Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Update RoBERTa vocabulary files #255

Merged
merged 6 commits into from
Nov 30, 2019
Merged

Update RoBERTa vocabulary files #255

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
35890c2
Update RoBERTa vocabulary files
gpengzhi Nov 27, 2019
7062f01
Update docs/requirements.txt
gpengzhi Nov 27, 2019
6f1aafd
Update .codecov.yml
gpengzhi Nov 28, 2019
5ff64a7
Disable codecov/patch
gpengzhi Nov 29, 2019
2c63646
Update .codecov.yml and docs/requirements
gpengzhi Nov 30, 2019
646de51
Update Pygments version
gpengzhi Nov 30, 2019
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
28 changes: 27 additions & 1 deletion texar/torch/data/tokenizers/bert_tokenizer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -76,6 +76,30 @@ class BERTTokenizer(PretrainedBERTMixin, TokenizerBase):
'scibert-basevocab-cased': 512,
}
_VOCAB_FILE_NAMES = {'vocab_file': 'vocab.txt'}
_VOCAB_FILE_MAP = {
'vocab_file': {
# Standard BERT
'bert-base-uncased': 'vocab.txt',
'bert-large-uncased': 'vocab.txt',
'bert-base-cased': 'vocab.txt',
'bert-large-cased': 'vocab.txt',
'bert-base-multilingual-uncased': 'vocab.txt',
'bert-base-multilingual-cased': 'vocab.txt',
'bert-base-chinese': 'vocab.txt',

# BioBERT
'biobert-v1.0-pmc': 'vocab.txt',
'biobert-v1.0-pubmed-pmc': 'vocab.txt',
'biobert-v1.0-pubmed': 'vocab.txt',
'biobert-v1.1-pubmed': 'vocab.txt',

# SciBERT
'scibert-scivocab-uncased': 'vocab.txt',
'scibert-scivocab-cased': 'vocab.txt',
'scibert-basevocab-uncased': 'vocab.txt',
'scibert-basevocab-cased': 'vocab.txt',
}
}

def __init__(self,
pretrained_model_name: Optional[str] = None,
Expand All @@ -93,8 +117,10 @@ def __init__(self,
}

if self.pretrained_model_dir is not None:
assert self.pretrained_model_name is not None
vocab_file = os.path.join(self.pretrained_model_dir,
self._VOCAB_FILE_NAMES['vocab_file'])
self._VOCAB_FILE_MAP['vocab_file']
[self.pretrained_model_name])
assert self.pretrained_model_name is not None
if self._MAX_INPUT_SIZE.get(self.pretrained_model_name):
self.max_len = self._MAX_INPUT_SIZE[self.pretrained_model_name]
Expand Down
35 changes: 30 additions & 5 deletions texar/torch/data/tokenizers/gpt2_tokenizer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -70,6 +70,24 @@ class GPT2Tokenizer(TokenizerBase, PretrainedGPT2Mixin):
'vocab_file': 'encoder.json',
'merges_file': 'vocab.bpe',
}
_VOCAB_FILE_MAP = {
'vocab_file': {
'gpt2-small': 'encoder.json',
'gpt2-medium': 'encoder.json',
'gpt2-large': 'encoder.json',
'gpt2-xl': 'encoder.json',
'117M': 'encoder.json',
'345M': 'encoder.json',
},
'merges_file': {
gpengzhi marked this conversation as resolved.
Show resolved Hide resolved
'gpt2-small': 'vocab.bpe',
'gpt2-medium': 'vocab.bpe',
'gpt2-large': 'vocab.bpe',
'gpt2-xl': 'vocab.bpe',
'117M': 'vocab.bpe',
'345M': 'vocab.bpe',
},
}

def __init__(self,
pretrained_model_name: Optional[str] = None,
Expand All @@ -84,10 +102,13 @@ def __init__(self,
}

if self.pretrained_model_dir is not None:
assert self.pretrained_model_name is not None
vocab_file = os.path.join(self.pretrained_model_dir,
self._VOCAB_FILE_NAMES['vocab_file'])
self._VOCAB_FILE_MAP['vocab_file']
[self.pretrained_model_name])
merges_file = os.path.join(self.pretrained_model_dir,
self._VOCAB_FILE_NAMES['merges_file'])
self._VOCAB_FILE_MAP['merges_file']
[self.pretrained_model_name])
assert pretrained_model_name is not None
if self._MAX_INPUT_SIZE.get(pretrained_model_name):
self.max_len = self._MAX_INPUT_SIZE[pretrained_model_name]
Expand Down Expand Up @@ -119,9 +140,8 @@ def __init__(self,

# Should haved added re.IGNORECASE so BPE merges can happen for
# capitalized versions of contractions
self.pat = re.compile(
r"""'s|'t|'re|'ve|'m|'ll|'d| ?\p{L}+| ?\p{N}+| ?
[^\s\p{L}\p{N}]+|\s+(?!\S)|\s+""")
self.pat = re.compile(r"""'s|'t|'re|'ve|'m|'ll|'d| ?\p{L}+| """ +
r""""?\p{N}+| ?[^\s\p{L}\p{N}]+|\s+(?!\S)|\s+""")

def _map_text_to_token(self, text: str) -> List[str]: # type: ignore
r"""Tokenize a string. """
Expand Down Expand Up @@ -301,6 +321,7 @@ def default_hparams() -> Dict[str, Any]:
"unk_token": "<|endoftext|>",
"pad_token": "<|endoftext|>",
"errors": "replace",
"name": "gpt2_tokenizer",
}

Here:
Expand Down Expand Up @@ -332,6 +353,9 @@ def default_hparams() -> Dict[str, Any]:
`"errors"`: str
Response when mapping tokens to text fails. The possible values are
`ignore`, `replace`, and `strict`.

`"name"`: str
Name of the tokenizer.
"""
return {
'pretrained_model_name': '117M',
Expand All @@ -343,6 +367,7 @@ def default_hparams() -> Dict[str, Any]:
'unk_token': '<|endoftext|>',
'pad_token': '<|endoftext|>',
'errors': 'replace',
'name': 'gpt2_tokenizer',
'@no_typecheck': ['pretrained_model_name'],
}

Expand Down
28 changes: 24 additions & 4 deletions texar/torch/data/tokenizers/roberta_tokenizer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -24,8 +24,7 @@
'RoBERTaTokenizer',
]

_GPT2_PATH = "https://dl.fbaipublicfiles.com/fairseq/gpt2_bpe/"
_CHECKPOINT_FILES = ["encoder.json", "vocab.bpe"]
_ROBERTA_PATH = "https://s3.amazonaws.com/models.huggingface.co/bert/"
gpengzhi marked this conversation as resolved.
Show resolved Hide resolved


class RoBERTaTokenizer(GPT2Tokenizer):
Expand All @@ -48,13 +47,29 @@ class RoBERTaTokenizer(GPT2Tokenizer):
"""

_MODEL2URL = {
'roberta-base': [_GPT2_PATH + f"{file}" for file in _CHECKPOINT_FILES],
'roberta-large': [_GPT2_PATH + f"{file}" for file in _CHECKPOINT_FILES],
'roberta-base': [
_ROBERTA_PATH + 'roberta-base-vocab.json',
_ROBERTA_PATH + 'roberta-base-merges.txt',
],
'roberta-large': [
_ROBERTA_PATH + 'roberta-large-vocab.json',
_ROBERTA_PATH + 'roberta-large-merges.txt',
],
}
_MAX_INPUT_SIZE = {
'roberta-base': 512,
'roberta-large': 512,
}
_VOCAB_FILE_MAP = {
'vocab_file': {
'roberta-base': 'roberta-base-vocab.json',
'roberta-large': 'roberta-large-vocab.json',
},
'merges_file': {
'roberta-base': 'roberta-base-merges.txt',
'roberta-large': 'roberta-large-merges.txt',
},
}

def encode_text(self, # type: ignore
text_a: str,
Expand Down Expand Up @@ -153,6 +168,7 @@ def default_hparams() -> Dict[str, Any]:
"pad_token": "<pad>",
"mask_token": "<mask>",
"errors": "replace",
"name": "roberta_tokenizer",
}

Here:
Expand Down Expand Up @@ -193,6 +209,9 @@ def default_hparams() -> Dict[str, Any]:
`"errors"`: str
Response when decoding fails. The possible values are
`ignore`, `replace`, and `strict`.

`"name"`: str
Name of the tokenizer.
"""
return {
'pretrained_model_name': 'roberta-base',
Expand All @@ -207,6 +226,7 @@ def default_hparams() -> Dict[str, Any]:
'pad_token': '<pad>',
'mask_token': '<mask>',
'errors': 'replace',
'name': 'roberta_tokenizer',
'@no_typecheck': ['pretrained_model_name'],
}

Expand Down
1 change: 1 addition & 0 deletions texar/torch/data/tokenizers/tokenizer_base.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -53,6 +53,7 @@ class TokenizerBase(ModuleBase):
_IS_PRETRAINED: bool
_MAX_INPUT_SIZE: Dict[str, Optional[int]]
_VOCAB_FILE_NAMES: Dict[str, str]
_VOCAB_FILE_MAP: Dict[str, Dict[str, str]]
_SPECIAL_TOKENS_ATTRIBUTES = ["bos_token", "eos_token", "unk_token",
"sep_token", "pad_token", "cls_token",
"mask_token", "additional_special_tokens"]
Expand Down
15 changes: 14 additions & 1 deletion texar/torch/data/tokenizers/xlnet_tokenizer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -67,6 +67,12 @@ class XLNetTokenizer(PretrainedXLNetMixin, TokenizerBase):
'xlnet-large-cased': None,
}
_VOCAB_FILE_NAMES = {'vocab_file': 'spiece.model'}
_VOCAB_FILE_MAP = {
'vocab_file': {
'xlnet-base-cased': 'spiece.model',
'xlnet-large-cased': 'spiece.model',
}
}

def __init__(self,
pretrained_model_name: Optional[str] = None,
Expand All @@ -85,8 +91,10 @@ def __init__(self,
}

if self.pretrained_model_dir is not None:
assert self.pretrained_model_name is not None
vocab_file = os.path.join(self.pretrained_model_dir,
self._VOCAB_FILE_NAMES['vocab_file'])
self._VOCAB_FILE_MAP['vocab_file']
[self.pretrained_model_name])
assert pretrained_model_name is not None
if self._MAX_INPUT_SIZE.get(pretrained_model_name):
self.max_len = self._MAX_INPUT_SIZE[pretrained_model_name]
Expand Down Expand Up @@ -304,6 +312,7 @@ def default_hparams() -> Dict[str, Any]:
"do_lower_case": False,
"remove_space": True,
"keep_accents": False,
"name": "xlnet_tokenizer",
}

Here:
Expand Down Expand Up @@ -349,6 +358,9 @@ def default_hparams() -> Dict[str, Any]:

`"keep_accents"`: bool
Whether to keep the accents in the text.

`"name"`: str
Name of the tokenizer.
"""
return {
'pretrained_model_name': 'xlnet-base-cased',
Expand All @@ -365,6 +377,7 @@ def default_hparams() -> Dict[str, Any]:
'do_lower_case': False,
'remove_space': True,
'keep_accents': False,
'name': 'xlnet_tokenizer',
'@no_typecheck': ['pretrained_model_name'],
}

Expand Down