Skip to content
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

Change shared_instance type from weakptr to shared_ptr #507

Merged
merged 2 commits into from
Oct 9, 2023
Merged

Change shared_instance type from weakptr to shared_ptr #507

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
7faed16
change shared_instances_ from weakptr to sharedptr
lvhan028 Sep 28, 2023
9f17d22
update
lvhan028 Sep 28, 2023
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
4 changes: 2 additions & 2 deletions src/turbomind/triton_backend/llama/LlamaTritonModel.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -273,7 +273,7 @@ LlamaTritonModel<T>::createModelInstance(int
std::shared_ptr<LlamaTritonSharedModelInstance<T>> instance;
{
std::lock_guard<std::mutex> lock(shared_mutexes_[device_id]);
instance = shared_instances_[device_id].lock();
instance = shared_instances_[device_id];
if (!instance) {
instance = createSharedModelInstance(device_id, rank, nccl_params, custom_all_reduce_comm);
instance->llm->setFfiLock(ffi_lock_);
Expand Down Expand Up @@ -347,7 +347,7 @@ LlamaTritonModel<T>::createNcclParams(const int node_id, const int device_id_sta
// create nccl group when there are non-occupied devices
for (int i = 0; i < device_count; ++i) {
std::lock_guard<std::mutex> lock(shared_mutexes_[i]);
if (shared_instances_[i].expired()) {
if (shared_instances_[i] == nullptr) {
need_nccl_params = true;
break;
}
Expand Down
5 changes: 2 additions & 3 deletions src/turbomind/triton_backend/llama/LlamaTritonModel.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -108,9 +108,8 @@ struct LlamaTritonModel: public AbstractTransformerModel {

std::shared_ptr<typename ft::LlamaV2<T>::SharedState> shared_state_;

// weak_ptr is used so that the instances get released when all strong references are gone
std::vector<std::weak_ptr<LlamaTritonSharedModelInstance<T>>> shared_instances_;
std::deque<std::mutex> shared_mutexes_; // is locking really needed?
std::vector<std::shared_ptr<LlamaTritonSharedModelInstance<T>>> shared_instances_;
std::deque<std::mutex> shared_mutexes_; // is locking really needed?

bool is_fp16_;
int enable_custom_all_reduce_ = 0;
Expand Down