update <|endoftext|> tokenizer id from 50257 to 50256

refactor tokenizer related files with eos token

fixed mismatched vocab_size between model and tokenizer

add tokenizer