from datasets import load_from_disk
from tokenizers import trainers, Tokenizer, normalizers, ByteLevelBPETokenizer

# load dataset
dataset = load_from_disk("/researchdisk/lm_training_dataset_full")["train"]

# Instantiate tokenizer
tokenizer = ByteLevelBPETokenizer()

def batch_iterator(batch_size=5000):
    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("./tokenizer.json")