CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-10-27 08:21:30 +00:00
Code Issues Packages Projects Releases Wiki Activity

llama : deprecate llama_kv_self_ API (#14030)

Browse Source 
* llama : deprecate llama_kv_self_ API

ggml-ci

* llama : allow llama_memory_(nullptr)

ggml-ci

* memory : add flag for optional data clear in llama_memory_clear

ggml-ci
This commit is contained in:
Georgi Gerganov
2025-06-06 14:11:15 +03:00
committed by GitHub
parent 487a5e0401
commit 745aa5319b
34 changed files with 206 additions and 127 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
tools/batched-bench/batched-bench.cpp
View File

@@ -57,6 +57,8 @@ int main(int argc, char ** argv) {
return 1;
}
auto * mem = llama_get_memory(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_self_clear(ctx);
llama_memory_clear(mem, false);
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_self_seq_cp(ctx, 0, i, -1, -1);
llama_memory_seq_cp(mem, 0, i, -1, -1);
}
}