The default model (Mistral) specifies <s> as the BOS token in the prompt template: <s>[INST] {prompt} [/INST].
However, this causes llama-cpp-python to output the following warning:
llama_tokenize_internal: Added a BOS token to the prompt as specified by the model but the prompt also starts with a BOS token. So now the final prompt starts with 2 BOS tokens. Are you sure this is what you want?
... which suggests the BOS token is being added by llama-cpp-python. Thus, <s> should be removed from the default prompt template.
The default model (Mistral) specifies
<s>
as the BOS token in the prompt template:<s>[INST] {prompt} [/INST]
.However, this causes
llama-cpp-python
to output the following warning:... which suggests the BOS token is being added by
llama-cpp-python
. Thus,<s>
should be removed from the default prompt template.