[Question] Is concurrent read/write across multiple CUDA streams thread-safe in HierarchicalKV?

[HomeHow]

Does HierarchicalKV natively support thread-safe concurrent read and write operations across multiple different CUDA streams on the same instance?

Example Case:

• Host Thread A: Invokes .find(..., stream_A) targeting HashTable_Instance
• Host Thread B: Simultaneously invokes .insert_or_assign(..., stream_B) targeting the exact same HashTable_Instance

Thanks

[9syj]

It is thread-safe when api_lock is true