eric-haibin-lin commented on issue #8817: Creating a monodimensional sparse array returns an error URL: https://github.com/apache/incubator-mxnet/issues/8817#issuecomment-350393045 From an implementation point of view, it's hard to enable such mono-dimensional array. The 1-D shape could be passed directly to the backend, whereas all the sparse operators in the backend assumes `shape[0]` is num_rows and `shape[1]` is num_cols at the moment. Accepting 1-D shape for csr would require **significant efforts** to check these cases for each operator in the backend. `zeros` is not the only operator affected, cases include: - `mx.sym.var('a', shape=10, stype='csr')` - generate a random uniform csr ndarray with 1-D shape - dot with 1-D csr - convert a dense 1-D array to csr - etc.. I suggest that we throw an exception for 1-D shape for `sparse.zeros` to guide users the expected usage instead.
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
