diff options
| author | Georgi Gerganov <ggerganov@gmail.com> | 2024-02-27 14:35:51 +0200 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2024-02-27 14:35:51 +0200 |
| commit | 9d533a77d0c3850ce09d736bc1baa67fd6ad27b3 (patch) | |
| tree | 25adffcbb0f7c13a8578279456a4937ed73ae3f6 /examples/passkey | |
| parent | cbbd1efa06f8c09f9dff58ff9d9af509cc4c152b (diff) | |
llama : fix defrag bugs + add parameter (#5735)
* llama : fix defrag bugs + enable by default
ggml-ci
* llama : add defrag_thold parameter
ggml-ci
* llama : cont
* llama : disable log message
ggml-ci
* llama : fix graph size check during defrag
Diffstat (limited to 'examples/passkey')
| -rw-r--r-- | examples/passkey/passkey.cpp | 4 |
1 files changed, 2 insertions, 2 deletions
diff --git a/examples/passkey/passkey.cpp b/examples/passkey/passkey.cpp index 47de67a9..2cbc9e1f 100644 --- a/examples/passkey/passkey.cpp +++ b/examples/passkey/passkey.cpp @@ -182,7 +182,7 @@ int main(int argc, char ** argv) { llama_kv_cache_seq_rm (ctx, 0, n_keep , n_keep + n_discard); llama_kv_cache_seq_add(ctx, 0, n_keep + n_discard, n_ctx, -n_discard); - llama_kv_cache_defrag (ctx); + //llama_kv_cache_defrag (ctx); llama_kv_cache_update (ctx); n_past = llama_kv_cache_seq_pos_max(ctx, 0) + 1; @@ -213,7 +213,7 @@ int main(int argc, char ** argv) { llama_kv_cache_seq_rm (ctx, 0, n_keep , n_keep + n_discard); llama_kv_cache_seq_add(ctx, 0, n_keep + n_discard, n_ctx, -n_discard); - llama_kv_cache_defrag (ctx); + //llama_kv_cache_defrag (ctx); llama_kv_cache_update (ctx); n_past = llama_kv_cache_seq_pos_max(ctx, 0) + 1; |