indonesian-roberta-base / train_tokenizer.py
w11wo's picture
oscar tokenizers
cd9ad5b
raw
history blame
617 Bytes
from datasets import load_dataset
from tokenizers import ByteLevelBPETokenizer
# load dataset
dataset = load_dataset("oscar", "unshuffled_deduplicated_id", split="train")
# Instantiate tokenizer
tokenizer = ByteLevelBPETokenizer()
def batch_iterator(batch_size=1000):
for i in range(0, len(dataset), batch_size):
yield dataset[i : i + batch_size]["text"]
# Customized training
tokenizer.train_from_iterator(
batch_iterator(),
vocab_size=50265,
min_frequency=2,
special_tokens=["<s>", "<pad>", "</s>", "<unk>", "<mask>",],
)
# Save files to disk
tokenizer.save(f"./tokenizer.json")