fix sampling in chat stream

fix slow long sequence inference

Update modeling_qwen.py

update tokenization and readme

deprecate argument stream in model.chat()

update support for flash attn

fix chat streaming

add support for flash attn 2

fix kwargs in generate method and update readme

update config and streaming generation

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