提交 53da530d 编写于 作者: L Luo Tao

package avg_gpu_backward

上级 1bca3cfd
...@@ -159,4 +159,10 @@ extern void hl_sequence_avg_forward(real* dst, ...@@ -159,4 +159,10 @@ extern void hl_sequence_avg_forward(real* dst,
int width, int width,
const int mode); const int mode);
extern void hl_sequence_avg_backward(real* dst,
real* src,
const int* starts,
int height,
int width,
const int mode);
#endif /* HL_SEQUENCE_H_ */ #endif /* HL_SEQUENCE_H_ */
...@@ -57,4 +57,10 @@ inline void hl_sequence_avg_forward(real* dst, ...@@ -57,4 +57,10 @@ inline void hl_sequence_avg_forward(real* dst,
int width, int width,
const int mode) {} const int mode) {}
inline void hl_sequence_avg_backward(real* dst,
real* src,
const int* starts,
int height,
int width,
const int mode) {}
#endif // HL_SEQUENCE_STUB_H_ #endif // HL_SEQUENCE_STUB_H_
...@@ -325,12 +325,12 @@ __global__ void KeSequenceAvgForward(real* dst, ...@@ -325,12 +325,12 @@ __global__ void KeSequenceAvgForward(real* dst,
int seqLength = end - start; int seqLength = end - start;
if (seqLength == 0) return; if (seqLength == 0) return;
real sum = 0.0; real sum = 0.0;
for (int i = 0; i < seqLength; i++) { for (int i = start; i < end; i++) {
sum += src[(start + i) * width + col]; sum += src[i * width + col];
} }
sum = mode == 1 ? sum : sum = mode == 1 ? sum :
(mode == 0 ? sum / seqLength : sum * my_rsqrt((real)seqLength)); (mode == 0 ? sum / seqLength : sum * my_rsqrt((real)seqLength));
dst[row * width + col] = sum; dst[gid] = sum;
} }
} }
...@@ -354,3 +354,48 @@ void hl_sequence_avg_forward(real* dst, ...@@ -354,3 +354,48 @@ void hl_sequence_avg_forward(real* dst,
(dst, src, starts, height, width, mode); (dst, src, starts, height, width, mode);
CHECK_SYNC("hl_sequence_avg_forward failed"); CHECK_SYNC("hl_sequence_avg_forward failed");
} }
__global__ void KeSequenceAvgBackward(real* dst,
real* src,
const int* starts,
int height,
int width,
const int mode) {
int gid = blockIdx.x * blockDim.x + threadIdx.x;
int row = gid / width;
int col = gid % width;
if (gid < height * width) {
int start = starts[row];
int end = starts[row + 1];
int seqLength = end - start;
if (seqLength == 0) return;
real grad = src[gid];
grad = mode == 1 ? grad :
(mode == 0 ? grad / seqLength : grad * my_rsqrt((real)seqLength));
for (int i = start; i < end; i++) {
dst[i * width + col] += grad;
}
}
}
void hl_sequence_avg_backward(real* dst,
real* src,
const int* starts,
int height,
int width,
const int mode) {
CHECK_NOTNULL(dst);
CHECK_NOTNULL(src);
CHECK_NOTNULL(starts);
int block = 512;
int grid = DIVUP(width * height, 512);
CHECK(mode == 0 || mode == 1 || mode == 2)
<< "mode error in hl_sequence_avg_backward!";
KeSequenceAvgBackward<<< grid, block, 0, STREAM_DEFAULT >>>
(dst, src, starts, height, width, mode);
CHECK_SYNC("hl_sequence_avg_backward failed");
}
...@@ -26,8 +26,6 @@ bool AverageLayer::init(const LayerMap& layerMap, ...@@ -26,8 +26,6 @@ bool AverageLayer::init(const LayerMap& layerMap,
const ParameterMap& parameterMap) { const ParameterMap& parameterMap) {
SequencePoolLayer::init(layerMap, parameterMap); SequencePoolLayer::init(layerMap, parameterMap);
dataMtx_ = Matrix::create(nullptr, 1, 1, false, useGpu_);
outMtx_ = Matrix::create(nullptr, 1, getSize(), false, useGpu_);
// average strategy // average strategy
if (config_.average_strategy() == "average") { if (config_.average_strategy() == "average") {
mode_ = kAverage; mode_ = kAverage;
...@@ -60,43 +58,9 @@ void AverageLayer::forward(PassType passType) { ...@@ -60,43 +58,9 @@ void AverageLayer::forward(PassType passType) {
void AverageLayer::backward(const UpdateCallback& callback) { void AverageLayer::backward(const UpdateCallback& callback) {
SequencePoolLayer::backward(callback); SequencePoolLayer::backward(callback);
const int* starts = startPositions_->getData(false); if (getInputGrad(0)) {
MatrixPtr grad = getInputGrad(0); getInputGrad(0)->sequenceAvgBackward(
*getOutputGrad(), *startPositions_->getVector(useGpu_), mode_);
if (grad) {
size_t dim = getSize();
real* gradientData = getInputGrad(0)->getData();
real* gradient = getOutputGrad()->getData();
size_t numSequences = startPositions_->getSize() - 1;
for (size_t sequenceId = 0; sequenceId < numSequences; ++sequenceId) {
// TODO(Dangqingqing) optimization for GPU
int sequenceLength = starts[sequenceId + 1] - starts[sequenceId];
if (0 == sequenceLength) {
// empty sequence
continue;
}
dataMtx_->setData(
gradientData + starts[sequenceId] * dim, sequenceLength, dim);
outMtx_->setData(gradient + sequenceId * dim);
switch (mode_) {
case kAverage: {
// plain average
dataMtx_->addBias(*outMtx_, 1.0f / sequenceLength);
break;
}
case kSum: {
// sum instead of average
dataMtx_->addBias(*outMtx_, 1.0f);
break;
}
case kAverageSquareRootN: {
// divide by square root of sequenceLength
dataMtx_->addBias(*outMtx_, 1.0f / sqrt(sequenceLength));
break;
}
default: { LOG(FATAL) << "should not reach here"; }
}
}
} }
} }
......
...@@ -45,8 +45,6 @@ public: ...@@ -45,8 +45,6 @@ public:
void backward(const UpdateCallback& callback = nullptr) override; void backward(const UpdateCallback& callback = nullptr) override;
protected: protected:
MatrixPtr outMtx_;
MatrixPtr dataMtx_;
int mode_; int mode_;
}; };
} // namespace paddle } // namespace paddle
...@@ -483,6 +483,20 @@ void GpuMatrix::sequenceAvgForward(Matrix& a, ...@@ -483,6 +483,20 @@ void GpuMatrix::sequenceAvgForward(Matrix& a,
hl_sequence_avg_forward(dst, src, starts, height, width, mode); hl_sequence_avg_forward(dst, src, starts, height, width, mode);
} }
void GpuMatrix::sequenceAvgBackward(Matrix& a,
const IVector& startsPos,
int mode) {
size_t height = a.getHeight();
size_t width = getWidth();
CHECK_EQ(height, startsPos.getSize() - 1);
CHECK_EQ(width, a.getWidth());
real* dst = getData();
real* src = a.getData();
const int* starts = startsPos.getData();
hl_sequence_avg_backward(dst, src, starts, height, width, mode);
}
/* this = scaleAB*(a*b) + scaleT*this */ /* this = scaleAB*(a*b) + scaleT*this */
void GpuMatrix::mul(const GpuMatrix& a, void GpuMatrix::mul(const GpuMatrix& a,
const GpuMatrix& b, const GpuMatrix& b,
...@@ -2304,6 +2318,41 @@ void CpuMatrix::sequenceAvgForward(Matrix& a, ...@@ -2304,6 +2318,41 @@ void CpuMatrix::sequenceAvgForward(Matrix& a,
} }
} }
void CpuMatrix::sequenceAvgBackward(Matrix& a,
const IVector& startsPos,
int mode) {
size_t height = a.getHeight();
size_t width = getWidth();
CHECK_EQ(height, startsPos.getSize() - 1);
CHECK_EQ(width, a.getWidth());
real* dst = getData();
real* src = a.getData();
const int* starts = startsPos.getData();
MatrixPtr outMtx = Matrix::create(nullptr, 1, width, false, false);
MatrixPtr dataMtx = Matrix::create(nullptr, 1, width, false, false);
for (size_t i = 0; i < height; ++i) {
int sequenceLength = starts[i + 1] - starts[i];
if (0 == sequenceLength) {
// empty sequence
continue;
}
outMtx->setData(dst + starts[i] * width, sequenceLength, width);
dataMtx->setData(src + i * width);
if (mode == 0) {
// plain average
outMtx->addBias(*dataMtx, 1.0f / sequenceLength);
} else if (mode == 1) {
// sum instead of average
outMtx->addBias(*dataMtx, 1.0f);
} else if (mode == 2) {
// divide by square root of sequenceLength
outMtx->addBias(*dataMtx, 1.0f / std::sqrt(sequenceLength));
} else {
LOG(FATAL) << "should not reach here";
}
}
}
/* this = scaleAB*(a*b) + scaleT*this*/ /* this = scaleAB*(a*b) + scaleT*this*/
void CpuMatrix::mul(const Matrix& a, void CpuMatrix::mul(const Matrix& a,
const Matrix& b, const Matrix& b,
......
...@@ -461,6 +461,12 @@ public: ...@@ -461,6 +461,12 @@ public:
LOG(FATAL) << "Not implemented"; LOG(FATAL) << "Not implemented";
} }
virtual void sequenceAvgBackward(Matrix& a,
const IVector& startsPos,
int mode) {
LOG(FATAL) << "Not implemented";
}
/** /**
* @code * @code
* this = scaleAB*(a*b) + scaleT*this * this = scaleAB*(a*b) + scaleT*this
...@@ -1203,6 +1209,7 @@ public: ...@@ -1203,6 +1209,7 @@ public:
void collectSharedBias(Matrix& a, real scale); void collectSharedBias(Matrix& a, real scale);
void sequenceAvgForward(Matrix& a, const IVector& startsPos, int mode); void sequenceAvgForward(Matrix& a, const IVector& startsPos, int mode);
void sequenceAvgBackward(Matrix& a, const IVector& startsPos, int mode);
/** /**
* @code * @code
...@@ -1619,6 +1626,7 @@ public: ...@@ -1619,6 +1626,7 @@ public:
void collectSharedBias(Matrix& a, real scale); void collectSharedBias(Matrix& a, real scale);
void sequenceAvgForward(Matrix& a, const IVector& startsPos, int mode); void sequenceAvgForward(Matrix& a, const IVector& startsPos, int mode);
void sequenceAvgBackward(Matrix& a, const IVector& startsPos, int mode);
/** /**
* @code * @code
......
...@@ -685,7 +685,7 @@ TEST(SMatrix, topK) { ...@@ -685,7 +685,7 @@ TEST(SMatrix, topK) {
} }
} }
void testMatrixSequenceAvgForward(int batchSize, int inputDim, int mode) { void testMatrixSequenceAvg(int batchSize, int inputDim, int mode) {
MatrixPtr cpuInput = std::make_shared<CpuMatrix>(batchSize, inputDim); MatrixPtr cpuInput = std::make_shared<CpuMatrix>(batchSize, inputDim);
MatrixPtr gpuInput = std::make_shared<GpuMatrix>(batchSize, inputDim); MatrixPtr gpuInput = std::make_shared<GpuMatrix>(batchSize, inputDim);
cpuInput->randomizeUniform(); cpuInput->randomizeUniform();
...@@ -706,15 +706,25 @@ void testMatrixSequenceAvgForward(int batchSize, int inputDim, int mode) { ...@@ -706,15 +706,25 @@ void testMatrixSequenceAvgForward(int batchSize, int inputDim, int mode) {
gpuOutput->sequenceAvgForward(*gpuInput, *gpuSequence, mode); gpuOutput->sequenceAvgForward(*gpuInput, *gpuSequence, mode);
TensorCheckErr(*cpuOutput, *gpuOutput); TensorCheckErr(*cpuOutput, *gpuOutput);
MatrixPtr cpuInGrad = std::make_shared<CpuMatrix>(batchSize, inputDim);
MatrixPtr gpuInGrad = std::make_shared<GpuMatrix>(batchSize, inputDim);
cpuInGrad->randomizeUniform();
gpuInGrad->copyFrom(*cpuInGrad);
cpuInGrad->sequenceAvgBackward(*cpuOutput, *cpuSequence, mode);
gpuInGrad->sequenceAvgBackward(*gpuOutput, *gpuSequence, mode);
TensorCheckErr(*cpuInGrad, *gpuInGrad);
} }
TEST(Matrix, sequenceAvgForward) { TEST(Matrix, sequenceAvg) {
for (auto batchSize : {10, 128, 6000}) { for (auto batchSize : {10, 128, 6000}) {
for (auto inputDim : {32, 100, 512}) { for (auto inputDim : {32, 100, 512}) {
for (auto mode : {0, 1, 2}) { for (auto mode : {0, 1, 2}) {
VLOG(3) << " batchSize=" << batchSize << " inputDim=" << inputDim VLOG(3) << " batchSize=" << batchSize << " inputDim=" << inputDim
<< " mode=" << mode; << " mode=" << mode;
testMatrixSequenceAvgForward(batchSize, inputDim, mode); testMatrixSequenceAvg(batchSize, inputDim, mode);
} }
} }
} }
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册