max_len = 20
vocab = build_vocab_from_iterator(
    list(map(str,range(10))) # Arabic symbols
    + ['I','V','X','L','C','D','M'], # Roman symbols
    specials=['<bos>', '<pad>', '<eos>'] # Special symbols
)
vocab_size = len(vocab)
pad_idx = vocab['<pad>']
collate_fn = partial(collate, pad_idx=pad_idx, max_len=max_len)
proc = Processor(vocab)
train_ds = NumberDataset.from_file('train', processor=proc)
valid_ds = NumberDataset.from_file('valid', processor=proc)
train_dl = DataLoader(train_ds, batch_size=10, collate_fn=collate_fn)
valid_dl = DataLoader(valid_ds, batch_size=10, collate_fn=collate_fn)