serverinfill tokens correction

This commit is contained in:
vvhg1 2023-10-06 18:35:50 +02:00
parent 8bd24b2e5c
commit 6796e7450c

View File

@ -344,9 +344,10 @@ struct llama_server_context
void loadInfill()
{
auto prefix_tokens = tokenize(params.input_prefix, true); // always add BOS
auto suffix_tokens = tokenize(params.input_suffix, true); // always add BOS
auto prefix_tokens = tokenize(params.input_prefix, false);
auto suffix_tokens = tokenize(params.input_suffix, false);
prefix_tokens.insert(prefix_tokens.begin(), llama_token_prefix(ctx));
prefix_tokens.insert(prefix_tokens.begin(), llama_token_bos(ctx)); // always add BOS
prefix_tokens.insert(prefix_tokens.end(), llama_token_suffix(ctx));
prefix_tokens.insert(prefix_tokens.end(), suffix_tokens.begin(), suffix_tokens.end());
prefix_tokens.push_back(llama_token_middle(ctx));