ikawrakow/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-04-27 09:53:40 +00:00
Code Issues Packages Projects Releases Wiki Activity

Server: add string ban (#1185)

Browse Source 
* server: add string ban

* increase rewind limit

* init n_buffer

---------

Co-authored-by: firecoperana <firecoperana>
This commit is contained in:
firecoperana
2026-02-05 00:12:34 -06:00
committed by GitHub
parent a335cff664
commit 8d952ff183
5 changed files with 337 additions and 53 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
examples/server/server-context.h
View File

@@ -83,6 +83,16 @@ struct server_slot {
std::string stopping_word;
stop_type stop;
// For context rewind/ token buffer
size_t n_buffer = 0;
int32_t rewind_count = 0;
bool rewind_status = false;
std::unordered_map<llama_token, float> logit_bias;
std::vector<std::string>ban_phrases;
completion_token_outputs token_buffer;
float ban_phrases_bias = 0;
int32_t banned_n = 1;
server_prompt server_cached_prompt;
void prompt_save(server_prompt_cache& prompt_cache) const;
@@ -316,6 +326,12 @@ struct server_context {
bool accept_special_token(const server_slot& slot, const llama_token token);
bool has_next_token(const completion_token_output& result, server_slot& slot);
void send_token_results(completion_token_outputs& results, server_slot& slot, int32_t n = 0);
void buffer_and_check_string_ban(server_slot& slot, completion_token_output& result);
json model_meta() const;
// Re-aggregates all active vectors and updates the model state