Skip to content

server: fix system_tokens being erased in kv_cache; #6312

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 2 commits into
base: master
Choose a base branch
from
+6 −6
Open

server: fix system_tokens being erased in kv_cache; #6312

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
694dcd9
server: fix system_tokens being erased in kv_cache;
Mar 26, 2024
1440d44
Merge branch 'master' into master
MasterYi1024 Mar 28, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 6 additions & 6 deletions examples/server/server.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1699,7 +1699,7 @@ struct server_context {
if (slot.is_processing() && (int) system_tokens.size() + slot.n_past >= slot.n_ctx - 1) {
// Shift context
const int n_keep = slot.params.n_keep + add_bos_token;
const int n_left = (int) system_tokens.size() + slot.n_past - n_keep;
const int n_left = slot.n_past - n_keep;
const int n_discard = slot.params.n_discard ? slot.params.n_discard : (n_left / 2);

LOG_INFO("slot context shift", {
Expand All @@ -1714,8 +1714,8 @@ struct server_context {
{"n_cache_tokens", slot.cache_tokens.size()}
});

llama_kv_cache_seq_rm (ctx, slot.id + 1, n_keep , n_keep + n_discard);
llama_kv_cache_seq_add(ctx, slot.id + 1, n_keep + n_discard, system_tokens.size() + slot.n_past, -n_discard);
llama_kv_cache_seq_rm (ctx, slot.id + 1, system_tokens.size() + n_keep , system_tokens.size() + n_keep + n_discard);
llama_kv_cache_seq_add(ctx, slot.id + 1, system_tokens.size() + n_keep + n_discard, system_tokens.size() + slot.n_past, -n_discard);

if (slot.params.cache_prompt) {
for (size_t i = n_keep + n_discard; i < slot.cache_tokens.size(); i++) {
Expand Down Expand Up @@ -1857,18 +1857,18 @@ struct server_context {

// if input prompt is too big, truncate it (if group attention self-extend is disabled)
if (slot.ga_n == 1 && slot.n_prompt_tokens >= slot.n_ctx) {
const int n_left = slot.n_ctx - slot.params.n_keep;
const int n_left = slot.n_ctx - slot.params.n_keep - system_tokens.size();

const int n_block_size = n_left / 2;
const int erased_blocks = (slot.n_prompt_tokens - slot.params.n_keep - n_block_size) / n_block_size;
const int erased_blocks = (slot.n_prompt_tokens - slot.params.n_keep - system_tokens.size() - n_block_size) / n_block_size;

std::vector<llama_token> new_tokens(
prompt_tokens.begin(),
prompt_tokens.begin() + slot.params.n_keep);

new_tokens.insert(
new_tokens.end(),
prompt_tokens.begin() + slot.params.n_keep + erased_blocks * n_block_size,
prompt_tokens.begin() + slot.params.n_keep + system_tokens.size() + erased_blocks * n_block_size,
prompt_tokens.end());

prompt_tokens = std::move(new_tokens);
Expand Down