xu-song's picture
add more tokenizers
f4973d4
raw
history blame
421 Bytes
import os
from transformers import AutoTokenizer, BloomTokenizerFast
CURRENT_DIR = os.path.dirname(os.path.abspath(__file__))
TOKENIZER_DIR = os.path.join(CURRENT_DIR, "moss-moon-003-sft")
tokenizer = AutoTokenizer.from_pretrained(TOKENIZER_DIR, trust_remote_code=True)
# vocab_size = len(tokenizer.get_vocab())
# vocab_size = tokenizer.vocab_size
tokenizer.comments = "在gpt2词典基础上,扩充了5万中文"