server : insert the extra context before the FIM prefix

Author: ggerganov

examples/server/server.cpp

@@ -1987,7 +1987,8 @@ struct server_context {
 
                                     SLT_DBG(slot, "extra: n_ctx = %d, n_extra_take = %d, n_extra = %d\n", slot.n_ctx, n_extra_take, (int) slot.extra_tokens.size());
 
-                                    embd_inp.insert(embd_inp.begin() + 1, slot.extra_tokens.end() - n_extra_take, slot.extra_tokens.end());
+                                    // put the extra context before the FIM prefix
+                                    embd_inp.insert(embd_inp.begin(), slot.extra_tokens.end() - n_extra_take, slot.extra_tokens.end());
 
                                     embd_inp.insert(embd_inp.end(), embd_end.begin(), embd_end.end());
                                     embd_inp.push_back(llama_token_fim_mid(model));