change config and fix training tokenizer script

change to proper vocab_size

add pre-generated config.json and trained Thai tokenizer.json