printf(" -p PROMPT, --prompt PROMPT read prompt from the argument.\n");
printf(" --stdin read prompt from standard input.\n");
printf(" --no-bos do not ever add a BOS token to the prompt, even if normally the model uses a BOS token.\n");
+ printf(" --no-escape do not escape input (such as \\n, \\t, etc.).\n");
printf(" --no-parse-special do not parse control tokens.\n");
printf(" --log-disable disable logs. Makes stderr quiet when loading the model.\n");
printf(" --show-count print the total number of tokens.\n");
// variables where to put any arguments we see.
bool printing_ids = false;
bool no_bos = false;
+ bool no_escape = false;
bool no_parse_special = false;
bool disable_logging = false;
bool show_token_count = false;
else if (arg == "--no-bos") {
no_bos = true;
}
+ else if (arg == "--no-escape") {
+ no_escape = true;
+ }
else if (arg == "--no-parse-special") {
no_parse_special = true;
}
const bool model_wants_add_bos = llama_add_bos_token(model);
const bool add_bos = model_wants_add_bos && !no_bos;
const bool parse_special = !no_parse_special;
+ const bool escape = !no_escape;
+
+ if (escape) {
+ string_process_escapes(prompt);
+ }
std::vector<llama_token> tokens;
tokens = common_tokenize(model, prompt, add_bos, parse_special);