My point is that the DLManagerd Tensor should manage its own deletion and you do not have to add dlpack field for mx.nd.NDArray. Even when the frontend NDArray release, the internal data will be kept alive until the pycapsule get released and the DLManagedTensor's deleter get called
[ Full content available at: https://github.com/apache/incubator-mxnet/pull/12047 ] This message was relayed via gitbox.apache.org for [email protected]
