Given this PR is to support N-D tensor, I wouldn't call an N-D array of size greater than 2^31-1 a special case though. Using int here simply leaves door for an overflow bug in the future. A common way in this module to index array size is to use the defined type index_t. Please see https://github.com/apache/incubator-mxnet/blob/ba8a9d13e1b549d061f1933c463cfad5e7bdd7aa/src/operator/nn/softmax-inl.h#L57
[ Full content available at: https://github.com/apache/incubator-mxnet/pull/12430 ] This message was relayed via gitbox.apache.org for [email protected]
