AlenkaF commented on issue #38868: URL: https://github.com/apache/arrow/issues/38868#issuecomment-1827491085
> The first option. I'm not sure the second is an option - I'm assuming non-contiguous (chunked) arrays can't be represented as DLPack tensors? Ah, maybe the issue desc isn't clear. What I meant was, are we talking about chunks of tensors or chunks of primitive arrays. Then we can make it clearer and just talk about tensor extension arrays vs primitive arrays, as chunking can't be supported because DLPack needs a contiguous buffer. So for the second part, for primitive arrays (not chunked), this is already being worked on in https://github.com/apache/arrow/issues/33984. What I would really like to see is to have DLPack support for `FixedShapeTensorArray` also (or nested in general?). This is the case I was having in mind when opening this issue and the question I think is: would it be correct to present the whole tensor extension storage array as a DLPack tensor or would it be more correct to represent specific tensor in the `FixedShapeTensorArray` (list element) as a DLPack tensor? -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
