update config about model precision, fix apply_rotary_pos_emb

update readme and fix convert_tokens_to_string

support cpu inference, format file (#9)

Update modeling_qwen.py, fix logn bug

fix flash-attention usage

add resource files