fix(llama.cpp): fix eos without cache (#1852)

This commit is contained in:
Ettore Di Giacinto
2024-03-18 18:59:24 +01:00
committed by GitHub
parent b202bfaaa0
commit fa9e330fc6

View File

@ -1084,7 +1084,7 @@ struct llama_server_context
slot.has_next_token = false;
}
if (!slot.cache_tokens.empty() && result.tok == llama_token_eos(model))
if (result.tok == llama_token_eos(model))
{
slot.stopped_eos = true;
slot.has_next_token = false;