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

Eliminate tokenizes post, add option "special" to tokenize

Browse source 
Eliminate tab compression from modified files.
This commit is contained in:
Troy Beukema 2023-10-23 11:48:37 -04:00
parent 5872e4f4da
commit 4150e74d04
2 changed files with 1762 additions and 1776 deletions
Download patch file Download diff file Expand all files Collapse all files

10
examples/server/README.md
View file

@ -222,18 +222,12 @@ node index.js
*Options:*
`special`: Boolean indicating if special tokens should be tokenized (default false)
`content`: Set the text to tokenize.
Note that the special `BOS` token is not added in front of the text and also a space character is not inserted automatically as it is for `/completion`.
- **POST** `/tokenizes`: Tokenize a given text with special tokens.
*Options:*
`content`: Set the text to tokenize with special tokens.
Note that the special `BOS` token is not added in front of the text and also a space character is not inserted automatically as it is for `/completion`.
- **POST** `/detokenize`: Convert tokens to text.
*Options:*

16
examples/server/server.cpp
View file

@ -176,6 +176,7 @@ struct slot_params
{
bool stream = true;
bool cache_prompt = false; // remember the prompt to avoid reprocessing all prompt
bool special = false; // use in tokenizer call to tokenize special tokens
uint32_t seed = -1; // RNG seed
int32_t n_keep = 0; // number of tokens to keep from initial prompt
@ -747,6 +748,7 @@ struct llama_server_context
slot->params.stream = json_value(data, "stream", false);
slot->params.cache_prompt = json_value(data, "cache_prompt", false);
slot->params.special = json_value(data, "special", false);
slot->params.n_predict = json_value(data, "n_predict", default_params.n_predict);
slot->sparams.top_k = json_value(data, "top_k", default_sparams.top_k);
slot->sparams.top_p = json_value(data, "top_p", default_sparams.top_p);
@ -2494,22 +2496,12 @@ int main(int argc, char **argv)
svr.Post("/tokenize", [&llama](const httplib::Request &req, httplib::Response &res)
{
const json body = json::parse(req.body);
std::vector<llama_token> tokens;
if (body.count("content") != 0)
{
tokens = llama.tokenize(body["content"], false);
}
const json data = format_tokenizer_response(tokens);
return res.set_content(data.dump(), "application/json");
});
bool special = json_value(body, "special", false);
svr.Post("/tokenizes", [&llama](const httplib::Request &req, httplib::Response &res)
{
const json body = json::parse(req.body);
std::vector<llama_token> tokens;
if (body.count("content") != 0)
{
tokens = llama.tokenize(body["content"], false, true);
tokens = llama.tokenize(body["content"], false, special);
}
const json data = format_tokenizer_response(tokens);
return res.set_content(data.dump(), "application/json");