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

parallel : process system prompt once + configurable paramters + llama API

Browse source
This commit is contained in:
Georgi Gerganov 2023-09-19 17:00:42 +03:00
parent 82e20e9ba0
commit 4b5f3cd6bf
No known key found for this signature in database
GPG key ID: 449E073F9DC10735
9 changed files with 187 additions and 93 deletions
Download patch file Download diff file Expand all files Collapse all files

4
examples/llama-bench/llama-bench.cpp
View file

@ -977,7 +977,7 @@ int main(int argc, char ** argv) {
test t(inst, lmodel, ctx);
llama_kv_cache_rm_tokens(ctx, -1, -1);
llama_kv_cache_tokens_rm(ctx, -1, -1);
// warmup run
if (t.n_prompt > 0) {
@ -988,7 +988,7 @@ int main(int argc, char ** argv) {
}
for (int i = 0; i < params.reps; i++) {
llama_kv_cache_rm_tokens(ctx, -1, -1);
llama_kv_cache_tokens_rm(ctx, -1, -1);
uint64_t t_start = get_time_ns();
if (t.n_prompt > 0) {