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

llama : more robust cell_max heuristic + wip shift

Browse source
This commit is contained in:
Georgi Gerganov 2023-09-18 17:15:25 +03:00
parent 4d76d762ef
commit f015b26689
No known key found for this signature in database
GPG key ID: 449E073F9DC10735
3 changed files with 39 additions and 52 deletions
Download patch file Download diff file Expand all files Collapse all files

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

@ -977,6 +977,8 @@ int main(int argc, char ** argv) {
test t(inst, lmodel, ctx);
llama_kv_cache_keep_seq(ctx, -1);
// warmup run
if (t.n_prompt > 0) {
test_prompt(ctx, std::min(2, t.n_batch), 0, t.n_batch, t.n_threads);
@ -986,6 +988,8 @@ int main(int argc, char ** argv) {
}
for (int i = 0; i < params.reps; i++) {
llama_kv_cache_keep_seq(ctx, -1);
uint64_t t_start = get_time_ns();
if (t.n_prompt > 0) {
test_prompt(ctx, t.n_prompt, 0, t.n_batch, t.n_threads);