1Revision history for Perl extension AI::MXNet 21.5 Sun Feb 16 19:56:17 PST 2020 3 - Runtime features 4 - INT64 Tensor support 5 61.4 Mon Feb 18 11:54:07 PST 2019 7 - Two more gluon loss classes 8 - Visualization fixes 9 - Gluon rnn rework, including hybridization 10 - Exposed GPU memory info to perl level. 11 121.33 Thu Oct 4 13:25:56 PDT 2018 13 - Added randn function. 14 - Internal SELU function on C++ layer. 15 - Predict now accepts ndarray as well. 16 - Gluon: Only warn when the blocks are unregistered. 17 - Gluon: Better sparse support. 18 - Gluon: Improved block summary. 19 - Added validation docs for MXNet installation for Perl. 20 - Flexible perl env for examples. 21 - Gluon: Custom dtypes for the symbol block 22 - Separate eval metric for the epoch level. 23 241.32 Sun Aug 5 14:25:31 PDT 2018 25 - Several new metric classes 26 - Expanded documentation 27 - Bugfixes. 28 291.31 Tue Jul 10 21:19:13 PDT 2018 30 - Memory leak fix for Gluon API 31 - Added summary function for Gluon models 32 - Improved graphviz vizualizations 33 - Added artistic style transfer example for Gluon API 34 351.3 Tue Jun 26 20:57:40 PDT 2018 36 - Major Gluon update towards parity with Python's API. 37 - Miscellaneous bugfixes and improvements. 38 - New Engine API. 39 - Module::reshape moved to C++ backend. 40 - Examples were updated to work on multi-gpu boxes 41 421.23 Thu Apr 19 15:38:10 PDT 2018 43 - Support for image operations on symbols and ndarrays. 44 451.22 Sat Apr 14 17:51:55 PDT 2018 46 - Parity with Python Gluon Loss classes 47 481.21 Sun Apr 8 12:08:44 PDT 2018 49 - Support for linear algebra operations on symbols and ndarrays. 50 511.2 Sun Mar 4 16:29:19 PST 2018 52 - Support for sparse tensors 53 541.1 Sun Oct 1 10:19:08 PDT 2017 55 - Major update, added support for new imperative MXNet interface Gluon and realtime GPU kernels callable from Perl space. 56 - Bugfixes for distributed training. 57 - Miscellaneous fixes and performance improvements. 58 591.0102 Sun Aug 6 16:55:08 PDT 2017 60 - bugfixes in Image.pm, updated tests, added PearsonCorrelation metric, added Convolutional RNN modules. 61 621.0101 Sun Jul 2 17:16:01 PDT 2017 63 - reworked CachedOp, two new optimizers, auto module reshape, using strings to index the kvstore. 64 651.01 Sat Jun 10 23:57:27 PDT 2017 66 - sync with python. 67 680.9507 Thu May 11 17:04:44 PDT 2017 69 - added AutoGrad, bugfixes. 70 710.9506 Sat Apr 29 20:26:50 PDT 2017 72 - Ftrl optimizer, new tests, bugfixes. 73 740.9505 Sun Apr 23 21:26:04 PDT 2017 75 - Perplexity bugfix, two new examples. 76 770.9504 Wed Apr 19 18:59:45 PDT 2017 78 - LR Scheduler bugfix. 79 800.9503 Wed Apr 19 13:33:57 PDT 2017 81 - added an example of generation of inferred text via pre-trained RNN. 82 - bugfixes/tests. 83 840.9502 Sat Apr 15 17:18:21 PDT 2017 85 - optimizations/bugfixes. 86 870.9501 Sat Apr 8 13:01:00 PDT 2017 88 - ZoneoutCell, nd inferred reshape and moveaxis, cosmetic changes to Image iter, 89 pod reworked to be readable via metacpan. 90 910.95 Sun Mar 26 17:42:02 PDT 2017 92 - docs, bugfixes, tests in order to be visible on https://mxnet.io 93 940.03 Tue Feb 14 07:28:11 PST 2017 95 - sync up with current state of the Python inteface. 96 - high level RNN support. 97 980.02 Tue Feb 14 07:28:11 PST 2017 99 - prepared for inclusion to the mxnet code repository. 100 101