ikawrakow/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-04-28 18:32:04 +00:00
Code Issues Packages Projects Releases Wiki Activity

Revert kv cache name changes

Browse Source
This commit is contained in:
firecoperana
2026-01-17 08:42:43 -06:00
parent 79185862e5
commit 93abfb66ab
15 changed files with 62 additions and 62 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
examples/lookahead/lookahead.cpp
View File

@@ -96,7 +96,7 @@ int main(int argc, char ** argv) {
llama_decode(ctx, llama_batch_get_one(&inp.back(), 1, n_input - 1, 0));
for (int s = 1; s < W + G + 1; ++s) {
llama_memory_seq_cp(ctx, 0, s, -1, -1);
llama_kv_cache_seq_cp(ctx, 0, s, -1, -1);
}
const auto t_enc_end = ggml_time_us();
@@ -438,17 +438,17 @@ int main(int argc, char ** argv) {
// KV cache management
// if no verification token matched, we simply remove all cells from this batch -> no fragmentation
llama_memory_seq_rm(ctx, -1, n_past, -1);
llama_kv_cache_seq_rm(ctx, -1, n_past, -1);
if (seq_id_best != 0) {
// if a verification token matched, we keep the best sequence and remove the rest
// this leads to some KV cache fragmentation
llama_kv_cache_seq_keep(ctx, seq_id_best);
llama_memory_seq_cp (ctx, seq_id_best, 0, -1, -1);
llama_memory_seq_rm (ctx, seq_id_best, -1, -1);
llama_kv_cache_seq_cp (ctx, seq_id_best, 0, -1, -1);
llama_kv_cache_seq_rm (ctx, seq_id_best, -1, -1);
for (int s = 1; s < W + G + 1; ++s) {
llama_memory_seq_cp(ctx, 0, s, -1, -1);
llama_kv_cache_seq_cp(ctx, 0, s, -1, -1);
}
}
}