# UTILS AND BUILDING BLOCKS OF THE ARCHITECTURE # # UTILS AND BUILDING BLOCKS OF THE ARCHITECTURE #
...@@ -327,7 +328,6 @@ class DistilBertPreTrainedModel(PreTrainedModel): ...@@ -327,7 +328,6 @@ class DistilBertPreTrainedModel(PreTrainedModel):
""" """
config_class = DistilBertConfig config_class = DistilBertConfig
pretrained_model_archive_map = DISTILBERT_PRETRAINED_MODEL_ARCHIVE_MAP
load_tf_weights = None load_tf_weights = None
base_model_prefix = "distilbert" base_model_prefix = "distilbert"
......
src/transformers/modeling_electra.py
View file @ d4c2cb40
...@@ -14,14 +14,15 @@ from .modeling_bert import BertEmbeddings, BertEncoder, BertLayerNorm, BertPreTr ...@@ -14,14 +14,15 @@ from .modeling_bert import BertEmbeddings, BertEncoder, BertLayerNorm, BertPreTr
logger = logging.getLogger(__name__) logger = logging.getLogger(__name__)
ELECTRA_PRETRAINED_MODEL_ARCHIVE_MAP = { ELECTRA_PRETRAINED_MODEL_ARCHIVE_LIST = [
"google/electra-small-generator": "https://cdn.huggingface.co/google/electra-small-generator/pytorch_model.bin", "google/electra-small-generator",
"google/electra-base-generator": "https://cdn.huggingface.co/google/electra-base-generator/pytorch_model.bin", "google/electra-base-generator",
"google/electra-large-generator": "https://cdn.huggingface.co/google/electra-large-generator/pytorch_model.bin", "google/electra-large-generator",
"google/electra-small-discriminator": "https://cdn.huggingface.co/google/electra-small-discriminator/pytorch_model.bin", "google/electra-small-discriminator",
"google/electra-base-discriminator": "https://cdn.huggingface.co/google/electra-base-discriminator/pytorch_model.bin", "google/electra-base-discriminator",
"google/electra-large-discriminator": "https://cdn.huggingface.co/google/electra-large-discriminator/pytorch_model.bin", "google/electra-large-discriminator",
} # See all ELECTRA models at https://huggingface.co/models?filter=electra
]
def load_tf_weights_in_electra(model, config, tf_checkpoint_path, discriminator_or_generator="discriminator"): def load_tf_weights_in_electra(model, config, tf_checkpoint_path, discriminator_or_generator="discriminator"):
...@@ -160,7 +161,6 @@ class ElectraPreTrainedModel(BertPreTrainedModel): ...@@ -160,7 +161,6 @@ class ElectraPreTrainedModel(BertPreTrainedModel):
""" """
config_class = ElectraConfig config_class = ElectraConfig
pretrained_model_archive_map = ELECTRA_PRETRAINED_MODEL_ARCHIVE_MAP
load_tf_weights = load_tf_weights_in_electra load_tf_weights = load_tf_weights_in_electra
base_model_prefix = "electra" base_model_prefix = "electra"
......
src/transformers/modeling_flaubert.py
View file @ d4c2cb40
...@@ -35,12 +35,13 @@ from .modeling_xlm import ( ...@@ -35,12 +35,13 @@ from .modeling_xlm import (
logger = logging.getLogger(__name__) logger = logging.getLogger(__name__)
FLAUBERT_PRETRAINED_MODEL_ARCHIVE_MAP = { FLAUBERT_PRETRAINED_MODEL_ARCHIVE_LIST = [
"flaubert-small-cased": "https://cdn.huggingface.co/flaubert/flaubert_small_cased/pytorch_model.bin", "flaubert/flaubert_small_cased",
"flaubert-base-uncased": "https://cdn.huggingface.co/flaubert/flaubert_base_uncased/pytorch_model.bin", "flaubert/flaubert_base_uncased",
"flaubert-base-cased": "https://cdn.huggingface.co/flaubert/flaubert_base_cased/pytorch_model.bin", "flaubert/flaubert_base_cased",
"flaubert-large-cased": "https://cdn.huggingface.co/flaubert/flaubert_large_cased/pytorch_model.bin", "flaubert/flaubert_large_cased",
} # See all Flaubert models at https://huggingface.co/models?filter=flaubert
]
FLAUBERT_START_DOCSTRING = r""" FLAUBERT_START_DOCSTRING = r"""
...@@ -109,7 +110,6 @@ FLAUBERT_INPUTS_DOCSTRING = r""" ...@@ -109,7 +110,6 @@ FLAUBERT_INPUTS_DOCSTRING = r"""
class FlaubertModel(XLMModel): class FlaubertModel(XLMModel):
config_class = FlaubertConfig config_class = FlaubertConfig
pretrained_model_archive_map = FLAUBERT_PRETRAINED_MODEL_ARCHIVE_MAP
def __init__(self, config): # , dico, is_encoder, with_output): def __init__(self, config): # , dico, is_encoder, with_output):
super().__init__(config) super().__init__(config)
...@@ -304,7 +304,6 @@ class FlaubertWithLMHeadModel(XLMWithLMHeadModel): ...@@ -304,7 +304,6 @@ class FlaubertWithLMHeadModel(XLMWithLMHeadModel):
""" """
config_class = FlaubertConfig config_class = FlaubertConfig
pretrained_model_archive_map = FLAUBERT_PRETRAINED_MODEL_ARCHIVE_MAP
def __init__(self, config): def __init__(self, config):
super().__init__(config) super().__init__(config)
...@@ -324,7 +323,6 @@ class FlaubertForSequenceClassification(XLMForSequenceClassification): ...@@ -324,7 +323,6 @@ class FlaubertForSequenceClassification(XLMForSequenceClassification):
""" """
config_class = FlaubertConfig config_class = FlaubertConfig
pretrained_model_archive_map = FLAUBERT_PRETRAINED_MODEL_ARCHIVE_MAP
def __init__(self, config): def __init__(self, config):
super().__init__(config) super().__init__(config)
...@@ -344,7 +342,6 @@ class FlaubertForQuestionAnsweringSimple(XLMForQuestionAnsweringSimple): ...@@ -344,7 +342,6 @@ class FlaubertForQuestionAnsweringSimple(XLMForQuestionAnsweringSimple):
""" """
config_class = FlaubertConfig config_class = FlaubertConfig
pretrained_model_archive_map = FLAUBERT_PRETRAINED_MODEL_ARCHIVE_MAP
def __init__(self, config): def __init__(self, config):
super().__init__(config) super().__init__(config)
...@@ -364,7 +361,6 @@ class FlaubertForQuestionAnswering(XLMForQuestionAnswering): ...@@ -364,7 +361,6 @@ class FlaubertForQuestionAnswering(XLMForQuestionAnswering):
""" """
config_class = FlaubertConfig config_class = FlaubertConfig
pretrained_model_archive_map = FLAUBERT_PRETRAINED_MODEL_ARCHIVE_MAP
def __init__(self, config): def __init__(self, config):
super().__init__(config) super().__init__(config)
......
src/transformers/modeling_gpt2.py
View file @ d4c2cb40
...@@ -31,13 +31,14 @@ from .modeling_utils import Conv1D, PreTrainedModel, SequenceSummary, prune_conv ...@@ -31,13 +31,14 @@ from .modeling_utils import Conv1D, PreTrainedModel, SequenceSummary, prune_conv
logger = logging.getLogger(__name__) logger = logging.getLogger(__name__)
GPT2_PRETRAINED_MODEL_ARCHIVE_MAP = { GPT2_PRETRAINED_MODEL_ARCHIVE_LIST = [
"gpt2": "https://cdn.huggingface.co/gpt2-pytorch_model.bin", "gpt2",
"gpt2-medium": "https://cdn.huggingface.co/gpt2-medium-pytorch_model.bin", "gpt2-medium",
"gpt2-large": "https://cdn.huggingface.co/gpt2-large-pytorch_model.bin", "gpt2-large",
"gpt2-xl": "https://cdn.huggingface.co/gpt2-xl-pytorch_model.bin", "gpt2-xl",
"distilgpt2": "https://cdn.huggingface.co/distilgpt2-pytorch_model.bin", "distilgpt2",
} # See all GPT-2 models at https://huggingface.co/models?filter=gpt2
]
def load_tf_weights_in_gpt2(model, config, gpt2_checkpoint_path): def load_tf_weights_in_gpt2(model, config, gpt2_checkpoint_path):
...@@ -251,7 +252,6 @@ class GPT2PreTrainedModel(PreTrainedModel): ...@@ -251,7 +252,6 @@ class GPT2PreTrainedModel(PreTrainedModel):
""" """
config_class = GPT2Config config_class = GPT2Config
pretrained_model_archive_map = GPT2_PRETRAINED_MODEL_ARCHIVE_MAP
load_tf_weights = load_tf_weights_in_gpt2 load_tf_weights = load_tf_weights_in_gpt2
base_model_prefix = "transformer" base_model_prefix = "transformer"
......
src/transformers/modeling_longformer.py
View file @ d4c2cb40
...@@ -30,13 +30,14 @@ from .modeling_roberta import RobertaLMHead, RobertaModel ...@@ -30,13 +30,14 @@ from .modeling_roberta import RobertaLMHead, RobertaModel
logger = logging.getLogger(__name__) logger = logging.getLogger(__name__)
LONGFORMER_PRETRAINED_MODEL_ARCHIVE_MAP = { LONGFORMER_PRETRAINED_MODEL_ARCHIVE_LIST = [
"allenai/longformer-base-4096": "https://s3.amazonaws.com/models.huggingface.co/bert/allenai/longformer-base-4096/pytorch_model.bin", "allenai/longformer-base-4096",
"allenai/longformer-large-4096": "https://s3.amazonaws.com/models.huggingface.co/bert/allenai/longformer-large-4096/pytorch_model.bin", "allenai/longformer-large-4096",
"allenai/longformer-large-4096-finetuned-triviaqa": "https://s3.amazonaws.com/models.huggingface.co/bert/allenai/longformer-large-4096-finetuned-triviaqa/pytorch_model.bin", "allenai/longformer-large-4096-finetuned-triviaqa",
"allenai/longformer-base-4096-extra.pos.embd.only": "https://s3.amazonaws.com/models.huggingface.co/bert/allenai/longformer-base-4096-extra.pos.embd.only/pytorch_model.bin", "allenai/longformer-base-4096-extra.pos.embd.only",
"allenai/longformer-large-4096-extra.pos.embd.only": "https://s3.amazonaws.com/models.huggingface.co/bert/allenai/longformer-large-4096-extra.pos.embd.only/pytorch_model.bin", "allenai/longformer-large-4096-extra.pos.embd.only",
} # See all Longformer models at https://huggingface.co/models?filter=longformer
]
def _get_question_end_index(input_ids, sep_token_id): def _get_question_end_index(input_ids, sep_token_id):
...@@ -513,7 +514,6 @@ class LongformerModel(RobertaModel): ...@@ -513,7 +514,6 @@ class LongformerModel(RobertaModel):
""" """
config_class = LongformerConfig config_class = LongformerConfig
pretrained_model_archive_map = LONGFORMER_PRETRAINED_MODEL_ARCHIVE_MAP
base_model_prefix = "longformer" base_model_prefix = "longformer"
def __init__(self, config): def __init__(self, config):
...@@ -685,7 +685,6 @@ class LongformerModel(RobertaModel): ...@@ -685,7 +685,6 @@ class LongformerModel(RobertaModel):
@add_start_docstrings("""Longformer Model with a `language modeling` head on top. """, LONGFORMER_START_DOCSTRING) @add_start_docstrings("""Longformer Model with a `language modeling` head on top. """, LONGFORMER_START_DOCSTRING)
class LongformerForMaskedLM(BertPreTrainedModel): class LongformerForMaskedLM(BertPreTrainedModel):
config_class = LongformerConfig config_class = LongformerConfig
pretrained_model_archive_map = LONGFORMER_PRETRAINED_MODEL_ARCHIVE_MAP
base_model_prefix = "longformer" base_model_prefix = "longformer"
def __init__(self, config): def __init__(self, config):
...@@ -776,7 +775,6 @@ class LongformerForMaskedLM(BertPreTrainedModel): ...@@ -776,7 +775,6 @@ class LongformerForMaskedLM(BertPreTrainedModel):
) )
class LongformerForSequenceClassification(BertPreTrainedModel): class LongformerForSequenceClassification(BertPreTrainedModel):
config_class = LongformerConfig config_class = LongformerConfig
pretrained_model_archive_map = LONGFORMER_PRETRAINED_MODEL_ARCHIVE_MAP
base_model_prefix = "longformer" base_model_prefix = "longformer"
def __init__(self, config): def __init__(self, config):
...@@ -893,7 +891,6 @@ class LongformerClassificationHead(nn.Module): ...@@ -893,7 +891,6 @@ class LongformerClassificationHead(nn.Module):
) )
class LongformerForQuestionAnswering(BertPreTrainedModel): class LongformerForQuestionAnswering(BertPreTrainedModel):
config_class = LongformerConfig config_class = LongformerConfig
pretrained_model_archive_map = LONGFORMER_PRETRAINED_MODEL_ARCHIVE_MAP
base_model_prefix = "longformer" base_model_prefix = "longformer"
def __init__(self, config): def __init__(self, config):
...@@ -1018,7 +1015,6 @@ class LongformerForQuestionAnswering(BertPreTrainedModel): ...@@ -1018,7 +1015,6 @@ class LongformerForQuestionAnswering(BertPreTrainedModel):
) )
class LongformerForTokenClassification(BertPreTrainedModel): class LongformerForTokenClassification(BertPreTrainedModel):
config_class = LongformerConfig config_class = LongformerConfig
pretrained_model_archive_map = LONGFORMER_PRETRAINED_MODEL_ARCHIVE_MAP
base_model_prefix = "longformer" base_model_prefix = "longformer"
def __init__(self, config): def __init__(self, config):
...@@ -1119,7 +1115,6 @@ class LongformerForTokenClassification(BertPreTrainedModel): ...@@ -1119,7 +1115,6 @@ class LongformerForTokenClassification(BertPreTrainedModel):
) )
class LongformerForMultipleChoice(BertPreTrainedModel): class LongformerForMultipleChoice(BertPreTrainedModel):
config_class = LongformerConfig config_class = LongformerConfig
pretrained_model_archive_map = LONGFORMER_PRETRAINED_MODEL_ARCHIVE_MAP
base_model_prefix = "longformer" base_model_prefix = "longformer"
def __init__(self, config): def __init__(self, config):
......
src/transformers/modeling_marian.py
View file @ d4c2cb40
...@@ -18,6 +18,11 @@ ...@@ -18,6 +18,11 @@
from transformers.modeling_bart import BartForConditionalGeneration from transformers.modeling_bart import BartForConditionalGeneration
MARIAN_PRETRAINED_MODEL_ARCHIVE_LIST = [
# See all Marian models at https://huggingface.co/models?search=Helsinki-NLP
]
class MarianMTModel(BartForConditionalGeneration): class MarianMTModel(BartForConditionalGeneration):
r""" r"""
Pytorch version of marian-nmt's transformer.h (c++). Designed for the OPUS-NMT translation checkpoints. Pytorch version of marian-nmt's transformer.h (c++). Designed for the OPUS-NMT translation checkpoints.
...@@ -41,8 +46,6 @@ class MarianMTModel(BartForConditionalGeneration): ...@@ -41,8 +46,6 @@ class MarianMTModel(BartForConditionalGeneration):
""" """
pretrained_model_archive_map = {} # see https://huggingface.co/models?search=Helsinki-NLP
def prepare_logits_for_generation(self, logits, cur_len, max_length): def prepare_logits_for_generation(self, logits, cur_len, max_length):
logits[:, self.config.pad_token_id] = float("-inf") logits[:, self.config.pad_token_id] = float("-inf")
if cur_len == max_length - 1 and self.config.eos_token_id is not None: if cur_len == max_length - 1 and self.config.eos_token_id is not None:
......
src/transformers/modeling_openai.py
View file @ d4c2cb40
...@@ -33,7 +33,10 @@ from .modeling_utils import Conv1D, PreTrainedModel, SequenceSummary, prune_conv ...@@ -33,7 +33,10 @@ from .modeling_utils import Conv1D, PreTrainedModel, SequenceSummary, prune_conv
logger = logging.getLogger(__name__) logger = logging.getLogger(__name__)
OPENAI_GPT_PRETRAINED_MODEL_ARCHIVE_MAP = {"openai-gpt": "https://cdn.huggingface.co/openai-gpt-pytorch_model.bin"} OPENAI_GPT_PRETRAINED_MODEL_ARCHIVE_LIST = [
"openai-gpt",
# See all OpenAI GPT models at https://huggingface.co/models?filter=openai-gpt
]
def load_tf_weights_in_openai_gpt(model, config, openai_checkpoint_folder_path): def load_tf_weights_in_openai_gpt(model, config, openai_checkpoint_folder_path):
...@@ -252,7 +255,6 @@ class OpenAIGPTPreTrainedModel(PreTrainedModel): ...@@ -252,7 +255,6 @@ class OpenAIGPTPreTrainedModel(PreTrainedModel):
""" """
config_class = OpenAIGPTConfig config_class = OpenAIGPTConfig
pretrained_model_archive_map = OPENAI_GPT_PRETRAINED_MODEL_ARCHIVE_MAP
load_tf_weights = load_tf_weights_in_openai_gpt load_tf_weights = load_tf_weights_in_openai_gpt
base_model_prefix = "transformer" base_model_prefix = "transformer"
......
src/transformers/modeling_reformer.py
View file @ d4c2cb40
...@@ -35,10 +35,11 @@ from .modeling_utils import PreTrainedModel, apply_chunking_to_forward ...@@ -35,10 +35,11 @@ from .modeling_utils import PreTrainedModel, apply_chunking_to_forward
logger = logging.getLogger(__name__) logger = logging.getLogger(__name__)
REFORMER_PRETRAINED_MODEL_ARCHIVE_MAP = { REFORMER_PRETRAINED_MODEL_ARCHIVE_LIST = [
"google/reformer-crime-and-punishment": "https://cdn.huggingface.co/google/reformer-crime-and-punishment/pytorch_model.bin", "google/reformer-crime-and-punishment",
"google/reformer-enwik8": "https://cdn.huggingface.co/google/reformer-enwik8/pytorch_model.bin", "google/reformer-enwik8",
} # See all Reformer models at https://huggingface.co/models?filter=reformer
]
def mish(x): def mish(x):
...@@ -1373,7 +1374,6 @@ class ReformerPreTrainedModel(PreTrainedModel): ...@@ -1373,7 +1374,6 @@ class ReformerPreTrainedModel(PreTrainedModel):
""" """
config_class = ReformerConfig config_class = ReformerConfig
pretrained_model_archive_map = REFORMER_PRETRAINED_MODEL_ARCHIVE_MAP
base_model_prefix = "reformer" base_model_prefix = "reformer"
@property @property
......
src/transformers/modeling_roberta.py
View file @ d4c2cb40
...@@ -30,14 +30,15 @@ from .modeling_utils import create_position_ids_from_input_ids ...@@ -30,14 +30,15 @@ from .modeling_utils import create_position_ids_from_input_ids
logger = logging.getLogger(__name__) logger = logging.getLogger(__name__)
ROBERTA_PRETRAINED_MODEL_ARCHIVE_MAP = { ROBERTA_PRETRAINED_MODEL_ARCHIVE_LIST = [
"roberta-base": "https://cdn.huggingface.co/roberta-base-pytorch_model.bin", "roberta-base",
"roberta-large": "https://cdn.huggingface.co/roberta-large-pytorch_model.bin", "roberta-large",
"roberta-large-mnli": "https://cdn.huggingface.co/roberta-large-mnli-pytorch_model.bin", "roberta-large-mnli",
"distilroberta-base": "https://cdn.huggingface.co/distilroberta-base-pytorch_model.bin", "distilroberta-base",
"roberta-base-openai-detector": "https://cdn.huggingface.co/roberta-base-openai-detector-pytorch_model.bin", "roberta-base-openai-detector",
"roberta-large-openai-detector": "https://cdn.huggingface.co/roberta-large-openai-detector-pytorch_model.bin", "roberta-large-openai-detector",
} # See all RoBERTa models at https://huggingface.co/models?filter=roberta
]
class RobertaEmbeddings(BertEmbeddings): class RobertaEmbeddings(BertEmbeddings):
...@@ -142,7 +143,6 @@ class RobertaModel(BertModel): ...@@ -142,7 +143,6 @@ class RobertaModel(BertModel):
""" """
config_class = RobertaConfig config_class = RobertaConfig
pretrained_model_archive_map = ROBERTA_PRETRAINED_MODEL_ARCHIVE_MAP
base_model_prefix = "roberta" base_model_prefix = "roberta"
def __init__(self, config): def __init__(self, config):
...@@ -161,7 +161,6 @@ class RobertaModel(BertModel): ...@@ -161,7 +161,6 @@ class RobertaModel(BertModel):
@add_start_docstrings("""RoBERTa Model with a `language modeling` head on top. """, ROBERTA_START_DOCSTRING) @add_start_docstrings("""RoBERTa Model with a `language modeling` head on top. """, ROBERTA_START_DOCSTRING)
class RobertaForMaskedLM(BertPreTrainedModel): class RobertaForMaskedLM(BertPreTrainedModel):
config_class = RobertaConfig config_class = RobertaConfig
pretrained_model_archive_map = ROBERTA_PRETRAINED_MODEL_ARCHIVE_MAP
base_model_prefix = "roberta" base_model_prefix = "roberta"
def __init__(self, config): def __init__(self, config):
...@@ -276,7 +275,6 @@ class RobertaLMHead(nn.Module): ...@@ -276,7 +275,6 @@ class RobertaLMHead(nn.Module):
) )
class RobertaForSequenceClassification(BertPreTrainedModel): class RobertaForSequenceClassification(BertPreTrainedModel):
config_class = RobertaConfig config_class = RobertaConfig
pretrained_model_archive_map = ROBERTA_PRETRAINED_MODEL_ARCHIVE_MAP
base_model_prefix = "roberta" base_model_prefix = "roberta"
def __init__(self, config): def __init__(self, config):
...@@ -367,7 +365,6 @@ class RobertaForSequenceClassification(BertPreTrainedModel): ...@@ -367,7 +365,6 @@ class RobertaForSequenceClassification(BertPreTrainedModel):
) )
class RobertaForMultipleChoice(BertPreTrainedModel): class RobertaForMultipleChoice(BertPreTrainedModel):
config_class = RobertaConfig config_class = RobertaConfig
pretrained_model_archive_map = ROBERTA_PRETRAINED_MODEL_ARCHIVE_MAP
base_model_prefix = "roberta" base_model_prefix = "roberta"
def __init__(self, config): def __init__(self, config):
...@@ -466,7 +463,6 @@ class RobertaForMultipleChoice(BertPreTrainedModel): ...@@ -466,7 +463,6 @@ class RobertaForMultipleChoice(BertPreTrainedModel):
) )
class RobertaForTokenClassification(BertPreTrainedModel): class RobertaForTokenClassification(BertPreTrainedModel):
config_class = RobertaConfig config_class = RobertaConfig
pretrained_model_archive_map = ROBERTA_PRETRAINED_MODEL_ARCHIVE_MAP
base_model_prefix = "roberta" base_model_prefix = "roberta"
def __init__(self, config): def __init__(self, config):
...@@ -586,7 +582,6 @@ class RobertaClassificationHead(nn.Module): ...@@ -586,7 +582,6 @@ class RobertaClassificationHead(nn.Module):
) )
class RobertaForQuestionAnswering(BertPreTrainedModel): class RobertaForQuestionAnswering(BertPreTrainedModel):
config_class = RobertaConfig config_class = RobertaConfig
pretrained_model_archive_map = ROBERTA_PRETRAINED_MODEL_ARCHIVE_MAP
base_model_prefix = "roberta" base_model_prefix = "roberta"
def __init__(self, config): def __init__(self, config):
......
src/transformers/modeling_t5.py
View file @ d4c2cb40
...@@ -36,13 +36,14 @@ logger = logging.getLogger(__name__) ...@@ -36,13 +36,14 @@ logger = logging.getLogger(__name__)
# This dict contrains shortcut names and associated url # This dict contrains shortcut names and associated url
# for the pretrained weights provided with the models # for the pretrained weights provided with the models
#################################################### ####################################################
T5_PRETRAINED_MODEL_ARCHIVE_MAP = { T5_PRETRAINED_MODEL_ARCHIVE_LIST = [
"t5-small": "https://cdn.huggingface.co/t5-small-pytorch_model.bin", "t5-small",
"t5-base": "https://cdn.huggingface.co/t5-base-pytorch_model.bin", "t5-base",
"t5-large": "https://cdn.huggingface.co/t5-large-pytorch_model.bin", "t5-large",
"t5-3b": "https://cdn.huggingface.co/t5-3b-pytorch_model.bin", "t5-3b",
"t5-11b": "https://cdn.huggingface.co/t5-11b-pytorch_model.bin", "t5-11b",
} # See all T5 models at https://huggingface.co/models?filter=t5
]
#################################################### ####################################################
...@@ -555,7 +556,6 @@ class T5PreTrainedModel(PreTrainedModel): ...@@ -555,7 +556,6 @@ class T5PreTrainedModel(PreTrainedModel):
""" """
config_class = T5Config config_class = T5Config
pretrained_model_archive_map = T5_PRETRAINED_MODEL_ARCHIVE_MAP
load_tf_weights = load_tf_weights_in_t5 load_tf_weights = load_tf_weights_in_t5
base_model_prefix = "transformer" base_model_prefix = "transformer"
......
src/transformers/modeling_tf_albert.py
View file @ d4c2cb40
...@@ -29,16 +29,17 @@ from .tokenization_utils import BatchEncoding ...@@ -29,16 +29,17 @@ from .tokenization_utils import BatchEncoding
logger = logging.getLogger(__name__) logger = logging.getLogger(__name__)
TF_ALBERT_PRETRAINED_MODEL_ARCHIVE_MAP = { TF_ALBERT_PRETRAINED_MODEL_ARCHIVE_LIST = [
"albert-base-v1": "https://cdn.huggingface.co/albert-base-v1-with-prefix-tf_model.h5", "albert-base-v1",
"albert-large-v1": "https://cdn.huggingface.co/albert-large-v1-with-prefix-tf_model.h5", "albert-large-v1",
"albert-xlarge-v1": "https://cdn.huggingface.co/albert-xlarge-v1-with-prefix-tf_model.h5", "albert-xlarge-v1",
"albert-xxlarge-v1": "https://cdn.huggingface.co/albert-xxlarge-v1-with-prefix-tf_model.h5", "albert-xxlarge-v1",
"albert-base-v2": "https://cdn.huggingface.co/albert-base-v2-with-prefix-tf_model.h5", "albert-base-v2",
"albert-large-v2": "https://cdn.huggingface.co/albert-large-v2-with-prefix-tf_model.h5", "albert-large-v2",
"albert-xlarge-v2": "https://cdn.huggingface.co/albert-xlarge-v2-with-prefix-tf_model.h5", "albert-xlarge-v2",
"albert-xxlarge-v2": "https://cdn.huggingface.co/albert-xxlarge-v2-with-prefix-tf_model.h5", "albert-xxlarge-v2",
} # See all ALBERT models at https://huggingface.co/models?filter=albert
]
class TFAlbertEmbeddings(tf.keras.layers.Layer): class TFAlbertEmbeddings(tf.keras.layers.Layer):
...@@ -440,7 +441,6 @@ class TFAlbertPreTrainedModel(TFPreTrainedModel): ...@@ -440,7 +441,6 @@ class TFAlbertPreTrainedModel(TFPreTrainedModel):
""" """
config_class = AlbertConfig config_class = AlbertConfig
pretrained_model_archive_map = TF_ALBERT_PRETRAINED_MODEL_ARCHIVE_MAP
base_model_prefix = "albert" base_model_prefix = "albert"
......
src/transformers/modeling_tf_auto.py
View file @ d4c2cb40
...@@ -34,7 +34,6 @@ from .configuration_auto import ( ...@@ -34,7 +34,6 @@ from .configuration_auto import (
) )
from .configuration_utils import PretrainedConfig from .configuration_utils import PretrainedConfig
from .modeling_tf_albert import ( from .modeling_tf_albert import (
TF_ALBERT_PRETRAINED_MODEL_ARCHIVE_MAP,
TFAlbertForMaskedLM, TFAlbertForMaskedLM,
TFAlbertForMultipleChoice, TFAlbertForMultipleChoice,
TFAlbertForPreTraining, TFAlbertForPreTraining,
...@@ -43,7 +42,6 @@ from .modeling_tf_albert import ( ...@@ -43,7 +42,6 @@ from .modeling_tf_albert import (
TFAlbertModel, TFAlbertModel,
) )
from .modeling_tf_bert import ( from .modeling_tf_bert import (
TF_BERT_PRETRAINED_MODEL_ARCHIVE_MAP,
TFBertForMaskedLM, TFBertForMaskedLM,
TFBertForMultipleChoice, TFBertForMultipleChoice,
TFBertForPreTraining, TFBertForPreTraining,
...@@ -52,40 +50,32 @@ from .modeling_tf_bert import ( ...@@ -52,40 +50,32 @@ from .modeling_tf_bert import (
TFBertForTokenClassification, TFBertForTokenClassification,
TFBertModel, TFBertModel,
) )
from .modeling_tf_ctrl import TF_CTRL_PRETRAINED_MODEL_ARCHIVE_MAP, TFCTRLLMHeadModel, TFCTRLModel from .modeling_tf_ctrl import TFCTRLLMHeadModel, TFCTRLModel
from .modeling_tf_distilbert import ( from .modeling_tf_distilbert import (
TF_DISTILBERT_PRETRAINED_MODEL_ARCHIVE_MAP,
TFDistilBertForMaskedLM, TFDistilBertForMaskedLM,
TFDistilBertForQuestionAnswering, TFDistilBertForQuestionAnswering,
TFDistilBertForSequenceClassification, TFDistilBertForSequenceClassification,
TFDistilBertForTokenClassification, TFDistilBertForTokenClassification,
TFDistilBertModel, TFDistilBertModel,
) )
from .modeling_tf_gpt2 import TF_GPT2_PRETRAINED_MODEL_ARCHIVE_MAP, TFGPT2LMHeadModel, TFGPT2Model from .modeling_tf_gpt2 import TFGPT2LMHeadModel, TFGPT2Model
from .modeling_tf_openai import TF_OPENAI_GPT_PRETRAINED_MODEL_ARCHIVE_MAP, TFOpenAIGPTLMHeadModel, TFOpenAIGPTModel from .modeling_tf_openai import TFOpenAIGPTLMHeadModel, TFOpenAIGPTModel
from .modeling_tf_roberta import ( from .modeling_tf_roberta import (
TF_ROBERTA_PRETRAINED_MODEL_ARCHIVE_MAP,
TFRobertaForMaskedLM, TFRobertaForMaskedLM,
TFRobertaForQuestionAnswering, TFRobertaForQuestionAnswering,
TFRobertaForSequenceClassification, TFRobertaForSequenceClassification,
TFRobertaForTokenClassification, TFRobertaForTokenClassification,
TFRobertaModel, TFRobertaModel,
) )
from .modeling_tf_t5 import TF_T5_PRETRAINED_MODEL_ARCHIVE_MAP, TFT5ForConditionalGeneration, TFT5Model from .modeling_tf_t5 import TFT5ForConditionalGeneration, TFT5Model
from .modeling_tf_transfo_xl import ( from .modeling_tf_transfo_xl import TFTransfoXLLMHeadModel, TFTransfoXLModel
TF_TRANSFO_XL_PRETRAINED_MODEL_ARCHIVE_MAP,
TFTransfoXLLMHeadModel,
TFTransfoXLModel,
)
from .modeling_tf_xlm import ( from .modeling_tf_xlm import (
TF_XLM_PRETRAINED_MODEL_ARCHIVE_MAP,
TFXLMForQuestionAnsweringSimple, TFXLMForQuestionAnsweringSimple,
TFXLMForSequenceClassification, TFXLMForSequenceClassification,
TFXLMModel, TFXLMModel,
TFXLMWithLMHeadModel, TFXLMWithLMHeadModel,
) )
from .modeling_tf_xlnet import ( from .modeling_tf_xlnet import (
TF_XLNET_PRETRAINED_MODEL_ARCHIVE_MAP,
TFXLNetForQuestionAnsweringSimple, TFXLNetForQuestionAnsweringSimple,
TFXLNetForSequenceClassification, TFXLNetForSequenceClassification,
TFXLNetForTokenClassification, TFXLNetForTokenClassification,
...@@ -97,24 +87,6 @@ from .modeling_tf_xlnet import ( ...@@ -97,24 +87,6 @@ from .modeling_tf_xlnet import (
logger = logging.getLogger(__name__) logger = logging.getLogger(__name__)
TF_ALL_PRETRAINED_MODEL_ARCHIVE_MAP = dict(
(key, value)
for pretrained_map in [
TF_BERT_PRETRAINED_MODEL_ARCHIVE_MAP,
TF_OPENAI_GPT_PRETRAINED_MODEL_ARCHIVE_MAP,
TF_TRANSFO_XL_PRETRAINED_MODEL_ARCHIVE_MAP,
TF_GPT2_PRETRAINED_MODEL_ARCHIVE_MAP,
TF_CTRL_PRETRAINED_MODEL_ARCHIVE_MAP,
TF_XLNET_PRETRAINED_MODEL_ARCHIVE_MAP,
TF_XLM_PRETRAINED_MODEL_ARCHIVE_MAP,
TF_ROBERTA_PRETRAINED_MODEL_ARCHIVE_MAP,
TF_DISTILBERT_PRETRAINED_MODEL_ARCHIVE_MAP,
TF_ALBERT_PRETRAINED_MODEL_ARCHIVE_MAP,
TF_T5_PRETRAINED_MODEL_ARCHIVE_MAP,
]
for key, value, in pretrained_map.items()
)
TF_MODEL_MAPPING = OrderedDict( TF_MODEL_MAPPING = OrderedDict(
[ [
(T5Config, TFT5Model), (T5Config, TFT5Model),
...@@ -208,20 +180,17 @@ class TFAutoModel(object): ...@@ -208,20 +180,17 @@ class TFAutoModel(object):
The `from_pretrained()` method takes care of returning the correct model class instance The `from_pretrained()` method takes care of returning the correct model class instance
based on the `model_type` property of the config object, or when it's missing, based on the `model_type` property of the config object, or when it's missing,
falling back to using pattern matching on the `pretrained_model_name_or_path` string. falling back to using pattern matching on the `pretrained_model_name_or_path` string:
- `t5`: TFT5Model (T5 model)
The base model class to instantiate is selected as the first pattern matching - `distilbert`: TFDistilBertModel (DistilBERT model)
in the `pretrained_model_name_or_path` string (in the following order): - `roberta`: TFRobertaModel (RoBERTa model)
- contains `t5`: TFT5Model (T5 model) - `bert`: TFBertModel (Bert model)
- contains `distilbert`: TFDistilBertModel (DistilBERT model) - `openai-gpt`: TFOpenAIGPTModel (OpenAI GPT model)
- contains `roberta`: TFRobertaModel (RoBERTa model) - `gpt2`: TFGPT2Model (OpenAI GPT-2 model)
- contains `bert`: TFBertModel (Bert model) - `transfo-xl`: TFTransfoXLModel (Transformer-XL model)
- contains `openai-gpt`: TFOpenAIGPTModel (OpenAI GPT model) - `xlnet`: TFXLNetModel (XLNet model)
- contains `gpt2`: TFGPT2Model (OpenAI GPT-2 model) - `xlm`: TFXLMModel (XLM model)
- contains `transfo-xl`: TFTransfoXLModel (Transformer-XL model) - `ctrl`: TFCTRLModel (CTRL model)
- contains `xlnet`: TFXLNetModel (XLNet model)
- contains `xlm`: TFXLMModel (XLM model)
- contains `ctrl`: TFCTRLModel (CTRL model)
This class cannot be instantiated using `__init__()` (throws an error). This class cannot be instantiated using `__init__()` (throws an error).
""" """
...@@ -276,17 +245,18 @@ class TFAutoModel(object): ...@@ -276,17 +245,18 @@ class TFAutoModel(object):
r""" Instantiates one of the base model classes of the library r""" Instantiates one of the base model classes of the library
from a pre-trained model configuration. from a pre-trained model configuration.
The model class to instantiate is selected as the first pattern matching The `from_pretrained()` method takes care of returning the correct model class instance
in the `pretrained_model_name_or_path` string (in the following order): based on the `model_type` property of the config object, or when it's missing,
- contains `t5`: TFT5Model (T5 model) falling back to using pattern matching on the `pretrained_model_name_or_path` string:
- contains `distilbert`: TFDistilBertModel (DistilBERT model) - `t5`: TFT5Model (T5 model)
- contains `roberta`: TFRobertaModel (RoBERTa model) - `distilbert`: TFDistilBertModel (DistilBERT model)
- contains `bert`: TFTFBertModel (Bert model) - `roberta`: TFRobertaModel (RoBERTa model)
- contains `openai-gpt`: TFOpenAIGPTModel (OpenAI GPT model) - `bert`: TFTFBertModel (Bert model)
- contains `gpt2`: TFGPT2Model (OpenAI GPT-2 model) - `openai-gpt`: TFOpenAIGPTModel (OpenAI GPT model)
- contains `transfo-xl`: TFTransfoXLModel (Transformer-XL model) - `gpt2`: TFGPT2Model (OpenAI GPT-2 model)
- contains `xlnet`: TFXLNetModel (XLNet model) - `transfo-xl`: TFTransfoXLModel (Transformer-XL model)
- contains `ctrl`: TFCTRLModel (CTRL model) - `xlnet`: TFXLNetModel (XLNet model)
- `ctrl`: TFCTRLModel (CTRL model)
Params: Params:
pretrained_model_name_or_path: either: pretrained_model_name_or_path: either:
...@@ -424,21 +394,18 @@ class TFAutoModelForPreTraining(object): ...@@ -424,21 +394,18 @@ class TFAutoModelForPreTraining(object):
The `from_pretrained()` method takes care of returning the correct model class instance The `from_pretrained()` method takes care of returning the correct model class instance
based on the `model_type` property of the config object, or when it's missing, based on the `model_type` property of the config object, or when it's missing,
falling back to using pattern matching on the `pretrained_model_name_or_path` string. falling back to using pattern matching on the `pretrained_model_name_or_path` string:
- `t5`: :class:`~transformers.TFT5ModelWithLMHead` (T5 model)
The model class to instantiate is selected as the first pattern matching - `distilbert`: :class:`~transformers.TFDistilBertForMaskedLM` (DistilBERT model)
in the `pretrained_model_name_or_path` string (in the following order): - `albert`: :class:`~transformers.TFAlbertForPreTraining` (ALBERT model)
- contains `t5`: :class:`~transformers.TFT5ModelWithLMHead` (T5 model) - `roberta`: :class:`~transformers.TFRobertaForMaskedLM` (RoBERTa model)
- contains `distilbert`: :class:`~transformers.TFDistilBertForMaskedLM` (DistilBERT model) - `bert`: :class:`~transformers.TFBertForPreTraining` (Bert model)
- contains `albert`: :class:`~transformers.TFAlbertForPreTraining` (ALBERT model) - `openai-gpt`: :class:`~transformers.TFOpenAIGPTLMHeadModel` (OpenAI GPT model)
- contains `roberta`: :class:`~transformers.TFRobertaForMaskedLM` (RoBERTa model) - `gpt2`: :class:`~transformers.TFGPT2LMHeadModel` (OpenAI GPT-2 model)
- contains `bert`: :class:`~transformers.TFBertForPreTraining` (Bert model) - `transfo-xl`: :class:`~transformers.TFTransfoXLLMHeadModel` (Transformer-XL model)
- contains `openai-gpt`: :class:`~transformers.TFOpenAIGPTLMHeadModel` (OpenAI GPT model) - `xlnet`: :class:`~transformers.TFXLNetLMHeadModel` (XLNet model)
- contains `gpt2`: :class:`~transformers.TFGPT2LMHeadModel` (OpenAI GPT-2 model) - `xlm`: :class:`~transformers.TFXLMWithLMHeadModel` (XLM model)
- contains `transfo-xl`: :class:`~transformers.TFTransfoXLLMHeadModel` (Transformer-XL model) - `ctrl`: :class:`~transformers.TFCTRLLMHeadModel` (Salesforce CTRL model)
- contains `xlnet`: :class:`~transformers.TFXLNetLMHeadModel` (XLNet model)
- contains `xlm`: :class:`~transformers.TFXLMWithLMHeadModel` (XLM model)
- contains `ctrl`: :class:`~transformers.TFCTRLLMHeadModel` (Salesforce CTRL model)
The model is set in evaluation mode by default using `model.eval()` (Dropout modules are deactivated) The model is set in evaluation mode by default using `model.eval()` (Dropout modules are deactivated)
To train the model, you should first set it back in training mode with `model.train()` To train the model, you should first set it back in training mode with `model.train()`
...@@ -525,20 +492,17 @@ class TFAutoModelWithLMHead(object): ...@@ -525,20 +492,17 @@ class TFAutoModelWithLMHead(object):
The `from_pretrained()` method takes care of returning the correct model class instance The `from_pretrained()` method takes care of returning the correct model class instance
based on the `model_type` property of the config object, or when it's missing, based on the `model_type` property of the config object, or when it's missing,
falling back to using pattern matching on the `pretrained_model_name_or_path` string. falling back to using pattern matching on the `pretrained_model_name_or_path` string:
- `t5`: TFT5ForConditionalGeneration (T5 model)
The model class to instantiate is selected as the first pattern matching - `distilbert`: TFDistilBertForMaskedLM (DistilBERT model)
in the `pretrained_model_name_or_path` string (in the following order): - `roberta`: TFRobertaForMaskedLM (RoBERTa model)
- contains `t5`: TFT5ForConditionalGeneration (T5 model) - `bert`: TFBertForMaskedLM (Bert model)
- contains `distilbert`: TFDistilBertForMaskedLM (DistilBERT model) - `openai-gpt`: TFOpenAIGPTLMHeadModel (OpenAI GPT model)
- contains `roberta`: TFRobertaForMaskedLM (RoBERTa model) - `gpt2`: TFGPT2LMHeadModel (OpenAI GPT-2 model)
- contains `bert`: TFBertForMaskedLM (Bert model) - `transfo-xl`: TFTransfoXLLMHeadModel (Transformer-XL model)
- contains `openai-gpt`: TFOpenAIGPTLMHeadModel (OpenAI GPT model) - `xlnet`: TFXLNetLMHeadModel (XLNet model)
- contains `gpt2`: TFGPT2LMHeadModel (OpenAI GPT-2 model) - `xlm`: TFXLMWithLMHeadModel (XLM model)
- contains `transfo-xl`: TFTransfoXLLMHeadModel (Transformer-XL model) - `ctrl`: TFCTRLLMHeadModel (CTRL model)
- contains `xlnet`: TFXLNetLMHeadModel (XLNet model)
- contains `xlm`: TFXLMWithLMHeadModel (XLM model)
- contains `ctrl`: TFCTRLLMHeadModel (CTRL model)
This class cannot be instantiated using `__init__()` (throws an error). This class cannot be instantiated using `__init__()` (throws an error).
""" """
...@@ -595,20 +559,17 @@ class TFAutoModelWithLMHead(object): ...@@ -595,20 +559,17 @@ class TFAutoModelWithLMHead(object):
The `from_pretrained()` method takes care of returning the correct model class instance The `from_pretrained()` method takes care of returning the correct model class instance
based on the `model_type` property of the config object, or when it's missing, based on the `model_type` property of the config object, or when it's missing,
falling back to using pattern matching on the `pretrained_model_name_or_path` string. falling back to using pattern matching on the `pretrained_model_name_or_path` string:
- `t5`: TFT5ForConditionalGeneration (T5 model)
The model class to instantiate is selected as the first pattern matching - `distilbert`: TFDistilBertForMaskedLM (DistilBERT model)
in the `pretrained_model_name_or_path` string (in the following order): - `roberta`: TFRobertaForMaskedLM (RoBERTa model)
- contains `t5`: TFT5ForConditionalGeneration (T5 model) - `bert`: TFBertForMaskedLM (Bert model)
- contains `distilbert`: TFDistilBertForMaskedLM (DistilBERT model) - `openai-gpt`: TFOpenAIGPTLMHeadModel (OpenAI GPT model)
- contains `roberta`: TFRobertaForMaskedLM (RoBERTa model) - `gpt2`: TFGPT2LMHeadModel (OpenAI GPT-2 model)
- contains `bert`: TFBertForMaskedLM (Bert model) - `transfo-xl`: TFTransfoXLLMHeadModel (Transformer-XL model)
- contains `openai-gpt`: TFOpenAIGPTLMHeadModel (OpenAI GPT model) - `xlnet`: TFXLNetLMHeadModel (XLNet model)
- contains `gpt2`: TFGPT2LMHeadModel (OpenAI GPT-2 model) - `xlm`: TFXLMWithLMHeadModel (XLM model)
- contains `transfo-xl`: TFTransfoXLLMHeadModel (Transformer-XL model) - `ctrl`: TFCTRLLMHeadModel (CTRL model)
- contains `xlnet`: TFXLNetLMHeadModel (XLNet model)
- contains `xlm`: TFXLMWithLMHeadModel (XLM model)
- contains `ctrl`: TFCTRLLMHeadModel (CTRL model)
Params: Params:
pretrained_model_name_or_path: either: pretrained_model_name_or_path: either:
...@@ -694,12 +655,9 @@ class TFAutoModelForMultipleChoice: ...@@ -694,12 +655,9 @@ class TFAutoModelForMultipleChoice:
The `from_pretrained()` method takes care of returning the correct model class instance The `from_pretrained()` method takes care of returning the correct model class instance
based on the `model_type` property of the config object, or when it's missing, based on the `model_type` property of the config object, or when it's missing,
falling back to using pattern matching on the `pretrained_model_name_or_path` string. falling back to using pattern matching on the `pretrained_model_name_or_path` string:
- `albert`: TFAlbertForMultipleChoice (Albert model)
The model class to instantiate is selected as the first pattern matching - `bert`: TFBertForMultipleChoice (Bert model)
in the `pretrained_model_name_or_path` string (in the following order):
- contains `albert`: TFAlbertForMultipleChoice (Albert model)
- contains `bert`: TFBertForMultipleChoice (Bert model)
This class cannot be instantiated using `__init__()` (throws an error). This class cannot be instantiated using `__init__()` (throws an error).
""" """
...@@ -751,12 +709,9 @@ class TFAutoModelForMultipleChoice: ...@@ -751,12 +709,9 @@ class TFAutoModelForMultipleChoice:
The `from_pretrained()` method takes care of returning the correct model class instance The `from_pretrained()` method takes care of returning the correct model class instance
based on the `model_type` property of the config object, or when it's missing, based on the `model_type` property of the config object, or when it's missing,
falling back to using pattern matching on the `pretrained_model_name_or_path` string. falling back to using pattern matching on the `pretrained_model_name_or_path` string:
- `albert`: TFRobertaForMultiple (Albert model)
The model class to instantiate is selected as the first pattern matching - `bert`: TFBertForMultipleChoice (Bert model)
in the `pretrained_model_name_or_path` string (in the following order):
- contains `albert`: TFRobertaForMultiple (Albert model)
- contains `bert`: TFBertForMultipleChoice (Bert model)
The model is set in evaluation mode by default using `model.eval()` (Dropout modules are deactivated) The model is set in evaluation mode by default using `model.eval()` (Dropout modules are deactivated)
To train the model, you should first set it back in training mode with `model.train()` To train the model, you should first set it back in training mode with `model.train()`
...@@ -847,15 +802,12 @@ class TFAutoModelForSequenceClassification(object): ...@@ -847,15 +802,12 @@ class TFAutoModelForSequenceClassification(object):
The `from_pretrained()` method takes care of returning the correct model class instance The `from_pretrained()` method takes care of returning the correct model class instance
based on the `model_type` property of the config object, or when it's missing, based on the `model_type` property of the config object, or when it's missing,
falling back to using pattern matching on the `pretrained_model_name_or_path` string. falling back to using pattern matching on the `pretrained_model_name_or_path` string:
- `distilbert`: TFDistilBertForSequenceClassification (DistilBERT model)
The model class to instantiate is selected as the first pattern matching - `roberta`: TFRobertaForSequenceClassification (RoBERTa model)
in the `pretrained_model_name_or_path` string (in the following order): - `bert`: TFBertForSequenceClassification (Bert model)
- contains `distilbert`: TFDistilBertForSequenceClassification (DistilBERT model) - `xlnet`: TFXLNetForSequenceClassification (XLNet model)
- contains `roberta`: TFRobertaForSequenceClassification (RoBERTa model) - `xlm`: TFXLMForSequenceClassification (XLM model)
- contains `bert`: TFBertForSequenceClassification (Bert model)
- contains `xlnet`: TFXLNetForSequenceClassification (XLNet model)
- contains `xlm`: TFXLMForSequenceClassification (XLM model)
This class cannot be instantiated using `__init__()` (throws an error). This class cannot be instantiated using `__init__()` (throws an error).
""" """
...@@ -910,15 +862,12 @@ class TFAutoModelForSequenceClassification(object): ...@@ -910,15 +862,12 @@ class TFAutoModelForSequenceClassification(object):
The `from_pretrained()` method takes care of returning the correct model class instance The `from_pretrained()` method takes care of returning the correct model class instance
based on the `model_type` property of the config object, or when it's missing, based on the `model_type` property of the config object, or when it's missing,
falling back to using pattern matching on the `pretrained_model_name_or_path` string. falling back to using pattern matching on the `pretrained_model_name_or_path` string:
- `distilbert`: TFDistilBertForSequenceClassification (DistilBERT model)
The model class to instantiate is selected as the first pattern matching - `roberta`: TFRobertaForSequenceClassification (RoBERTa model)
in the `pretrained_model_name_or_path` string (in the following order): - `bert`: TFBertForSequenceClassification (Bert model)
- contains `distilbert`: TFDistilBertForSequenceClassification (DistilBERT model) - `xlnet`: TFXLNetForSequenceClassification (XLNet model)
- contains `roberta`: TFRobertaForSequenceClassification (RoBERTa model) - `xlm`: TFXLMForSequenceClassification (XLM model)
- contains `bert`: TFBertForSequenceClassification (Bert model)
- contains `xlnet`: TFXLNetForSequenceClassification (XLNet model)
- contains `xlm`: TFXLMForSequenceClassification (XLM model)
The model is set in evaluation mode by default using `model.eval()` (Dropout modules are deactivated) The model is set in evaluation mode by default using `model.eval()` (Dropout modules are deactivated)
To train the model, you should first set it back in training mode with `model.train()` To train the model, you should first set it back in training mode with `model.train()`
...@@ -1009,16 +958,13 @@ class TFAutoModelForQuestionAnswering(object): ...@@ -1009,16 +958,13 @@ class TFAutoModelForQuestionAnswering(object):
The `from_pretrained()` method takes care of returning the correct model class instance The `from_pretrained()` method takes care of returning the correct model class instance
based on the `model_type` property of the config object, or when it's missing, based on the `model_type` property of the config object, or when it's missing,
falling back to using pattern matching on the `pretrained_model_name_or_path` string. falling back to using pattern matching on the `pretrained_model_name_or_path` string:
- `distilbert`: TFDistilBertForQuestionAnswering (DistilBERT model)
The model class to instantiate is selected as the first pattern matching - `albert`: TFAlbertForQuestionAnswering (ALBERT model)
in the `pretrained_model_name_or_path` string (in the following order): - `roberta`: TFRobertaForQuestionAnswering (RoBERTa model)
- contains `distilbert`: TFDistilBertForQuestionAnswering (DistilBERT model) - `bert`: TFBertForQuestionAnswering (Bert model)
- contains `albert`: TFAlbertForQuestionAnswering (ALBERT model) - `xlnet`: TFXLNetForQuestionAnswering (XLNet model)
- contains `roberta`: TFRobertaForQuestionAnswering (RoBERTa model) - `xlm`: TFXLMForQuestionAnswering (XLM model)
- contains `bert`: TFBertForQuestionAnswering (Bert model)
- contains `xlnet`: TFXLNetForQuestionAnswering (XLNet model)
- contains `xlm`: TFXLMForQuestionAnswering (XLM model)
This class cannot be instantiated using `__init__()` (throws an error). This class cannot be instantiated using `__init__()` (throws an error).
""" """
...@@ -1074,16 +1020,13 @@ class TFAutoModelForQuestionAnswering(object): ...@@ -1074,16 +1020,13 @@ class TFAutoModelForQuestionAnswering(object):
The `from_pretrained()` method takes care of returning the correct model class instance The `from_pretrained()` method takes care of returning the correct model class instance
based on the `model_type` property of the config object, or when it's missing, based on the `model_type` property of the config object, or when it's missing,
falling back to using pattern matching on the `pretrained_model_name_or_path` string. falling back to using pattern matching on the `pretrained_model_name_or_path` string:
- `distilbert`: TFDistilBertForQuestionAnswering (DistilBERT model)
The model class to instantiate is selected as the first pattern matching - `albert`: TFAlbertForQuestionAnswering (ALBERT model)
in the `pretrained_model_name_or_path` string (in the following order): - `roberta`: TFRobertaForQuestionAnswering (RoBERTa model)
- contains `distilbert`: TFDistilBertForQuestionAnswering (DistilBERT model) - `bert`: TFBertForQuestionAnswering (Bert model)
- contains `albert`: TFAlbertForQuestionAnswering (ALBERT model) - `xlnet`: TFXLNetForQuestionAnswering (XLNet model)
- contains `roberta`: TFRobertaForQuestionAnswering (RoBERTa model) - `xlm`: TFXLMForQuestionAnswering (XLM model)
- contains `bert`: TFBertForQuestionAnswering (Bert model)
- contains `xlnet`: TFXLNetForQuestionAnswering (XLNet model)
- contains `xlm`: TFXLMForQuestionAnswering (XLM model)
The model is set in evaluation mode by default using `model.eval()` (Dropout modules are deactivated) The model is set in evaluation mode by default using `model.eval()` (Dropout modules are deactivated)
To train the model, you should first set it back in training mode with `model.train()` To train the model, you should first set it back in training mode with `model.train()`
...@@ -1215,14 +1158,11 @@ class TFAutoModelForTokenClassification: ...@@ -1215,14 +1158,11 @@ class TFAutoModelForTokenClassification:
The `from_pretrained()` method takes care of returning the correct model class instance The `from_pretrained()` method takes care of returning the correct model class instance
based on the `model_type` property of the config object, or when it's missing, based on the `model_type` property of the config object, or when it's missing,
falling back to using pattern matching on the `pretrained_model_name_or_path` string. falling back to using pattern matching on the `pretrained_model_name_or_path` string:
- `bert`: BertForTokenClassification (Bert model)
The model class to instantiate is selected as the first pattern matching - `xlnet`: XLNetForTokenClassification (XLNet model)
in the `pretrained_model_name_or_path` string (in the following order): - `distilbert`: DistilBertForTokenClassification (DistilBert model)
- contains `bert`: BertForTokenClassification (Bert model) - `roberta`: RobertaForTokenClassification (Roberta model)
- contains `xlnet`: XLNetForTokenClassification (XLNet model)
- contains `distilbert`: DistilBertForTokenClassification (DistilBert model)
- contains `roberta`: RobertaForTokenClassification (Roberta model)
The model is set in evaluation mode by default using `model.eval()` (Dropout modules are deactivated) The model is set in evaluation mode by default using `model.eval()` (Dropout modules are deactivated)
To train the model, you should first set it back in training mode with `model.train()` To train the model, you should first set it back in training mode with `model.train()`
......
src/transformers/modeling_tf_bert.py
View file @ d4c2cb40
...@@ -30,28 +30,29 @@ from .tokenization_utils import BatchEncoding ...@@ -30,28 +30,29 @@ from .tokenization_utils import BatchEncoding
logger = logging.getLogger(__name__) logger = logging.getLogger(__name__)
TF_BERT_PRETRAINED_MODEL_ARCHIVE_MAP = { TF_BERT_PRETRAINED_MODEL_ARCHIVE_LIST = [
"bert-base-uncased": "https://cdn.huggingface.co/bert-base-uncased-tf_model.h5", "bert-base-uncased",
"bert-large-uncased": "https://cdn.huggingface.co/bert-large-uncased-tf_model.h5", "bert-large-uncased",
"bert-base-cased": "https://cdn.huggingface.co/bert-base-cased-tf_model.h5", "bert-base-cased",
"bert-large-cased": "https://cdn.huggingface.co/bert-large-cased-tf_model.h5", "bert-large-cased",
"bert-base-multilingual-uncased": "https://cdn.huggingface.co/bert-base-multilingual-uncased-tf_model.h5", "bert-base-multilingual-uncased",
"bert-base-multilingual-cased": "https://cdn.huggingface.co/bert-base-multilingual-cased-tf_model.h5", "bert-base-multilingual-cased",
"bert-base-chinese": "https://cdn.huggingface.co/bert-base-chinese-tf_model.h5", "bert-base-chinese",
"bert-base-german-cased": "https://cdn.huggingface.co/bert-base-german-cased-tf_model.h5", "bert-base-german-cased",
"bert-large-uncased-whole-word-masking": "https://cdn.huggingface.co/bert-large-uncased-whole-word-masking-tf_model.h5", "bert-large-uncased-whole-word-masking",
"bert-large-cased-whole-word-masking": "https://cdn.huggingface.co/bert-large-cased-whole-word-masking-tf_model.h5", "bert-large-cased-whole-word-masking",
"bert-large-uncased-whole-word-masking-finetuned-squad": "https://cdn.huggingface.co/bert-large-uncased-whole-word-masking-finetuned-squad-tf_model.h5", "bert-large-uncased-whole-word-masking-finetuned-squad",
"bert-large-cased-whole-word-masking-finetuned-squad": "https://cdn.huggingface.co/bert-large-cased-whole-word-masking-finetuned-squad-tf_model.h5", "bert-large-cased-whole-word-masking-finetuned-squad",
"bert-base-cased-finetuned-mrpc": "https://cdn.huggingface.co/bert-base-cased-finetuned-mrpc-tf_model.h5", "bert-base-cased-finetuned-mrpc",
"bert-base-japanese": "https://cdn.huggingface.co/cl-tohoku/bert-base-japanese/tf_model.h5", "cl-tohoku/bert-base-japanese",
"bert-base-japanese-whole-word-masking": "https://cdn.huggingface.co/cl-tohoku/bert-base-japanese-whole-word-masking/tf_model.h5", "cl-tohoku/bert-base-japanese-whole-word-masking",
"bert-base-japanese-char": "https://cdn.huggingface.co/cl-tohoku/bert-base-japanese-char/tf_model.h5", "cl-tohoku/bert-base-japanese-char",
"bert-base-japanese-char-whole-word-masking": "https://cdn.huggingface.co/cl-tohoku/bert-base-japanese-char-whole-word-masking/tf_model.h5", "cl-tohoku/bert-base-japanese-char-whole-word-masking",
"bert-base-finnish-cased-v1": "https://cdn.huggingface.co/TurkuNLP/bert-base-finnish-cased-v1/tf_model.h5", "TurkuNLP/bert-base-finnish-cased-v1",
"bert-base-finnish-uncased-v1": "https://cdn.huggingface.co/TurkuNLP/bert-base-finnish-uncased-v1/tf_model.h5", "TurkuNLP/bert-base-finnish-uncased-v1",
"bert-base-dutch-cased": "https://cdn.huggingface.co/wietsedv/bert-base-dutch-cased/tf_model.h5", "wietsedv/bert-base-dutch-cased",
} # See all BERT models at https://huggingface.co/models?filter=bert
]
def gelu(x): def gelu(x):
...@@ -585,7 +586,6 @@ class TFBertPreTrainedModel(TFPreTrainedModel): ...@@ -585,7 +586,6 @@ class TFBertPreTrainedModel(TFPreTrainedModel):
""" """
config_class = BertConfig config_class = BertConfig
pretrained_model_archive_map = TF_BERT_PRETRAINED_MODEL_ARCHIVE_MAP
base_model_prefix = "bert" base_model_prefix = "bert"
......
src/transformers/modeling_tf_camembert.py
View file @ d4c2cb40
...@@ -30,7 +30,9 @@ from .modeling_tf_roberta import ( ...@@ -30,7 +30,9 @@ from .modeling_tf_roberta import (
logger = logging.getLogger(__name__) logger = logging.getLogger(__name__)
TF_CAMEMBERT_PRETRAINED_MODEL_ARCHIVE_MAP = {} TF_CAMEMBERT_PRETRAINED_MODEL_ARCHIVE_LIST = [
# See all CamemBERT models at https://huggingface.co/models?filter=camembert
]
CAMEMBERT_START_DOCSTRING = r""" CAMEMBERT_START_DOCSTRING = r"""
...@@ -72,7 +74,6 @@ class TFCamembertModel(TFRobertaModel): ...@@ -72,7 +74,6 @@ class TFCamembertModel(TFRobertaModel):
""" """
config_class = CamembertConfig config_class = CamembertConfig
pretrained_model_archive_map = TF_CAMEMBERT_PRETRAINED_MODEL_ARCHIVE_MAP
@add_start_docstrings( @add_start_docstrings(
...@@ -85,7 +86,6 @@ class TFCamembertForMaskedLM(TFRobertaForMaskedLM): ...@@ -85,7 +86,6 @@ class TFCamembertForMaskedLM(TFRobertaForMaskedLM):
""" """
config_class = CamembertConfig config_class = CamembertConfig
pretrained_model_archive_map = TF_CAMEMBERT_PRETRAINED_MODEL_ARCHIVE_MAP
@add_start_docstrings( @add_start_docstrings(
...@@ -100,7 +100,6 @@ class TFCamembertForSequenceClassification(TFRobertaForSequenceClassification): ...@@ -100,7 +100,6 @@ class TFCamembertForSequenceClassification(TFRobertaForSequenceClassification):
""" """
config_class = CamembertConfig config_class = CamembertConfig
pretrained_model_archive_map = TF_CAMEMBERT_PRETRAINED_MODEL_ARCHIVE_MAP
@add_start_docstrings( @add_start_docstrings(
...@@ -115,4 +114,3 @@ class TFCamembertForTokenClassification(TFRobertaForTokenClassification): ...@@ -115,4 +114,3 @@ class TFCamembertForTokenClassification(TFRobertaForTokenClassification):
""" """
config_class = CamembertConfig config_class = CamembertConfig
pretrained_model_archive_map = TF_CAMEMBERT_PRETRAINED_MODEL_ARCHIVE_MAP
src/transformers/modeling_tf_ctrl.py
View file @ d4c2cb40
...@@ -29,7 +29,10 @@ from .tokenization_utils import BatchEncoding ...@@ -29,7 +29,10 @@ from .tokenization_utils import BatchEncoding
logger = logging.getLogger(__name__) logger = logging.getLogger(__name__)
TF_CTRL_PRETRAINED_MODEL_ARCHIVE_MAP = {"ctrl": "https://cdn.huggingface.co/ctrl-tf_model.h5"} TF_CTRL_PRETRAINED_MODEL_ARCHIVE_LIST = [
"ctrl"
# See all CTRL models at https://huggingface.co/models?filter=ctrl
]
def angle_defn(pos, i, d_model_size): def angle_defn(pos, i, d_model_size):
...@@ -379,7 +382,6 @@ class TFCTRLPreTrainedModel(TFPreTrainedModel): ...@@ -379,7 +382,6 @@ class TFCTRLPreTrainedModel(TFPreTrainedModel):
""" """
config_class = CTRLConfig config_class = CTRLConfig
pretrained_model_archive_map = TF_CTRL_PRETRAINED_MODEL_ARCHIVE_MAP
base_model_prefix = "transformer" base_model_prefix = "transformer"
......
src/transformers/modeling_tf_distilbert.py
View file @ d4c2cb40
...@@ -31,14 +31,15 @@ from .tokenization_utils import BatchEncoding ...@@ -31,14 +31,15 @@ from .tokenization_utils import BatchEncoding
logger = logging.getLogger(__name__) logger = logging.getLogger(__name__)
TF_DISTILBERT_PRETRAINED_MODEL_ARCHIVE_MAP = { TF_DISTILBERT_PRETRAINED_MODEL_ARCHIVE_LIST = [
"distilbert-base-uncased": "https://cdn.huggingface.co/distilbert-base-uncased-tf_model.h5", "distilbert-base-uncased",
"distilbert-base-uncased-distilled-squad": "https://cdn.huggingface.co/distilbert-base-uncased-distilled-squad-tf_model.h5", "distilbert-base-uncased-distilled-squad",
"distilbert-base-cased": "https://cdn.huggingface.co/distilbert-base-cased-tf_model.h5", "distilbert-base-cased",
"distilbert-base-cased-distilled-squad": "https://cdn.huggingface.co/distilbert-base-cased-distilled-squad-tf_model.h5", "distilbert-base-cased-distilled-squad",
"distilbert-base-multilingual-cased": "https://cdn.huggingface.co/distilbert-base-multilingual-cased-tf_model.h5", "distilbert-base-multilingual-cased",
"distilbert-base-uncased-finetuned-sst-2-english": "https://cdn.huggingface.co/distilbert-base-uncased-finetuned-sst-2-english-tf_model.h5", "distilbert-base-uncased-finetuned-sst-2-english",
} # See all DistilBERT models at https://huggingface.co/models?filter=distilbert
]
# UTILS AND BUILDING BLOCKS OF THE ARCHITECTURE # # UTILS AND BUILDING BLOCKS OF THE ARCHITECTURE #
...@@ -467,7 +468,6 @@ class TFDistilBertPreTrainedModel(TFPreTrainedModel): ...@@ -467,7 +468,6 @@ class TFDistilBertPreTrainedModel(TFPreTrainedModel):
""" """
config_class = DistilBertConfig config_class = DistilBertConfig
pretrained_model_archive_map = TF_DISTILBERT_PRETRAINED_MODEL_ARCHIVE_MAP
base_model_prefix = "distilbert" base_model_prefix = "distilbert"
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment