141 lines
6.0 KiB
C++
141 lines
6.0 KiB
C++
/*******************************************************************************
|
|
* Copyright (c) 2015-2019 Skymind, Inc.
|
|
*
|
|
* This program and the accompanying materials are made available under the
|
|
* terms of the Apache License, Version 2.0 which is available at
|
|
* https://www.apache.org/licenses/LICENSE-2.0.
|
|
*
|
|
* Unless required by applicable law or agreed to in writing, software
|
|
* distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
|
|
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
|
|
* License for the specific language governing permissions and limitations
|
|
* under the License.
|
|
*
|
|
* SPDX-License-Identifier: Apache-2.0
|
|
******************************************************************************/
|
|
|
|
//
|
|
// @author Yurii Shyrma, created on 14.02.2018
|
|
//
|
|
|
|
// implementation of operation for LSTM cell with peep hole connections:
|
|
// http://www.bioinf.jku.at/publications/older/2604.pdf
|
|
// S. Hochreiter and J. Schmidhuber. "Long Short-Term Memory". Neural Computation, 9(8):1735-1780, 1997.
|
|
// and
|
|
// https://research.google.com/pubs/archive/43905.pdf
|
|
// Hasim Sak, Andrew Senior, and Francoise Beaufays. "Long short-term memory recurrent neural network architectures for large scale acoustic modeling." INTERSPEECH, 2014.
|
|
|
|
|
|
#include <ops/declarable/helpers/lstm.h>
|
|
#include <VariableSpace.h>
|
|
#include <ops/declarable/CustomOperations.h>
|
|
#include <ops/declarable/helpers/transforms.h>
|
|
#include <ops/declarable/helpers/legacy_helpers.h>
|
|
#include <array/NDArrayList.h>
|
|
#include <iterator>
|
|
#include <MmulHelper.h>
|
|
|
|
namespace nd4j {
|
|
namespace ops {
|
|
namespace helpers {
|
|
|
|
/////////////////////////////////////////////////////////////////////////////
|
|
void lstmBlockTimeLoop(const NDArray* maxSeqLength, const NDArray* xSeq, const NDArray* c0, const NDArray* y0,
|
|
const NDArray* W, const NDArray* Wci, const NDArray* Wcf, const NDArray* Wco, const NDArray* b,
|
|
const NDArray* iSeq, const NDArray* cSeq, const NDArray* fSeq, const NDArray* oSeq, const NDArray* zSeq,
|
|
const NDArray* hSeq, const NDArray* ySeq, const std::vector<double>& params, const int dataFormat){
|
|
|
|
int seqLen, bS, nIn, nOut;
|
|
|
|
if(dataFormat == 0) {
|
|
seqLen = xSeq->sizeAt(0);
|
|
bS = xSeq->sizeAt(1);
|
|
nIn = xSeq->sizeAt(2);
|
|
nOut = iSeq->sizeAt(2);
|
|
}
|
|
else if(dataFormat == 1) {
|
|
seqLen = xSeq->sizeAt(2);
|
|
bS = xSeq->sizeAt(0);
|
|
nIn = xSeq->sizeAt(1);
|
|
nOut = iSeq->sizeAt(1);
|
|
}
|
|
else if(dataFormat == 2) {
|
|
seqLen = xSeq->sizeAt(1);
|
|
bS = xSeq->sizeAt(0);
|
|
nIn = xSeq->sizeAt(2);
|
|
nOut = iSeq->sizeAt(2);
|
|
}
|
|
|
|
const std::vector<Nd4jLong> inSliceShape({bS,nIn});
|
|
const std::vector<Nd4jLong> outSliceShape({bS,nOut});
|
|
|
|
auto c_t1 = const_cast<NDArray*>(c0);
|
|
auto y_t1 = const_cast<NDArray*>(y0);
|
|
|
|
// loop through time steps
|
|
for (int t = 0; t < seqLen; ++t) {
|
|
|
|
auto xt = timeSubset(xSeq, t, dataFormat);
|
|
|
|
auto it = timeSubset(iSeq, t, dataFormat);
|
|
auto ct = timeSubset(cSeq, t, dataFormat);
|
|
auto ft = timeSubset(fSeq, t, dataFormat);
|
|
auto ot = timeSubset(oSeq, t, dataFormat);
|
|
auto zt = timeSubset(zSeq, t, dataFormat);
|
|
auto ht = timeSubset(hSeq, t, dataFormat);
|
|
auto yt = timeSubset(ySeq, t, dataFormat);
|
|
|
|
helpers::lstmBlockCell(&xt, c_t1, y_t1, W, Wci, Wcf, Wco, b, &it, &ct, &ft, &ot, &zt, &ht, &yt, params);
|
|
|
|
if(t != 0) {
|
|
delete c_t1;
|
|
delete y_t1;
|
|
}
|
|
|
|
if(t < seqLen - 1) {
|
|
c_t1 = new NDArray(std::move(ct));
|
|
y_t1 = new NDArray(std::move(yt));
|
|
}
|
|
}
|
|
}
|
|
|
|
|
|
|
|
//////////////////////////////////////////////////////////////////////////
|
|
void lstmTimeLoop(nd4j::LaunchContext * context, const NDArray* x, const NDArray* h0, const NDArray* c0, const NDArray* Wx, const NDArray* Wh, const NDArray* Wc, const NDArray* Wp, const NDArray* b,
|
|
NDArray* h, NDArray* c, const std::vector<double>& params) {
|
|
|
|
// x input [time x bS x nIn]
|
|
// h0 initial cell output (at time step = 0) [bS x numProj], in case of projection=false -> numProj == numUnits !!!
|
|
// c0 initial cell state (at time step = 0) [bS x numUnits],
|
|
|
|
// Wx input-to-hidden weights, [nIn x 4*numUnits]
|
|
// Wh hidden-to-hidden weights, [numProj x 4*numUnits]
|
|
// Wc diagonal weights for peephole connections [3*numUnits]
|
|
// Wp projection weights [numUnits x numProj]
|
|
// b biases, [4*numUnits]
|
|
|
|
// h cell outputs [time x bS x numProj], that is per each time step
|
|
// c cell states [time x bS x numUnits] that is per each time step
|
|
|
|
const int time = x->sizeAt(0);
|
|
|
|
NDArray currentH(*h0);
|
|
NDArray currentC(*c0);
|
|
|
|
// loop through time steps
|
|
for (int t = 0; t < time; ++t) {
|
|
auto xt = (*x)({t,t+1, 0,0, 0,0});
|
|
auto ht = (*h)({t,t+1, 0,0, 0,0});
|
|
auto ct = (*c)({t,t+1, 0,0, 0,0});
|
|
|
|
helpers::lstmCell(context, &xt,¤tH,¤tC, Wx,Wh,Wc,Wp, b, &ht, &ct, params);
|
|
currentH.assign(ht);
|
|
currentC.assign(ct);
|
|
}
|
|
}
|
|
|
|
|
|
}
|
|
}
|
|
} |