Implementation of recurrent array memory from "Recurrent Memory Array Structures" Technical Report
https://arxiv.org/abs/1607.03085
It is OK to use this code for research and other non-commercial purposes, please cite the report:
@article{rocki2016recurrent,
title={Recurrent Memory Array Structures},
author={Rocki, Kamil},
journal={arXiv preprint arXiv:1607.03085},
year={2016}
}
to compile (with CUDA):
make PRECISE_MATH=0 cuda
or for precise FP64 (useful for debugging)
make PRECISE_MATH=1 cuda
CUDA version is most recent, CPU versions are outdated, OpenCL version is not really implemented
(Waiting for CUDA 8 final and heterogenous lambdas)
If you still wish to compile the CPU version run:
make cpu
run like this
./deeplstm N B S GPU (test_every_seconds)
example (512 hidden nodes, 100 BPTT steps, batchsize = 64, GPU id = 0, run test every 1000s):
./deeplstm 512 100 64 0 1000
Kamil M Rocki (kmrocki@us.ibm.com)
Copyright (c) 2016, IBM Corporation. All rights reserved.
This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.
You should have received a copy of the GNU General Public License along with this program; if not, write to the Free Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA