Skip to content

kv-cache : initialize per-stream views after buffer allocation (#23737) #27738

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
ByungHyun21 wants to merge 1 commit into from
Closed

kv-cache : initialize per-stream views after buffer allocation (#23737) #27738

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
20 changes: 20 additions & 0 deletions src/llama-kv-cache.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -292,6 +292,26 @@ llama_kv_cache::llama_kv_cache(
ctxs_bufs.emplace_back(std::move(ctx), buf);
}

// the per-stream views were created before the k/v tensors were allocated, so
// they may still have data == NULL while their storage lives on view_src —
// depending on the allocation path taken. Direct IO (checkpoint serialization,
// stream-to-stream copies) accesses these views through
// ggml_backend_tensor_get/set, which requires an initialized tensor on
// backends that address via tensor->data. Initialize them here (#23737).
if (!hparams.no_alloc) {
for (const auto & layer : layers) {
for (ggml_tensor * view : layer.k_stream) {
if (view != nullptr && view->data == nullptr && view->view_src != nullptr && view->view_src->data != nullptr) {
GGML_ASSERT(ggml_backend_view_init(view) == GGML_STATUS_SUCCESS);
}
}
for (ggml_tensor * view : layer.v_stream) {
if (view != nullptr && view->data == nullptr && view->view_src != nullptr && view->view_src->data != nullptr) {
GGML_ASSERT(ggml_backend_view_init(view) == GGML_STATUS_SUCCESS);
}
}
}
}
{
const size_t memory_size_k = size_k_bytes();
const size_t memory_size_v = size_v_bytes();
Expand Down