From f2c4bb679bb3f247c12888b878ec4dddff358e49 Mon Sep 17 00:00:00 2001 From: Yibing Liu Date: Sun, 21 Jan 2018 09:24:40 -0800 Subject: [PATCH] Add lstm with recurrent projection operator --- paddle/operators/lstmp_op.cc | 296 ++++++++++++++ paddle/operators/lstmp_op.cu.cc | 24 ++ paddle/operators/lstmp_op.h | 384 ++++++++++++++++++ python/paddle/v2/fluid/tests/test_lstmp_op.py | 314 ++++++++++++++ 4 files changed, 1018 insertions(+) create mode 100644 paddle/operators/lstmp_op.cc create mode 100644 paddle/operators/lstmp_op.cu.cc create mode 100644 paddle/operators/lstmp_op.h create mode 100644 python/paddle/v2/fluid/tests/test_lstmp_op.py diff --git a/paddle/operators/lstmp_op.cc b/paddle/operators/lstmp_op.cc new file mode 100644 index 00000000000..4c7f7713eea --- /dev/null +++ b/paddle/operators/lstmp_op.cc @@ -0,0 +1,296 @@ +/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserve. + +Licensed under the Apache License, Version 2.0 (the "License"); +you may not use this file except in compliance with the License. +You may obtain a copy of the License at + +http://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. */ + +#include "paddle/operators/lstmp_op.h" + +namespace paddle { +namespace operators { + +class LSTMPOp : public framework::OperatorWithKernel { + public: + using framework::OperatorWithKernel::OperatorWithKernel; + + void InferShape(framework::InferShapeContext* ctx) const override { + PADDLE_ENFORCE(ctx->HasInput("Input"), + "Input(Input) of LSTMP should not be null."); + PADDLE_ENFORCE(ctx->HasInput("Weight"), + "Input(Weight) of LSTMP should not be null."); + PADDLE_ENFORCE(ctx->HasInput("ProjWeight"), + "Input(ProjWeight) of LSTMP should not be null."); + PADDLE_ENFORCE(ctx->HasInput("Bias"), + "Input(Bias) of LSTMP should not be null."); + + PADDLE_ENFORCE(ctx->HasOutput("Projection"), + "Output(Projection) of LSTMP should not be null."); + PADDLE_ENFORCE(ctx->HasOutput("Cell"), + "Output(Cell) of LSTMP should not be null."); + PADDLE_ENFORCE(ctx->HasOutput("BatchGate"), + "Output(BatchGate) of LSTMP should not be null."); + PADDLE_ENFORCE(ctx->HasOutput("BatchCellPreAct"), + "Output(BatchGate) of LSTMP should not be null."); + + auto in_dims = ctx->GetInputDim("Input"); + PADDLE_ENFORCE_EQ(in_dims.size(), 2, "Input(X)'s rank must be 2."); + + if (ctx->HasInput("H0")) { + PADDLE_ENFORCE(ctx->HasInput("C0"), + "Input(C0) and Input(H0) of LSTMP should not " + "be null at the same time."); + auto h_dims = ctx->GetInputDim("H0"); + auto c_dims = ctx->GetInputDim("C0"); + PADDLE_ENFORCE(h_dims == c_dims, + "The dimension of Input(H0) and Input(C0) " + "should be the same."); + } + + int frame_size = in_dims[1] / 4; + auto w_dims = ctx->GetInputDim("Weight"); + auto proj_dims = ctx->GetInputDim("ProjWeight"); + PADDLE_ENFORCE_EQ(w_dims.size(), 2, + "The rank of Input(Weight) should be 2."); + PADDLE_ENFORCE_EQ(w_dims[0], proj_dims[1], + "The first dimension of Input(Weight) " + "should be %d.", + proj_dims[1]); + PADDLE_ENFORCE_EQ(w_dims[1], 4 * frame_size, + "The second dimension of Input(Weight) " + "should be 4 * %d.", + frame_size); + + PADDLE_ENFORCE_EQ(proj_dims.size(), 2, + "The rank of Input(ProjWeight) should be 2."); + PADDLE_ENFORCE_EQ(proj_dims[0], frame_size, + "The first dimension of Input(ProjWeight) " + "should be %d.", + frame_size); + + auto b_dims = ctx->GetInputDim("Bias"); + PADDLE_ENFORCE_EQ(b_dims.size(), 2, "The rank of Input(Bias) should be 2."); + PADDLE_ENFORCE_EQ(b_dims[0], 1, + "The first dimension of Input(Bias) should be 1."); + + if (ctx->Attrs().Get("use_peepholes")) { + PADDLE_ENFORCE_EQ(b_dims[1], 7 * frame_size, + "The second dimension of Input(Bias) should be " + "7 * %d if enable peepholes connection", + frame_size); + } else { + PADDLE_ENFORCE_EQ(b_dims[1], 4 * frame_size, + "The second dimension of Input(Bias) should be " + "4 * %d if disable peepholes connection", + frame_size); + } + + framework::DDim out_dims({in_dims[0], frame_size}); + framework::DDim proj_out_dims({in_dims[0], proj_dims[1]}); + ctx->SetOutputDim("Projection", proj_out_dims); + ctx->SetOutputDim("Cell", out_dims); + ctx->SetOutputDim("BatchGate", in_dims); + ctx->SetOutputDim("BatchCellPreAct", out_dims); + ctx->ShareLoD("Input", "Projection"); + ctx->ShareLoD("Input", "Cell"); + } + + protected: + framework::OpKernelType GetExpectedKernelType( + const framework::ExecutionContext& ctx) const override { + return framework::OpKernelType( + framework::ToDataType(ctx.Input("Input")->type()), + ctx.device_context()); + } +}; + +class LSTMPOpMaker : public framework::OpProtoAndCheckerMaker { + public: + LSTMPOpMaker(OpProto* proto, OpAttrChecker* op_checker) + : OpProtoAndCheckerMaker(proto, op_checker) { + AddInput("Input", + "(LoDTensor) the first input is a LodTensor, which support " + "variable-time length input sequence. The underlying tensor in " + "this LoDTensor is a matrix with shape (T X 4D), where T is the " + "total time steps in this mini-batch, D is the hidden size."); + AddInput("H0", + "(Tensor, optional) the initial hidden state is an optional " + "input. This is a tensor with shape (N x D), where N is the " + "batch size and D is the hidden size.") + .AsDispensable(); + AddInput("C0", + "(Tensor, optional) the initial cell state is an optional " + "input. This is a tensor with shape (N x D), where N is the " + "batch size. `H0` and `C0` can be NULL but only at the same time") + .AsDispensable(); + AddInput("Weight", + "(Tensor) the learnable hidden-hidden weights." + " - The shape is (P x 4D), where P is the recurrent projection " + "layer size and D is the hidden size. " + " - Weight = {W_cr, W_ir, W_fr, W_or}"); + AddInput("ProjWeight", + "(Tensor) the learnable weight `W_rh` of the projection layer." + " - The shape is (D x P), where P is the recurrent projection " + "layer size and D is the hidden size."); + AddInput("Bias", + "(Tensor) the learnable weights, which contains two parts: " + "input-hidden bias weight and peephole connections weight if " + "setting `use_peepholes` True. " + "1. `use_peepholes = False` " + " - The shape is (1 x 4D). " + " - Bias = {b_c, b_i, b_f, b_o}." + "2. `use_peepholes = True` " + " - The shape is (1 x 7D). " + " - Bias = {b_c, b_i, b_f, b_o, W_ic, W_fc, W_oc}."); + AddOutput("Projection", + "(LoDTensor) the projection of the hidden state of LSTMP " + "operator. The shape is (T x P), and lod is the same with the " + "`Input`."); + AddOutput("Cell", + "(LoDTensor) the cell state of LSTMP operator. " + "The shape is (T x D), and lod is the same with the `Input`."); + AddOutput("BatchGate", + "(LoDTensor) This LoDTensor contains input gate, forget gate " + "and output gate after the nonlinear computation. This " + "LoDTensor has the same shape as the reorganized input, which " + "is also be called batch input. The LoD size is 2. The first " + "LoD is the batch offsets and the second LoD contains the " + "indexes, which denote the position of reorganized sequence " + "in the raw input.") + .AsIntermediate(); + AddOutput("BatchCellPreAct", + "(LoDTensor) This LoDTensor is obtained in the forward and used " + "in the backward.") + .AsIntermediate(); + AddAttr("use_peepholes", + "(bool, defalut: True) " + "whether to enable diagonal/peephole connections.") + .SetDefault(true); + AddAttr("is_reverse", + "(bool, defalut: False) " + "whether to compute reversed LSTMP.") + .SetDefault(false); + AddAttr( + "gate_activation", + "(string, default: sigmoid)" + "The activation for input gate, forget gate and output " + "gate, `sigmoid` by default.") + .SetDefault("sigmoid") + .InEnum({"sigmoid", "tanh", "relu", "identity"}); + AddAttr("cell_activation", + "(string, default: tanh)" + "The activation for cell output, `tanh` by defalut.") + .SetDefault("tanh") + .InEnum({"sigmoid", "tanh", "relu", "identity"}); + AddAttr("candidate_activation", + "(string, default: tanh)" + "The activation for candidate hidden state, " + "`tanh` by default.") + .SetDefault("tanh") + .InEnum({"sigmoid", "tanh", "relu", "identity"}); + AddComment(R"DOC( +Long-Short Term Memory with Recurrent Projection (LSTMP) Operator. + +LATMP is stand LSTM appended by a recurrent projection layer to reduce the +number of parameters, espeacially when the output size is relative large. +The formula is as follows: + +$$ +i_t = \sigma(W_{ix}x_{t} + W_{ih}r_{t-1} + W_{ic}c_{t-1} + b_i) \\ + +f_t = \sigma(W_{fx}x_{t} + W_{fh}r_{t-1} + W_{fc}c_{t-1} + b_f) \\ + +c_t = f_t \odot c_{t-1} + i_t \odot act_g(W_{cx}x_t + W_{ch}r_{t-1} + b_c) \\ + +o_t = \sigma(W_{ox}x_{t} + W_{oh}r_{t-1} + W_{oc}c_t + b_o) \\ + +h_t = o_t \odot act_h(c_t) + +r_t = W_{rh}h_t +$$ + +where the W terms denote weight matrices (e.g. $W_{xi}$ is the matrix +of weights from the input gate to the input), $W_{ic}, W_{fc}, W_{oc}$ +are diagonal weight matrices for peephole connections. In our implementation, +we use vectors to reprenset these diagonal weight matrices. The b terms +denote bias vectors ($b_i$ is the input gate bias vector), $\sigma$ +is the non-line activations, such as logistic sigmoid function, and +$i, f, o$ and $c$ are the input gate, forget gate, output gate, +and cell activation vectors, respectively, all of which have the same size as +the cell output activation vector $h$. $r$ denotes the recurrent projection +layer. + +The $\odot$ is the element-wise product of the vectors. $act_g$ and $act_h$ +are the cell input and cell output activation functions and `tanh` is usually +used for them. + +Note that these $W_{xi}x_{t}, W_{xf}x_{t}, W_{xc}x_{t}, W_{xo}x_{t}$ +operations on the input $x_{t}$ are NOT included in this operator. +Users can choose to use fully-connect operator before LSTMP operator. + +)DOC"); + } +}; + +class LSTMPGradOp : public framework::OperatorWithKernel { + public: + using framework::OperatorWithKernel::OperatorWithKernel; + + void InferShape(framework::InferShapeContext* ctx) const override { + PADDLE_ENFORCE(ctx->HasInput("Input"), + "Input(Input) of LSTMP should not be null."); + PADDLE_ENFORCE(ctx->HasInput("Hidden"), + "Input(Hidden) of LSTMP should not be null."); + PADDLE_ENFORCE(ctx->HasInput("Cell"), + "Input(Cell) of LSTMP should not be null."); + PADDLE_ENFORCE(ctx->HasInput("Weight"), + "Input(Weight) of LSTMP should not be null."); + PADDLE_ENFORCE(ctx->HasInput("Bias"), + "Input(Bias) of LSTMP should not be null."); + + PADDLE_ENFORCE(ctx->HasInput("BatchGate"), + "Input(BatchGate) of LSTMP should not be null."); + PADDLE_ENFORCE(ctx->HasInput("BatchCellPreAct"), + "Input(BatchGate) of LSTMP should not be null."); + + auto SetOutGradDim = [&ctx](const std::string& name) { + auto g_name = framework::GradVarName(name); + if (ctx->HasOutput(g_name)) + ctx->SetOutputDim(g_name, ctx->GetInputDim(name)); + }; + + SetOutGradDim("Input"); + SetOutGradDim("Weight"); + SetOutGradDim("Bias"); + SetOutGradDim("H0"); + SetOutGradDim("C0"); + } + + protected: + framework::OpKernelType GetExpectedKernelType( + const framework::ExecutionContext& ctx) const override { + return framework::OpKernelType( + framework::ToDataType(ctx.Input("Input")->type()), + ctx.device_context()); + } +}; + +} // namespace operators +} // namespace paddle + +namespace ops = paddle::operators; +REGISTER_OP(lstmp, ops::LSTMPOp, ops::LSTMPOpMaker, lstmp_grad, + ops::LSTMPGradOp); +REGISTER_OP_CPU_KERNEL( + lstmp, ops::LSTMPKernel, + ops::LSTMPKernel); +REGISTER_OP_CPU_KERNEL( + lstmp_grad, ops::LSTMPGradKernel, + ops::LSTMPGradKernel); diff --git a/paddle/operators/lstmp_op.cu.cc b/paddle/operators/lstmp_op.cu.cc new file mode 100644 index 00000000000..7fcbcfecc87 --- /dev/null +++ b/paddle/operators/lstmp_op.cu.cc @@ -0,0 +1,24 @@ +/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserve. + +Licensed under the Apache License, Version 2.0 (the "License"); +you may not use this file except in compliance with the License. +You may obtain a copy of the License at + + http://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. */ + +#include "paddle/operators/lstmp_op.h" + +namespace ops = paddle::operators; +REGISTER_OP_CUDA_KERNEL( + lstmp, ops::LSTMPKernel, + ops::LSTMPKernel); +REGISTER_OP_CUDA_KERNEL( + lstmp_grad, + ops::LSTMPGradKernel, + ops::LSTMPGradKernel); diff --git a/paddle/operators/lstmp_op.h b/paddle/operators/lstmp_op.h new file mode 100644 index 00000000000..f5a38b2ff50 --- /dev/null +++ b/paddle/operators/lstmp_op.h @@ -0,0 +1,384 @@ +/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserve. + +Licensed under the Apache License, Version 2.0 (the "License"); +you may not use this file except in compliance with the License. +You may obtain a copy of the License at + +http://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. */ + +#pragma once +#include "paddle/framework/op_registry.h" +#include "paddle/operators/math/detail/activation_functions.h" +#include "paddle/operators/math/lstm_compute.h" +#include "paddle/operators/math/math_function.h" +#include "paddle/operators/math/sequence2batch.h" + +namespace paddle { +namespace operators { + +using LoDTensor = framework::LoDTensor; +using Tensor = framework::Tensor; + +template +inline void ReorderInitState(const DeviceContext& ctx, + const framework::Tensor& src, const size_t* index, + framework::Tensor* dst, bool indexed_src) { + math::CopyMatrixRowsFunctor row_shuffle; + dst->mutable_data(src.dims(), ctx.GetPlace()); + row_shuffle(ctx, src, index, *dst, indexed_src); +} + +template +class LSTMPKernel : public framework::OpKernel { + public: + void Compute(const framework::ExecutionContext& ctx) const override { + auto* input = ctx.Input("Input"); + auto* weight = ctx.Input("Weight"); + auto* proj_weight = ctx.Input("ProjWeight"); + auto* bias = ctx.Input("Bias"); + + auto* hidden_t0 = ctx.Input("H0"); + auto* cell_t0 = ctx.Input("C0"); + + auto* batch_gate = ctx.Output("BatchGate"); + batch_gate->mutable_data(ctx.GetPlace()); + auto* proj_out = ctx.Output("Projection"); + proj_out->mutable_data(ctx.GetPlace()); + auto* cell_out = ctx.Output("Cell"); + cell_out->mutable_data(ctx.GetPlace()); + + bool is_reverse = ctx.Attr("is_reverse"); + math::LoDTensor2BatchFunctor to_batch; + auto& device_ctx = ctx.template device_context(); + to_batch(device_ctx, *input, *batch_gate, true, is_reverse); + + auto in_dims = input->dims(); + int frame_size = static_cast(in_dims[1] / 4); + framework::DDim dims({in_dims[0], frame_size}); + framework::DDim proj_dims({in_dims[0], proj_weight->dims()[1]}); + + if (bias) { + Tensor b = *bias; + b.Resize({bias->numel(), 1}); + Tensor gate_bias = b.Slice(0, 4 * frame_size); + math::RowwiseAdd add_bias; + add_bias(device_ctx, *batch_gate, gate_bias, batch_gate); + } + + math::LstmMetaValue lstmp_value; + if (bias && ctx.Attr("use_peepholes")) { + T* bias_data = const_cast(bias->data()); + // the code style in LstmpMetaValue will be updated later. + + lstmp_value.check_ig = bias_data + 4 * frame_size; + lstmp_value.check_fg = lstmp_value.check_ig + frame_size; + lstmp_value.check_og = lstmp_value.check_fg + frame_size; + } else { + lstmp_value.check_ig = nullptr; + lstmp_value.check_fg = nullptr; + lstmp_value.check_og = nullptr; + } + lstmp_value.prev_state_value = nullptr; + Tensor ordered_c0; + const size_t* order = batch_gate->lod()[2].data(); + if (cell_t0) { + // Since the batch computing for LSTMP reorders the input sequence + // according to their length. The initialized cell state also needs + // to reorder. + ReorderInitState(device_ctx, *cell_t0, order, + &ordered_c0, true); + lstmp_value.prev_state_value = ordered_c0.data(); + } + + // Use the local variable as here. + LoDTensor batch_hidden, batch_proj, batch_cell; + auto* batch_cell_pre_act = ctx.Output("BatchCellPreAct"); + batch_hidden.mutable_data(dims, ctx.GetPlace()); // T x D + batch_proj.mutable_data(proj_dims, ctx.GetPlace()); // T x P + batch_cell.mutable_data(dims, ctx.GetPlace()); // T x D + batch_cell_pre_act->mutable_data(dims, ctx.GetPlace()); + + auto batch_starts = batch_gate->lod()[0]; + size_t num_batch = batch_starts.size() - 1; + auto gate_act = math::detail::GetActivationType( + ctx.Attr("gate_activation")); + auto cell_act = math::detail::GetActivationType( + ctx.Attr("cell_activation")); + auto cand_act = math::detail::GetActivationType( + ctx.Attr("candidate_activation")); + + for (size_t n = 0; n < num_batch; n++) { + int bstart = static_cast(batch_starts[n]); + int bend = static_cast(batch_starts[n + 1]); + + Tensor gate_t = batch_gate->Slice(bstart, bend); + Tensor hidden_t = batch_hidden.Slice(bstart, bend); + Tensor proj_t = batch_proj.Slice(bstart, bend); + Tensor cell_t = batch_cell.Slice(bstart, bend); + Tensor cell_pre_act_t = batch_cell_pre_act->Slice(bstart, bend); + + int cur_batch_size = bend - bstart; + + if (n > 0) { + int pre_h_start = static_cast(batch_starts[n - 1]); + int pre_h_end = pre_h_start + cur_batch_size; + auto pre_proj_t = batch_proj.Slice(pre_h_start, pre_h_end); + math::matmul(device_ctx, pre_proj_t, false, *weight, + false, static_cast(1.0), &gate_t, + static_cast(1.0)); + } else if (hidden_t0) { + // If n == 0 and there is no initialized hidden state, that is to say + // the H0 is zeros, the calculation W_h * H0 will be skiped. + // If n == 0 and there is initialized hidden state, calculate W_h * H0. + + // Since the batch computing for LSTMP reorders the input sequence + // according to their length. The initialized hidden state also needs + // to reorder. + Tensor ordered_h0, ordered_proj0; + ordered_proj0.Resize({1, proj_weight->dims()[1]}); + ordered_proj0.mutable_data(ctx.GetPlace()); + ReorderInitState(device_ctx, *hidden_t0, order, + &ordered_h0, true); + math::matmul(device_ctx, ordered_h0, false, + *proj_weight, false, static_cast(1.0), + &ordered_proj0, static_cast(0.0)); + math::matmul(device_ctx, ordered_proj0, false, + *weight, false, static_cast(1.0), + &gate_t, static_cast(1.0)); + } + + lstmp_value.gate_value = gate_t.data(); + lstmp_value.output_value = hidden_t.data(); + lstmp_value.state_value = cell_t.data(); + lstmp_value.state_active_value = cell_pre_act_t.data(); + math::LstmUnitFunctor::compute( + device_ctx, lstmp_value, frame_size, cur_batch_size, gate_act, + cell_act, cand_act); + lstmp_value.prev_state_value = lstmp_value.state_value; + math::matmul(device_ctx, hidden_t, false, *proj_weight, + false, static_cast(1.0), &proj_t, + static_cast(0.0)); + } + + math::Batch2LoDTensorFunctor to_seq; + batch_proj.set_lod(batch_gate->lod()); + // restore the output hidden in LoDTensor from the batch hidden + to_seq(device_ctx, batch_proj, *proj_out); + + batch_cell.set_lod(batch_gate->lod()); + // restore the output cell state in LoDTensor from the batch cell + to_seq(device_ctx, batch_cell, *cell_out); + } +}; + +template +class LSTMPGradKernel : public framework::OpKernel { + public: + void Compute(const framework::ExecutionContext& ctx) const override { + auto* input = ctx.Input("Input"); + auto* weight = ctx.Input("Weight"); + auto* bias = ctx.Input("Bias"); + + auto* proj_out = ctx.Input("Projection"); + auto* cell_out = ctx.Input("Cell"); + + auto* batch_gate = ctx.Input("BatchGate"); + auto* batch_cell_pre_act = ctx.Input("BatchCellPreAct"); + + auto* hidden_g = ctx.Input(framework::GradVarName("Projection")); + + auto* in_g = ctx.Output(framework::GradVarName("Input")); + auto* weight_g = ctx.Output(framework::GradVarName("Weight")); + auto* bias_g = ctx.Output(framework::GradVarName("Bias")); + + auto* h0 = ctx.Input("H0"); + auto* c0 = ctx.Input("C0"); + + auto* h0_g = ctx.Output(framework::GradVarName("H0")); + auto* c0_g = ctx.Output(framework::GradVarName("C0")); + + auto& device_ctx = ctx.template device_context(); + math::SetConstant zero; + if (weight_g) { + weight_g->mutable_data(ctx.GetPlace()); + zero(device_ctx, weight_g, static_cast(0.0)); + } + + // ordered_h0/c0 is the reordered hidden/cell initialization. + // ordered_h0_g/c0_g is the reordered gradient of hidden/cell + // initialization. + Tensor ordered_h0, ordered_c0, ordered_h0_g, ordered_c0_g; + const size_t* order = batch_gate->lod()[2].data(); + if (c0) { + ReorderInitState(device_ctx, *c0, order, &ordered_c0, + true); + } + if (c0 && c0_g) { + ordered_c0_g.mutable_data(c0_g->dims(), ctx.GetPlace()); + } + + auto in_dims = input->dims(); + auto out_dims = hidden_g->dims(); + int frame_size = static_cast(in_dims[1] / 4); + PADDLE_ENFORCE_EQ(frame_size, out_dims[1]); + + math::LstmMetaValue lstmp_value; + if (bias && ctx.Attr("use_peepholes")) { + T* bias_data = const_cast(bias->data()); + lstmp_value.check_ig = bias_data + 4 * frame_size; + lstmp_value.check_fg = lstmp_value.check_ig + frame_size; + lstmp_value.check_og = lstmp_value.check_fg + frame_size; + } else { + lstmp_value.check_ig = nullptr; + lstmp_value.check_fg = nullptr; + lstmp_value.check_og = nullptr; + } + + math::LstmMetaGrad lstmp_grad; + + if (bias && bias_g) { + bias_g->mutable_data(ctx.GetPlace()); + zero(device_ctx, bias_g, static_cast(0.0)); + } + if (bias && bias_g && ctx.Attr("use_peepholes")) { + T* bias_g_data = bias_g->data(); + lstmp_grad.check_ig_grad = bias_g_data + 4 * frame_size; + lstmp_grad.check_fg_grad = lstmp_grad.check_ig_grad + frame_size; + lstmp_grad.check_og_grad = lstmp_grad.check_fg_grad + frame_size; + } else { + lstmp_grad.check_ig_grad = nullptr; + lstmp_grad.check_fg_grad = nullptr; + lstmp_grad.check_og_grad = nullptr; + } + + math::LoDTensor2BatchFunctor to_batch; + + auto ToBatch = [&batch_gate, &to_batch]( + const DeviceContext& ctx, const framework::LoDTensor& src, + const framework::DDim& dims, framework::LoDTensor& dst) { + dst.mutable_data(dims, ctx.GetPlace()); + dst.set_lod(batch_gate->lod()); + to_batch(ctx, src, dst, false); + }; + + LoDTensor batch_proj, batch_proj_g, batch_cell; + ToBatch(device_ctx, *proj_out, out_dims, batch_proj); + ToBatch(device_ctx, *hidden_g, out_dims, batch_proj_g); + ToBatch(device_ctx, *cell_out, out_dims, batch_cell); + + LoDTensor batch_cell_g, batch_gate_g; + batch_cell_g.mutable_data(out_dims, ctx.GetPlace()); + // TODO(qingqing) support the case output cell has gradient. + // to_batch(device_ctx, *cell_g, batch_cell_g, false); + zero(device_ctx, &batch_cell_g, static_cast(0.0)); + batch_gate_g.mutable_data(batch_gate->dims(), ctx.GetPlace()); + batch_gate_g.set_lod(batch_gate->lod()); + + auto gate_act = math::detail::GetActivationType( + ctx.Attr("gate_activation")); + auto cell_act = math::detail::GetActivationType( + ctx.Attr("cell_activation")); + auto cand_act = math::detail::GetActivationType( + ctx.Attr("candidate_activation")); + + auto batch_starts = batch_gate->lod()[0]; + size_t num_batch = batch_starts.size() - 1; + for (int n = static_cast(num_batch) - 1; n >= 0; n--) { + int bstart = static_cast(batch_starts[n]); + int bend = static_cast(batch_starts[n + 1]); + + Tensor gate = batch_gate->Slice(bstart, bend); + Tensor cell = batch_cell.Slice(bstart, bend); + Tensor cell_pre_act = batch_cell_pre_act->Slice(bstart, bend); + lstmp_value.gate_value = gate.data(); + lstmp_value.state_value = cell.data(); + lstmp_value.state_active_value = cell_pre_act.data(); + + Tensor out_g = batch_proj_g.Slice(bstart, bend); + Tensor gate_g = batch_gate_g.Slice(bstart, bend); + Tensor cell_g = batch_cell_g.Slice(bstart, bend); + lstmp_grad.state_grad = cell_g.data(); + lstmp_grad.gate_grad = gate_g.data(); + lstmp_grad.output_grad = out_g.data(); + + if (n > 0) { + int bstart_pre = static_cast(batch_starts[n - 1]); + Tensor cell_pre = batch_cell.Slice(bstart_pre, bstart); + Tensor cell_pre_g = batch_cell_g.Slice(bstart_pre, bstart); + lstmp_value.prev_state_value = cell_pre.data(); + lstmp_grad.prev_state_grad = cell_pre_g.data(); + } else { + lstmp_value.prev_state_value = c0 ? ordered_c0.data() : nullptr; + lstmp_grad.prev_state_grad = c0_g ? ordered_c0_g.data() : nullptr; + } + + int cur_batch_size = bend - bstart; + math::LstmUnitGradFunctor::compute( + device_ctx, lstmp_value, lstmp_grad, frame_size, cur_batch_size, + gate_act, cell_act, cand_act); + + if (n > 0) { + int pre_h_start = static_cast(batch_starts[n - 1]); + int pre_h_end = pre_h_start + cur_batch_size; + auto pre_proj_g = batch_proj_g.Slice(pre_h_start, pre_h_end); + math::matmul(device_ctx, gate_g, false, *weight, true, + static_cast(1.0), &pre_proj_g, + static_cast(1.0)); + if (weight_g) { + /* backward weight */ + auto pre_proj = batch_proj.Slice(pre_h_start, pre_h_end); + math::matmul(device_ctx, pre_proj, true, gate_g, + false, static_cast(1.0), weight_g, + static_cast(1.0)); + } + } else { + if (h0 && weight_g) { + ReorderInitState(device_ctx, *h0, order, + &ordered_h0, true); + math::matmul(device_ctx, ordered_h0, true, gate_g, + false, static_cast(1.0), weight_g, + static_cast(1.0)); + } + if (h0 && h0_g) { + ordered_h0_g.mutable_data(h0_g->dims(), ctx.GetPlace()); + math::matmul(device_ctx, gate_g, false, *weight, + true, static_cast(1.0), + &ordered_h0_g, static_cast(0.0)); + } + } + } + + math::Batch2LoDTensorFunctor to_seq; + if (in_g) { + /* backward data */ + in_g->mutable_data(ctx.GetPlace()); + to_seq(device_ctx, batch_gate_g, *in_g); + } + if (bias && bias_g) { + /* backward bias */ + Tensor b_g = *bias_g; + b_g.Resize({bias_g->numel(), 1}); + Tensor gate_bias_g = b_g.Slice(0, 4 * frame_size); + math::ColwiseSum col_sum; + col_sum(device_ctx, batch_gate_g, &gate_bias_g); + } + + if (h0 && h0_g) { + ReorderInitState(device_ctx, ordered_h0_g, order, h0_g, + false); + } + if (c0 && c0_g) { + ReorderInitState(device_ctx, ordered_c0_g, order, c0_g, + false); + } + } +}; + +} // namespace operators +} // namespace paddle diff --git a/python/paddle/v2/fluid/tests/test_lstmp_op.py b/python/paddle/v2/fluid/tests/test_lstmp_op.py new file mode 100644 index 00000000000..e35712ec069 --- /dev/null +++ b/python/paddle/v2/fluid/tests/test_lstmp_op.py @@ -0,0 +1,314 @@ +# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserve. +# +#Licensed under the Apache License, Version 2.0 (the "License"); +#you may not use this file except in compliance with the License. +#You may obtain a copy of the License at +# +# http://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. +import unittest +import numpy as np +from op_test import OpTest + +SIGMOID_THRESHOLD_MIN = -40.0 +SIGMOID_THRESHOLD_MAX = 13.0 +EXP_MAX_INPUT = 40.0 + + +def identity(x): + return x + + +def sigmoid(x): + y = np.copy(x) + y[x < SIGMOID_THRESHOLD_MIN] = SIGMOID_THRESHOLD_MIN + y[x > SIGMOID_THRESHOLD_MAX] = SIGMOID_THRESHOLD_MAX + return 1. / (1. + np.exp(-y)) + + +def tanh(x): + y = -2. * x + y[y > EXP_MAX_INPUT] = EXP_MAX_INPUT + return (2. / (1. + np.exp(y))) - 1. + + +def relu(x): + return np.maximum(x, 0) + + +ACTVATION = { + 'identity': identity, + 'sigmoid': sigmoid, + 'tanh': tanh, + 'relu': relu +} + + +# LSTM with recurrent projection Layer +def lstmp( + input, # T x 4D + lod, # 1 x N + h0=None, # N x D + c0=None, # N x D + w_r=None, # P x 5D + w_rh=None, # D x P + w_b=None, # 1 x 4D + w_c=None, # 1 x 3D + is_reverse=False, + act_gate=None, + act_cell=None, + act_cand=None): + def _step(x, w_r, w_rh, w_c, r_pre, c_pre, act_gate, act_cell, act_cand): + g = np.dot(r_pre, w_r) # 1 x 4D + g = g + x + g = np.reshape(g, (1, g.size)) + c, g_i, g_f, g_o = np.split(g, 4, axis=1) + if w_c is None: + g_i = act_gate(g_i) # 1 x D + g_f = act_gate(g_f) # 1 x D + else: + w_ic, w_fc, _ = np.split(w_c, 3, axis=1) + g_i = act_gate(g_i + w_ic * c_pre) # 1 x D + g_f = act_gate(g_f + w_fc * c_pre) # 1 x D + c = g_f * c_pre + g_i * act_cand(c) # 1 x D + + if w_c is None: + g_o = act_gate(g_o) # 1 x D + else: + _, _, w_oc = np.split(w_c, 3, axis=1) + g_o = act_gate(g_o + w_oc * c) # 1 x D + h = g_o * act_cell(c) + # projection + r = np.dot(h, w_rh) + return r, c + + def _reverse(x, lod): + y = np.zeros_like(x) + for i in range(len(lod) - 1): + b, e = lod[i], lod[i + 1] + y[b:e, :] = np.flip(x[b:e, :], 0) + return y + + offset = lod[0] + batch_size = len(offset) - 1 + # recurrent projection state + projection = [] + cell = [] + input = _reverse(input, offset) if is_reverse else input + if w_b is not None: + input = input + np.tile(w_b, (offset[-1], 1)) + for i in range(batch_size): + # compute one sequence + seq_len = offset[i + 1] - offset[i] + x = input[offset[i]:offset[i + 1], :] + r_pre = np.dot(h0[i], w_rh) # 1 x P + c_pre = c0[i] # 1 x D + for j in range(seq_len): + # compute one step + r_pre, c_pre = _step(x[j], w_r, w_rh, w_c, r_pre, c_pre, act_gate, + act_cell, act_cand) + projection.append(r_pre.flatten()) + cell.append(c_pre.flatten()) + + projection = np.array(projection).astype('float64') + cell = np.array(cell).astype('float64') + + projection = _reverse(projection, offset) if is_reverse else projection + cell = _reverse(cell, offset) if is_reverse else cell + + assert projection.shape == (input.shape[0], w_r.shape[0]) # T x P + assert cell.shape == (input.shape[0], input.shape[1] / 4) # T x D + return projection, cell + + +class TestLstmOp(OpTest): + def set_argument(self): + self.lod = [[0, 2, 5, 7]] + # hidden size + self.D = 16 + # projection size + self.P = 10 + + self.act_gate = 'sigmoid' + self.act_cell = 'tanh' + self.act_cand = 'tanh' + + self.has_initial_state = False + self.is_reverse = False + self.use_peepholes = True + + def setUp(self): + self.set_argument() + self.op_type = 'lstmp' + + T = self.lod[0][-1] + N = len(self.lod[0]) - 1 + + x = np.random.normal(size=(T, 4 * self.D)).astype('float64') + if self.has_initial_state: + h0 = np.random.normal(size=(N, self.D)).astype('float64') + c0 = np.random.normal(size=(N, self.D)).astype('float64') + else: + h0 = np.zeros((N, self.D)).astype('float64') + c0 = np.zeros((N, self.D)).astype('float64') + w = np.random.normal(size=(self.P, 4 * self.D)).astype('float64') + if self.use_peepholes: + b = np.random.normal(size=(1, 7 * self.D)).astype('float64') + else: + b = np.random.normal(size=(1, 4 * self.D)).astype('float64') + + w_b = b[:, 0:4 * self.D] + w_c = b[:, 4 * self.D:] if self.use_peepholes else None + w_rh = np.random.normal(size=(self.D, self.P)).astype('float64') + r, c = lstmp(x, self.lod, h0, c0, w, w_rh, w_b, w_c, self.is_reverse, + ACTVATION[self.act_gate], ACTVATION[self.act_cell], + ACTVATION[self.act_cand]) + + self.inputs = {'Input': (x, self.lod), 'Weight': w, 'ProjWeight': w_rh} + + self.inputs['Bias'] = b + + if self.has_initial_state: + self.inputs['H0'] = h0 + self.inputs['C0'] = c0 + + self.outputs = { + 'Projection': (r, self.lod), + 'Cell': (c, self.lod), + } + self.attrs = { + 'use_peepholes': self.use_peepholes, + 'is_reverse': self.is_reverse, + 'gate_activation': self.act_gate, + 'cell_activation': self.act_cell, + 'candidate_activation': self.act_cand + } + + def test_check_output(self): + self.check_output(atol=1e-8) + + """ + def test_check_grad(self): + # TODO(qingqing) remove folowing lines after the check_grad is refined. + N = len(self.lod[0]) - 1 + self.outputs['BatchGate'] = np.zeros((N, 4 * self.D)).astype('float64') + self.outputs['BatchCellPreAct'] = np.zeros( + (N, self.D)).astype('float64') + self.check_grad( + ['Input', 'Weight', 'Bias'], ['Hidden'], max_relative_error=5e-4) + """ + + +""" +class TestLstmOpHasInitial(TestLstmOp): + def set_argument(self): + self.lod = [[0, 2, 5, 7]] + self.D = 16 + + self.act_gate = 'sigmoid' + self.act_cell = 'tanh' + self.act_cand = 'tanh' + + self.has_initial_state = True + self.is_reverse = True + self.use_peepholes = True + + def test_check_grad(self): + # TODO(qingqing) remove folowing lines after the check_grad is refined. + N = len(self.lod[0]) - 1 + self.outputs['BatchGate'] = np.zeros((N, 4 * self.D)).astype('float64') + self.outputs['BatchCellPreAct'] = np.zeros( + (N, self.D)).astype('float64') + self.check_grad( + ['Input', 'Weight', 'Bias', 'H0', 'C0'], ['Hidden'], + max_relative_error=5e-4) + + def test_check_grad_ingore_bias(self): + N = len(self.lod[0]) - 1 + self.outputs['BatchGate'] = np.zeros((N, 4 * self.D)).astype('float64') + self.outputs['BatchCellPreAct'] = np.zeros( + (N, self.D)).astype('float64') + self.check_grad( + ['Input', 'Weight'], ['Hidden'], + max_relative_error=5e-4, + no_grad_set=set('Bias')) + + def test_check_grad_ingore_weight(self): + N = len(self.lod[0]) - 1 + self.outputs['BatchGate'] = np.zeros((N, 4 * self.D)).astype('float64') + self.outputs['BatchCellPreAct'] = np.zeros( + (N, self.D)).astype('float64') + self.check_grad( + ['Input', 'Bias'], ['Hidden'], + max_relative_error=5e-4, + no_grad_set=set('Weight')) + + def test_check_grad_ingore_input(self): + N = len(self.lod[0]) - 1 + self.outputs['BatchGate'] = np.zeros((N, 4 * self.D)).astype('float64') + self.outputs['BatchCellPreAct'] = np.zeros( + (N, self.D)).astype('float64') + self.check_grad( + ['Weight', 'Bias'], ['Hidden'], + max_relative_error=5e-4, + no_grad_set=set('Input')) + + def test_check_grad_ingore_h0(self): + N = len(self.lod[0]) - 1 + self.outputs['BatchGate'] = np.zeros((N, 4 * self.D)).astype('float64') + self.outputs['BatchCellPreAct'] = np.zeros( + (N, self.D)).astype('float64') + self.check_grad( + ['Input', 'Weight', 'Bias', 'C0'], ['Hidden'], + max_relative_error=5e-4, + no_grad_set=set('H0')) + + def test_check_grad_ingore_c0(self): + N = len(self.lod[0]) - 1 + self.outputs['BatchGate'] = np.zeros((N, 4 * self.D)).astype('float64') + self.outputs['BatchCellPreAct'] = np.zeros( + (N, self.D)).astype('float64') + self.check_grad( + ['Input', 'Weight', 'Bias', 'H0'], ['Hidden'], + max_relative_error=5e-4, + no_grad_set=set('C0')) +""" + + +class TestLstmOpRerverse(TestLstmOp): + def set_argument(self): + self.lod = [[0, 2, 5, 7]] + self.D = 16 + self.P = 10 + + self.act_gate = 'sigmoid' + self.act_cell = 'tanh' + self.act_cand = 'tanh' + + self.has_initial_state = False + self.is_reverse = True + self.use_peepholes = True + + +class TestLstmOpNotUsePeepholes(TestLstmOp): + def set_argument(self): + self.lod = [[0, 2, 5, 7]] + self.D = 16 + self.P = 10 + + self.act_gate = 'sigmoid' + self.act_cell = 'tanh' + self.act_cand = 'tanh' + + self.has_initial_state = False + self.is_reverse = True + self.use_peepholes = False + + +if __name__ == '__main__': + unittest.main() -- GitLab