CTranslate2
CTranslate2 copied to clipboard
Support Baichuan2?
Baichuan2 is a generative model similar to llama, I find the following two differences:
-
- qkv merge as W_pack, so i change the file /ctranslate2/converters/transformers.py
- qkv merge as W_pack, so i change the file /ctranslate2/converters/transformers.py
-
- rotary changes to alibi, so i change the file /ctranslate2/converters/transformers.py
- rotary changes to alibi, so i change the file /ctranslate2/converters/transformers.py
But I found that the conversion only worked on simple sentences, complex sentences would not work.
Could you help me or is there a plan to support Baichuan2?
- link https://huggingface.co/baichuan-inc/Baichuan2-13B-Chat https://huggingface.co/baichuan-inc/Baichuan2-13B-Chat/blob/main/modeling_baichuan.py