nikola/llama_cpp
1
0
Fork 0
Code Issues Pull Requests Actions 22 Packages Projects Releases Wiki Activity

server: Fix undefined timing measurement errors in server context (#21201)

Browse Source 
Co-authored-by: Dan Hoffman <dhoffman@cyket.net>
This commit is contained in:
Dan Hoffman
2026-04-04 22:11:19 +08:00
committed by GitHub
co-authored by Dan Hoffman
parent d01f6274c0
commit 9c699074c9
2 changed files with 8 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tools/server/server-context.cpp
+2 -2
View File
@@ -155,8 +155,8 @@ struct server_slot {
int64_t t_start_process_prompt; int64_t t_start_process_prompt;
int64_t t_start_generation; int64_t t_start_generation;
double t_prompt_processing; // ms double t_prompt_processing = 0.0; // ms
double t_token_generation; // ms double t_token_generation = 0.0; // ms
std::function<void(int /* id_slot */)> callback_on_release; std::function<void(int /* id_slot */)> callback_on_release;
tools/server/server-task.h
+6 -6
View File
@@ -261,14 +261,14 @@ struct result_timings {
int32_t cache_n = -1; int32_t cache_n = -1;
int32_t prompt_n = -1; int32_t prompt_n = -1;
double prompt_ms; double prompt_ms = 0.0;
double prompt_per_token_ms; double prompt_per_token_ms = 0.0;
double prompt_per_second; double prompt_per_second = 0.0;
int32_t predicted_n = -1; int32_t predicted_n = -1;
double predicted_ms; double predicted_ms = 0.0;
double predicted_per_token_ms; double predicted_per_token_ms = 0.0;
double predicted_per_second; double predicted_per_second = 0.0;
// Optional speculative metrics - only included when > 0 // Optional speculative metrics - only included when > 0
int32_t draft_n = 0; int32_t draft_n = 0;