diff options
Diffstat (limited to 'examples/server/utils.hpp')
| -rw-r--r-- | examples/server/utils.hpp | 170 |
1 files changed, 0 insertions, 170 deletions
diff --git a/examples/server/utils.hpp b/examples/server/utils.hpp index e6a1f069..70be0748 100644 --- a/examples/server/utils.hpp +++ b/examples/server/utils.hpp @@ -411,177 +411,7 @@ static json oaicompat_completion_params_parse( return llama_params; } -static json format_final_response_oaicompat(const json & request, json result, const std::string & completion_id, bool streaming = false) { - bool stopped_word = result.count("stopped_word") != 0; - bool stopped_eos = json_value(result, "stopped_eos", false); - int num_tokens_predicted = json_value(result, "tokens_predicted", 0); - int num_prompt_tokens = json_value(result, "tokens_evaluated", 0); - std::string content = json_value(result, "content", std::string("")); - - std::string finish_reason = "length"; - if (stopped_word || stopped_eos) { - finish_reason = "stop"; - } - - json choices = - streaming ? json::array({json{{"finish_reason", finish_reason}, - {"index", 0}, - {"delta", json::object()}}}) - : json::array({json{{"finish_reason", finish_reason}, - {"index", 0}, - {"message", json{{"content", content}, - {"role", "assistant"}}}}}); - - std::time_t t = std::time(0); - - json res = json { - {"choices", choices}, - {"created", t}, - {"model", - json_value(request, "model", std::string(DEFAULT_OAICOMPAT_MODEL))}, - {"object", streaming ? "chat.completion.chunk" : "chat.completion"}, - {"usage", json { - {"completion_tokens", num_tokens_predicted}, - {"prompt_tokens", num_prompt_tokens}, - {"total_tokens", num_tokens_predicted + num_prompt_tokens} - }}, - {"id", completion_id} - }; - - if (server_verbose) { - res["__verbose"] = result; - } - - if (result.contains("completion_probabilities")) { - res["completion_probabilities"] = json_value(result, "completion_probabilities", json::array()); - } - - return res; -} - -// return value is vector as there is one case where we might need to generate two responses -static std::vector<json> format_partial_response_oaicompat(json result, const std::string & completion_id) { - if (!result.contains("model") || !result.contains("oaicompat_token_ctr")) { - return std::vector<json>({result}); - } - - bool first = json_value(result, "oaicompat_token_ctr", 0) == 0; - std::string modelname = json_value(result, "model", std::string(DEFAULT_OAICOMPAT_MODEL)); - - bool stopped_word = json_value(result, "stopped_word", false); - bool stopped_eos = json_value(result, "stopped_eos", false); - bool stopped_limit = json_value(result, "stopped_limit", false); - std::string content = json_value(result, "content", std::string("")); - - std::string finish_reason; - if (stopped_word || stopped_eos) { - finish_reason = "stop"; - } - if (stopped_limit) { - finish_reason = "length"; - } - - std::time_t t = std::time(0); - - json choices; - if (!finish_reason.empty()) { - choices = json::array({json{{"finish_reason", finish_reason}, - {"index", 0}, - {"delta", json::object()}}}); - } else { - if (first) { - if (content.empty()) { - choices = json::array({json{{"finish_reason", nullptr}, - {"index", 0}, - {"delta", json{{"role", "assistant"}}}}}); - } else { - // We have to send this as two updates to conform to openai behavior - json initial_ret = json{{"choices", json::array({json{ - {"finish_reason", nullptr}, - {"index", 0}, - {"delta", json{ - {"role", "assistant"} - }}}})}, - {"created", t}, - {"id", completion_id}, - {"model", modelname}, - {"object", "chat.completion.chunk"}}; - - json second_ret = json{ - {"choices", json::array({json{{"finish_reason", nullptr}, - {"index", 0}, - {"delta", json{ - {"content", content}}} - }})}, - {"created", t}, - {"id", completion_id}, - {"model", modelname}, - {"object", "chat.completion.chunk"}}; - - return std::vector<json>({initial_ret, second_ret}); - } - } else { - // Some idiosyncrasy in task processing logic makes several trailing calls - // with empty content, we ignore these at the calee site. - if (content.empty()) { - return std::vector<json>({json::object()}); - } - - choices = json::array({json{ - {"finish_reason", nullptr}, - {"index", 0}, - {"delta", - json{ - {"content", content}, - }}, - }}); - } - } - - json ret = json { - {"choices", choices}, - {"created", t}, - {"id", completion_id}, - {"model", modelname}, - {"object", "chat.completion.chunk"} - }; - if (!finish_reason.empty()) { - int num_tokens_predicted = json_value(result, "tokens_predicted", 0); - int num_prompt_tokens = json_value(result, "tokens_evaluated", 0); - ret.push_back({"usage", json { - {"completion_tokens", num_tokens_predicted}, - {"prompt_tokens", num_prompt_tokens}, - {"total_tokens", num_tokens_predicted + num_prompt_tokens} - }}); - } - - return std::vector<json>({ret}); -} - -static json format_embeddings_response_oaicompat(const json & request, const json & embeddings) { - json data = json::array(); - int i = 0; - for (auto & elem : embeddings) { - data.push_back(json{ - {"embedding", json_value(elem, "embedding", json::array())}, - {"index", i++}, - {"object", "embedding"} - }); - } - - json res = json { - {"model", json_value(request, "model", std::string(DEFAULT_OAICOMPAT_MODEL))}, - {"object", "list"}, - {"usage", json { - {"prompt_tokens", 0}, - {"total_tokens", 0} - }}, - {"data", data} - }; - - return res; -} static json format_tokenizer_response(const std::vector<llama_token> & tokens) { return json { |