@tqchen Hi! I have modified the PR, the DLManaged Tensor will manage itself. However, there is a problem in PR(DLManagedTensor manages itself). In this case, ```python import mxnet as mx pack = mx.nd.array([1,2,3]).to_dlpack_for_write() b = mx.nd.from_dlpack(pack) del pack # DLManagedTensor's deleter get called when `pack` get released. print (b) # b != [1,2,3] ```
Does it need to manage the release of dlpack in NDArray Chunk (C++) ? I add `deleter` in `TBlob` and `NDArray::Chunk` in the latest PR (add deleter for TBlob and Chunk in NDArray), and solve the problem. [ Full content available at: https://github.com/apache/incubator-mxnet/pull/12047 ] This message was relayed via gitbox.apache.org for [email protected]
