参考资料:https://mp.weixin.qq.com/s/hzPQHz9Si01ODdvU3i02ow
在tokenization文件里共有3个class和11个函数
1.def validate_case_matches_checkpoint
def validate_case_matches_checkpoint(do_lower_case, init_checkpoint):
"""Checks whether the casing config is consistent with the checkpoint name."""
if not init_checkpoint:
return
m = re.match("^.*?([A-Za-z0-9_-]+)/bert_model.ckpt", init_checkpoint)
if m is None:
return
model_name = m.group(1)
lower_models = [
"uncased_L-24_H-1024_A-16", "uncased_L-12_H-768_A-12",
"multilingual_L-12_H-768_A-12", "chinese_L-12_H-768_A-12"
]
cased_models = [
"cased_L-12_H-768_A-12", "cased_L-24_H-1024_A-16",
"multi_cased_L-12_H-768_A-12"
]
is_bad_config = False
if model_name in lower_models and not do_lower_case:
is_bad_config = True
actual_flag = "False"
case_name = "lowercased"
opposite_flag = "True"
if model_name in cased_models and do_lower_case:
is_bad_config = True
actual_flag = "True"
case_name = "cased"
opposite_flag = "False"
if is_bad_config:
raise ValueError(
"You passed in `--do_lower_case=%s` with `--init_checkpoint=%s`. "
"However, `%s` seems to be a %s model, so you "
"should pass in `--do_lower_case=%s` so that the fine-tuning matches "
"how the model was pre-training. If this error is wrong, please "
"just comment out this check." % (actual_flag, init_checkpoint,
model_name, case_name, opposite_flag))
2.def convert_to_unicode
def convert_to_unicode(text):
#将输入为utf-8的文本转化成Unicode编码
"""Converts `text` to Unicode (if it's not already), assuming utf-8 input."""
if six.PY3:
if isinstance(text, str):
return text
elif isinstance(text, bytes):
return text.decode("utf-8", "ignore")
else:
raise ValueError("Unsupported string type: %s" % (type(text)))
elif six.PY2:
if isinstance(text, str):
return text.decode("utf-8", "ignore")
elif isinstance(text, unicode):
return text
else:
raise ValueError("Unsupported string type: %s" % (type(text)))
else:
raise ValueError("Not running on Python2 or Python 3?")