vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

server : return stopping_word in the partial response

Browse source
This commit is contained in:
ZXED 2024-12-08 13:06:23 +03:00
parent ecc93d0558
commit c06405e10d
No known key found for this signature in database
GPG key ID: 637FB44813DCFD66
1 changed files with 4 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

5
examples/server/server.cpp
View file

@ -556,6 +556,7 @@ struct server_task_result_cmpl_partial : server_task_result {
int32_t n_decoded;
int32_t n_prompt_tokens;
std::string stopping_word;
stop_type stop = STOP_TYPE_NONE;
std::vector<completion_token_output> probs_output;
@ -586,6 +587,7 @@ struct server_task_result_cmpl_partial : server_task_result {
{"index", index},
{"content", content},
{"stop_type", stop_type_to_str(stop)},
{"stopping_word", stopping_word},
{"stop", is_stop},
{"id_slot", id_slot},
{"tokens_predicted", n_decoded},
@ -1892,7 +1894,8 @@ struct server_context {
res->n_decoded = slot.n_decoded;
res->n_prompt_tokens = slot.n_prompt_tokens;
res->stop = slot.stop;
res->stopping_word = slot.stopping_word;
res->stop = slot.stop;
res->verbose = slot.params.verbose;
res->oaicompat = slot.params.oaicompat;