Skip to content

server: prevent destructive memory wipe on seq_rm failure for hybrid models #22534

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

Closed
BlisteringViola wants to merge 1 commit into from
+11 −6
Closed

server: prevent destructive memory wipe on seq_rm failure for hybrid models #22534

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
17 changes: 11 additions & 6 deletions tools/server/server-context.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2559,12 +2559,17 @@ struct server_context_impl {
SLT_INF(slot, "n_tokens = %d, memory_seq_rm [%d, end)\n", slot.prompt.n_tokens(), p0);

if (!llama_memory_seq_rm(llama_get_memory(ctx), slot.id, p0, -1)) {
SLT_WRN(slot, "failed to truncate tokens with position >= %d - clearing the memory\n", p0);

slot.prompt_clear(true);

// there is no common part left
slot.n_prompt_tokens_cache = 0;
if (slot.ctx_seq_rm_type == COMMON_CONTEXT_SEQ_RM_TYPE_FULL && slot.n_prompt_tokens_cache > 0) {
// hybrid/recurrent: partial seq_rm always fails, but checkpoint restored valid state
SLT_INF(slot, "seq_rm failed (expected for hybrid) - keeping %d cached tokens from checkpoint\n", slot.n_prompt_tokens_cache);
} else {
SLT_WRN(slot, "failed to truncate tokens with position >= %d - clearing the memory\n", p0);

slot.prompt_clear(true);

// there is no common part left
slot.n_prompt_tokens_cache = 0;
}
}

// If using an alora, there may be uncached tokens that come
Expand Down