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

llama : cont

Browse source 
ggml-ci
This commit is contained in:
Georgi Gerganov 2025-01-13 14:56:52 +02:00
parent f78b396ee7
commit e4550fbafc
No known key found for this signature in database
GPG key ID: 449E073F9DC10735
19 changed files with 128 additions and 79 deletions
Download patch file Download diff file Expand all files Collapse all files

6
examples/batched-bench/batched-bench.cpp
View file

@ -57,6 +57,8 @@ int main(int argc, char ** argv) {
return 1;
}
llama_kv_cache * kv = llama_get_kv_cache(ctx);
const int32_t n_kv_max = llama_n_ctx(ctx);
llama_batch batch = llama_batch_init(n_kv_max, 0, 1);
@ -132,7 +134,7 @@ int main(int argc, char ** argv) {
const auto t_pp_start = ggml_time_us();
llama_kv_cache_clear(ctx);
llama_kv_cache_clear(kv);
if (!decode_helper(ctx, batch, ctx_params.n_batch)) {
LOG_ERR("%s: llama_decode() failed\n", __func__);
@ -141,7 +143,7 @@ int main(int argc, char ** argv) {
if (is_pp_shared) {
for (int32_t i = 1; i < pl; ++i) {
llama_kv_cache_seq_cp(ctx, 0, i, -1, -1);
llama_kv_cache_seq_cp(kv, 0, i, -1, -1);
}
}