vrr/prima.cpp
2
0
Fork 0
mirror of https://github.com/Lizonghang/prima.cpp.git synced 2025-09-10 13:14:33 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Merge branch 'dev' into feat/auto-exit

Browse source
This commit is contained in:
Li, Zonghang 2025-05-20 02:04:14 +08:00 • committed by GitHub
commit 7b0ededd24
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
10 changed files with 488 additions and 87 deletions
Download patch file Download diff file Expand all files Collapse all files

8
examples/main/main.cpp
View file

@ -351,6 +351,9 @@ int main(int argc, char ** argv) {
// remove any "future" tokens that we might have inherited from the previous session
llama_kv_cache_seq_rm(ctx, -1, n_matching_session_tokens, -1);
if (my_rank == 0) {
llama_send_kv_cache_seq_rm(ctx, -1, n_matching_session_tokens, -1);
}
}
LOG_DBG("recalculate the cached logits (check): embd_inp.size() %zu, n_matching_session_tokens %zu, embd_inp.size() %zu, session_tokens.size() %zu\n",
@ -596,6 +599,11 @@ int main(int argc, char ** argv) {
llama_kv_cache_seq_rm (ctx, 0, params.n_keep , params.n_keep + n_discard);
llama_kv_cache_seq_add(ctx, 0, params.n_keep + n_discard, n_past, -n_discard);
if (my_rank == 0) {
llama_send_kv_cache_seq_rm (ctx, 0, params.n_keep , params.n_keep + n_discard);
llama_send_kv_cache_seq_add(ctx, 0, params.n_keep + n_discard, n_past, -n_discard);
}
n_past -= n_discard;
LOG_DBG("after swap: n_past = %d\n", n_past);