leezu commented on a change in pull request #17841:
URL: https://github.com/apache/incubator-mxnet/pull/17841#discussion_r416854988
##########
File path: include/mxnet/io.h
##########
@@ -113,5 +120,98 @@ struct DataIteratorReg
*/
#define MXNET_REGISTER_IO_ITER(name) \
DMLC_REGISTRY_REGISTER(::mxnet::DataIteratorReg, DataIteratorReg, name)
+
+/*!
+ * \brief A random accessable dataset which provides GetLen() and GetItem().
+ * Unlike DataIter, it's a static lookup storage which is friendly to random
access.
+ * The dataset itself should NOT contain data processing, which should be
applied during
+ * data augmentation or transformation processes.
+ */
+class Dataset {
+ public:
+ /*!
+ * \brief Get the size of the dataset
+ */
+ virtual uint64_t GetLen(void) const = 0;
+ /*!
+ * \brief Create a copy of dataset for threaded worker
+ */
+ virtual Dataset* Clone(void) const = 0;
Review comment:
Is this used? If not, delete. If yes, why not use c++ copy mechanism?
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
[email protected]
