File tree Expand file tree Collapse file tree 2 files changed +17
-3
lines changed Expand file tree Collapse file tree 2 files changed +17
-3
lines changed Original file line number Diff line number Diff line change @@ -223,7 +223,11 @@ int main(int argc, char ** argv) {
223
223
inp_sfx.insert (inp_sfx.begin (), llama_token_suffix (model));
224
224
embd_inp = inp_pfx;
225
225
embd_inp.insert (embd_inp.end (), inp_sfx.begin (), inp_sfx.end ());
226
- embd_inp.push_back (llama_token_middle (model));
226
+
227
+ const llama_token middle_token = llama_token_middle (model);
228
+ if (middle_token >= 0 ) {
229
+ embd_inp.push_back (middle_token);
230
+ }
227
231
228
232
LOG (" prefix: \" %s\"\n " , log_tostr (params.input_prefix ));
229
233
LOG (" suffix: \" %s\"\n " , log_tostr (params.input_suffix ));
@@ -528,7 +532,12 @@ int main(int argc, char ** argv) {
528
532
inp_sfx.insert (inp_sfx.begin (), llama_token_suffix (model));
529
533
embd_inp = inp_pfx;
530
534
embd_inp.insert (embd_inp.end (), inp_sfx.begin (), inp_sfx.end ());
531
- embd_inp.push_back (llama_token_middle (model));
535
+
536
+ const llama_token middle_token = llama_token_middle (model);
537
+ if (middle_token >= 0 ) {
538
+ embd_inp.push_back (middle_token);
539
+ }
540
+
532
541
embd.clear ();
533
542
n_remain = params.n_predict ;
534
543
n_past = 0 ;
Original file line number Diff line number Diff line change @@ -2038,7 +2038,12 @@ struct server_context {
2038
2038
prefix_tokens.insert (prefix_tokens.begin (), llama_token_bos (model)); // always add BOS
2039
2039
prefix_tokens.insert (prefix_tokens.end (), llama_token_suffix (model));
2040
2040
prefix_tokens.insert (prefix_tokens.end (), suffix_tokens.begin (), suffix_tokens.end ());
2041
- prefix_tokens.push_back (llama_token_middle (model));
2041
+
2042
+ const llama_token middle_token = llama_token_middle (model);
2043
+ if (middle_token >= 0 ) {
2044
+ prefix_tokens.push_back (middle_token);
2045
+ }
2046
+
2042
2047
prompt_tokens = prefix_tokens;
2043
2048
} else {
2044
2049
prompt_tokens = tokenize (slot.prompt , system_prompt.empty ()); // add BOS if there isn't system prompt
You can’t perform that action at this time.
0 commit comments