vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2025-09-10 17:14:36 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

expose timing info in web api

Browse source
This commit is contained in:
Concedo 2023-07-11 18:56:06 +08:00
parent 7222877069
commit 1d1111e10f
4 changed files with 17 additions and 16 deletions
Download patch file Download diff file Expand all files Collapse all files

10
gpttype_adapter.cpp
View file

@ -33,8 +33,8 @@ std::string executable_path = "";
std::string lora_filename = "";
std::string lora_base = "";
bool generation_finished;
float prompt_process_time;
float prompt_eval_time;
float last_process_time = 0;
float last_eval_time = 0;
std::vector<std::string> generated_tokens;
//return val: 0=fail, 1=(original ggml, alpaca), 2=(ggmf), 3=(ggjt)
@ -869,8 +869,6 @@ generation_outputs gpttype_generate(const generation_inputs inputs, generation_o
bool stream_sse = inputs.stream_sse;
generation_finished = false; // Set current generation status
prompt_eval_time = 0;
prompt_process_time = 0;
generated_tokens.clear(); // New Generation, new tokens
if (params.repeat_last_n < 1)
@ -1449,8 +1447,8 @@ generation_outputs gpttype_generate(const generation_inputs inputs, generation_o
fflush(stdout);
output.status = 1;
generation_finished = true;
prompt_eval_time = pt2;
prompt_process_time = pt1;
last_eval_time = pt2;
last_process_time = pt1;
snprintf(output.text, sizeof(output.text), "%s", concat_output.c_str());
return output;