Wrapper around a CUDA graph.
This API is in beta and may change in future releases.
Begins capturing CUDA work on the current stream.
Ends CUDA graph capture on the current stream. After
replaymay be called on this instance.
Returns an opaque token representing the id of this graph’s memory pool. This id can optionally be passed to another graph’s
capture_begin, which hints the other graph may share the same memory pool.
Replays the CUDA work captured by this graph.
Deletes the graph currently held by this instance.