About the function of assign special token ids to the model.config object

Nobody has claimed this yet.

Hi @bofenghuang, I just read through the Qwen-7B code and found the following code fragment,
```
class QwenChatAdapter(BaseModelAdapter):
tokenizer.eos_token_id = config.eos_token_id
tokenizer.bos_token_id = config.bos_token_id
tokenizer.pad_token_id = generation_config.pad_token_id
model.config.eos_token_id = tokenizer.eos_token_id
model.config.bos_token_id = tokenizer.bos_token_id
model.config.pad_token_id = tokenizer.pad_token_id
```
I don't understand the function of adding the `xxx_token_id` to the `model.config.xxx` attribute.
I found some language model like `llama-2` also have the above code.

Could you tell me the funciton of the above code?

Thanks

The link to the above code is:
https://github.com/lm-sys/FastChat/blob/8432833f3bcdcea0fe48b6c28fac8a0f456b3d05/fastchat/model/model_adapter.py#L1322C1-L1330C59

Start with fastchat/model/model_adapter.py around the linked QwenChatAdapter lines 1322-1330, then compare the corresponding Llama-2 code mentioned in the issue. Explain the purpose of assigning the special token IDs to model.config and identify whether this behavior should be documented or clarified in the relevant project documentation.

Written by the indexing model from the issue text.