wehub-resource-sync
964 lines
43 KiB
C++
964 lines
43 KiB
C++
/* Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
|
|
|
|
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/fluid/inference/tensorrt/convert/op_converter.h"
|
|
#include "paddle/fluid/inference/tensorrt/plugin/qkv_to_context_plugin.h"
|
|
#include "paddle/fluid/inference/tensorrt/plugin/transformer_input_output_convert_plugin.h"
|
|
|
|
namespace paddle::inference::tensorrt {
|
|
|
|
class MultiheadMatMulOpConverter : public OpConverter {
|
|
public:
|
|
void operator()(const framework::proto::OpDesc& op,
|
|
const framework::Scope& scope,
|
|
bool test_mode) override {
|
|
VLOG(3) << "convert a multihead_matmul op to a corresponding tensorrt "
|
|
"network structure";
|
|
framework::OpDesc op_desc(op, nullptr);
|
|
// Declare inputs
|
|
auto* input = engine_->GetITensor(op_desc.Input("Input").front());
|
|
auto input_dims = input->getDimensions();
|
|
bool bias_qk_attr =
|
|
(op_desc.Inputs().find("BiasQK") == op_desc.Inputs().end()) ? false
|
|
: true;
|
|
|
|
// fc weights and fc bias
|
|
auto weight_name = op_desc.Input("W").front();
|
|
auto bias_name = op_desc.Input("Bias").front();
|
|
|
|
auto* weight_v = scope.FindVar(weight_name);
|
|
auto* weight_t = weight_v->GetMutable<phi::DenseTensor>();
|
|
|
|
auto* bias_v = scope.FindVar(bias_name);
|
|
auto* bias_t = bias_v->GetMutable<phi::DenseTensor>();
|
|
|
|
float* weight_data = nullptr;
|
|
bool qkv2context_plugin_int8 = op_desc.HasAttr("qkv2context_plugin_int8");
|
|
float in_scale = 0.;
|
|
|
|
if (op_desc.HasAttr("Input_scale")) {
|
|
in_scale = PADDLE_GET_CONST(float, op_desc.GetAttr("Input_scale"));
|
|
engine_->SetTensorDynamicRange(input, in_scale);
|
|
}
|
|
weight_data = const_cast<float*>(static_cast<const float*>(
|
|
engine_->GetFp32TrtWeight(weight_name, *weight_t).get().values));
|
|
|
|
float* bias_data = const_cast<float*>(static_cast<const float*>(
|
|
engine_->GetFp32TrtWeight(bias_name, *bias_t).get().values));
|
|
std::vector<float> weight_data_tmp;
|
|
weight_data_tmp.reserve(weight_t->numel());
|
|
memcpy(
|
|
weight_data_tmp.data(), weight_data, weight_t->numel() * sizeof(float));
|
|
|
|
// (hidden_in, 3, hidden_out)
|
|
const auto& weight_dims = weight_t->dims();
|
|
|
|
int hidden_in = weight_dims[0]; // channels_in
|
|
int three = weight_dims[1]; // channels_out
|
|
int hidden_out = weight_dims[2]; // channels_out
|
|
int m = hidden_in;
|
|
int n = three * hidden_out;
|
|
|
|
int head_number = PADDLE_GET_CONST(int, op_desc.GetAttr("head_number"));
|
|
|
|
nvinfer1::ILayer* layer = nullptr;
|
|
auto output_name = op_desc.Output("Out")[0];
|
|
bool flag_varseqlen = engine_->use_varseqlen() &&
|
|
!engine_->tensorrt_transformer_posid().empty() &&
|
|
!engine_->tensorrt_transformer_maskid().empty();
|
|
if (engine_->with_dynamic_shape()) {
|
|
if (!engine_->tensorrt_transformer_maskid().empty() &&
|
|
engine_->precision() != phi::DataType::FLOAT32 &&
|
|
platform::GetGPUComputeCapability(platform::GetCurrentDeviceId()) >=
|
|
75) {
|
|
nvinfer1::Weights weight{nvinfer1::DataType::kFLOAT,
|
|
static_cast<void*>(weight_data),
|
|
static_cast<int32_t>(weight_t->numel())};
|
|
nvinfer1::Weights bias{nvinfer1::DataType::kFLOAT,
|
|
static_cast<void*>(bias_data),
|
|
static_cast<int32_t>(bias_t->numel())};
|
|
|
|
nvinfer1::ITensor* mask_tensor;
|
|
nvinfer1::ITensor* pos_id_tensor;
|
|
nvinfer1::ITensor* max_seqlen_tensor;
|
|
if (flag_varseqlen) {
|
|
mask_tensor = engine_->GetITensor("qkv_plugin_mask");
|
|
pos_id_tensor = engine_->GetITensor("pos_id");
|
|
max_seqlen_tensor = engine_->GetITensor("max_seqlen_tensor");
|
|
} else {
|
|
auto* bias_qk_tensor =
|
|
engine_->GetITensor(op_desc.Input("BiasQK").front());
|
|
auto bias_qk_dims = bias_qk_tensor->getDimensions();
|
|
PADDLE_ENFORCE_EQ(bias_qk_dims.nbDims,
|
|
4,
|
|
common::errors::InvalidArgument(
|
|
"The rank of Multihead Matmul'BiasQK must be "
|
|
"4, but got rank is %d.",
|
|
bias_qk_dims.nbDims));
|
|
|
|
nvinfer1::Dims start_dims = bias_qk_dims;
|
|
start_dims.d[0] = 0;
|
|
start_dims.d[1] = 0;
|
|
start_dims.d[2] = 0;
|
|
start_dims.d[3] = 0;
|
|
nvinfer1::Dims size_dims = bias_qk_dims;
|
|
nvinfer1::Dims step_dims = bias_qk_dims;
|
|
step_dims.d[0] = 1;
|
|
step_dims.d[1] = 1;
|
|
step_dims.d[2] = 1;
|
|
step_dims.d[3] = 1;
|
|
auto* shape_tensor = Shape(bias_qk_tensor);
|
|
|
|
// (b,*,*,m) -> (b,1,1,m)
|
|
std::vector<nvinfer1::ITensor*> size_vec_tensor;
|
|
size_vec_tensor.push_back(GetEleTensorOfShape(shape_tensor, 0));
|
|
size_vec_tensor.push_back(Add1DConstantLayer(1));
|
|
size_vec_tensor.push_back(Add1DConstantLayer(1));
|
|
size_vec_tensor.push_back(GetEleTensorOfShape(shape_tensor, 3));
|
|
|
|
auto* size_tensor = Concat(size_vec_tensor);
|
|
auto* slice_layer = TRT_ENGINE_ADD_LAYER(engine_,
|
|
Slice,
|
|
*bias_qk_tensor,
|
|
start_dims,
|
|
size_dims,
|
|
step_dims);
|
|
slice_layer->setInput(2, *size_tensor);
|
|
|
|
// half -> bool
|
|
auto* cast_layer_0 = TRT_ENGINE_ADD_LAYER(
|
|
engine_, Identity, *slice_layer->getOutput(0));
|
|
cast_layer_0->setOutputType(0, nvinfer1::DataType::kBOOL);
|
|
|
|
// bool kNOT
|
|
auto* not_layer =
|
|
TRT_ENGINE_ADD_LAYER(engine_,
|
|
Unary,
|
|
*cast_layer_0->getOutput(0),
|
|
nvinfer1::UnaryOperation::kNOT);
|
|
|
|
// bool -> int32
|
|
auto* cast_layer_1 =
|
|
TRT_ENGINE_ADD_LAYER(engine_, Identity, *not_layer->getOutput(0));
|
|
cast_layer_1->setOutputType(0, nvinfer1::DataType::kINT32);
|
|
|
|
// Calculate the number of 1 : (b,1,1,m) -> (b)
|
|
uint32_t reduce_dim_0 = 0;
|
|
reduce_dim_0 |= 1 << 1; // 00000000000000000000000000000010
|
|
reduce_dim_0 |= 1 << 2; // 00000000000000000000000000000110
|
|
reduce_dim_0 |= 1 << 3; // 00000000000000000000000000001110
|
|
bool keep_dim = false;
|
|
nvinfer1::ReduceOperation reduce_type =
|
|
nvinfer1::ReduceOperation::kSUM;
|
|
auto* reduce_sum_layer =
|
|
TRT_ENGINE_ADD_LAYER(engine_,
|
|
Reduce,
|
|
*cast_layer_1->getOutput(0),
|
|
reduce_type,
|
|
reduce_dim_0,
|
|
keep_dim);
|
|
std::vector<nvinfer1::ITensor*> inputs_transformer;
|
|
inputs_transformer.emplace_back(input);
|
|
inputs_transformer.emplace_back(
|
|
reduce_sum_layer->getOutput(0)); // (b,m)
|
|
plugin::TransformerInputConvertPlugin* plugin =
|
|
new plugin::TransformerInputConvertPlugin();
|
|
nvinfer1::ILayer* transformer_input_layer = engine_->AddDynamicPlugin(
|
|
inputs_transformer.data(), inputs_transformer.size(), plugin);
|
|
|
|
input = transformer_input_layer->getOutput(0);
|
|
if (op_desc.HasAttr("Input_scale")) {
|
|
in_scale = PADDLE_GET_CONST(float, op_desc.GetAttr("Input_scale"));
|
|
engine_->SetTensorDynamicRange(input, in_scale);
|
|
}
|
|
mask_tensor = transformer_input_layer->getOutput(1);
|
|
pos_id_tensor = transformer_input_layer->getOutput(2);
|
|
max_seqlen_tensor = transformer_input_layer->getOutput(3);
|
|
}
|
|
if (engine_->with_interleaved()) {
|
|
VLOG(4) << "fused multihead_matmul op: use_varseqlen and "
|
|
"with_interleaved";
|
|
if (!op_desc.HasAttr("Input_scale")) {
|
|
PADDLE_THROW(
|
|
common::errors::Fatal("use with_interleaved must be int8."));
|
|
}
|
|
nvinfer1::ILayer* fc_layer = nullptr;
|
|
float dp_probs = 1.0 / 127.0;
|
|
nvinfer1::DimsHW nv_ksize(1, 1);
|
|
fc_layer = TRT_ENGINE_ADD_LAYER(
|
|
engine_, ConvolutionNd, *input, n, nv_ksize, weight, bias);
|
|
fc_layer->setName(
|
|
("Multihead: Convolution/FullyConnected: (Output: " +
|
|
output_name + ")")
|
|
.c_str());
|
|
PADDLE_ENFORCE_EQ(
|
|
op_desc.HasAttr("fc_out_threshold"),
|
|
true,
|
|
common::errors::InvalidArgument(
|
|
"must have out_threshold in multihead layers in int8 mode"));
|
|
float out_scale =
|
|
PADDLE_GET_CONST(float, op_desc.GetAttr("fc_out_threshold"));
|
|
engine_->SetTensorDynamicRange(fc_layer->getOutput(0), out_scale);
|
|
if (qkv2context_plugin_int8) {
|
|
dp_probs =
|
|
PADDLE_GET_CONST(float, op_desc.GetAttr("dp_probs")) / 127.0;
|
|
}
|
|
auto creator = GetPluginRegistry()->getPluginCreator(
|
|
"CustomQKVToContextPluginDynamic", "3");
|
|
assert(creator != nullptr);
|
|
std::vector<nvinfer1::PluginField> fields{
|
|
{"hidden_size",
|
|
&hidden_out,
|
|
nvinfer1::PluginFieldType::kINT32,
|
|
1},
|
|
{"num_heads",
|
|
&head_number,
|
|
nvinfer1::PluginFieldType::kINT32,
|
|
1}};
|
|
if (qkv2context_plugin_int8) {
|
|
fields.push_back({"dq_probs",
|
|
&dp_probs,
|
|
nvinfer1::PluginFieldType::kFLOAT32,
|
|
1});
|
|
}
|
|
std::unique_ptr<nvinfer1::PluginFieldCollection> plugin_collection(
|
|
new nvinfer1::PluginFieldCollection);
|
|
plugin_collection->nbFields = static_cast<int>(fields.size());
|
|
plugin_collection->fields = fields.data();
|
|
|
|
auto plugin = creator->createPlugin("CustomQKVToContextPluginDynamic",
|
|
plugin_collection.get());
|
|
plugin_collection.reset();
|
|
|
|
std::vector<nvinfer1::ITensor*> plugin_inputs;
|
|
plugin_inputs.emplace_back(fc_layer->getOutput(0));
|
|
plugin_inputs.emplace_back(pos_id_tensor);
|
|
plugin_inputs.emplace_back(
|
|
max_seqlen_tensor); // max_seqlen, eval_placeholder_3
|
|
auto plugin_layer = engine_->network()->addPluginV2(
|
|
plugin_inputs.data(), plugin_inputs.size(), *plugin);
|
|
ReplenishLayerAndOutput(
|
|
plugin_layer, "multihead_matmul", {output_name}, test_mode);
|
|
} else {
|
|
auto* reshape_before_matrix =
|
|
TRT_ENGINE_ADD_LAYER(engine_, Shuffle, *input);
|
|
|
|
std::vector<nvinfer1::ITensor*> reshape_before_tensor_matrix;
|
|
reshape_before_tensor_matrix.push_back(
|
|
GetEleTensorOfShape(Shape(input), 0));
|
|
reshape_before_tensor_matrix.push_back(
|
|
GetEleTensorOfShape(Shape(input), 1));
|
|
|
|
reshape_before_matrix->setInput(
|
|
1, *Concat(reshape_before_tensor_matrix));
|
|
reshape_before_matrix->setName(
|
|
("reshape_before_matrix(Output: " + output_name + ")").c_str());
|
|
auto* input = reshape_before_matrix->getOutput(0);
|
|
if (op_desc.HasAttr("Input_scale")) {
|
|
in_scale = PADDLE_GET_CONST(float, op_desc.GetAttr("Input_scale"));
|
|
engine_->SetTensorDynamicRange(input, in_scale);
|
|
}
|
|
int head_size = hidden_out / head_number;
|
|
// [hidden_in, 3, head_number, head_size] -> [hidden_in, head_number,
|
|
// 3, head_size]
|
|
auto transpose_weight_v2 = [](const float* src,
|
|
float* dst,
|
|
int three,
|
|
int head_number,
|
|
int head_size,
|
|
int hidden_in) {
|
|
for (int i = 0; i < hidden_in; ++i) {
|
|
for (int j = 0; j < three; ++j) {
|
|
for (int n = 0; n < head_number; ++n) {
|
|
for (int m = 0; m < head_size; ++m) {
|
|
dst[i * head_number * three * head_size +
|
|
n * three * head_size + j * head_size + m] =
|
|
src[i * three * head_number * head_size +
|
|
j * head_number * head_size + n * head_size + m];
|
|
}
|
|
}
|
|
}
|
|
}
|
|
};
|
|
// [3, head_number, head_size] -> [head_number, 3, head_size]
|
|
auto transpose_bias_v2 =
|
|
[](const float* src, float* dst, int N, int H) {
|
|
for (int i = 0; i < 3; ++i) {
|
|
for (int n = 0; n < N; ++n) {
|
|
for (int h = 0; h < H; ++h) {
|
|
dst[n * 3 * H + i * H + h] = src[i * N * H + n * H + h];
|
|
}
|
|
}
|
|
}
|
|
};
|
|
memcpy(weight_data_tmp.data(),
|
|
weight_data,
|
|
weight_t->numel() * sizeof(float));
|
|
transpose_weight_v2(weight_data_tmp.data(),
|
|
weight_data,
|
|
three,
|
|
head_number,
|
|
head_size,
|
|
hidden_in);
|
|
|
|
std::vector<float> bias_data_tmp;
|
|
bias_data_tmp.reserve(bias_t->numel());
|
|
memcpy(
|
|
bias_data_tmp.data(), bias_data, bias_t->numel() * sizeof(float));
|
|
transpose_bias_v2(
|
|
bias_data_tmp.data(), bias_data, head_number, head_size);
|
|
|
|
float dp_probs = 1.0 / 127.0;
|
|
|
|
nvinfer1::Dims trt_dims_weight;
|
|
trt_dims_weight.nbDims = 2;
|
|
trt_dims_weight.d[0] = m;
|
|
trt_dims_weight.d[1] = n;
|
|
auto* weight_tensor =
|
|
TRT_ENGINE_ADD_LAYER(engine_, Constant, trt_dims_weight, weight)
|
|
->getOutput(0);
|
|
|
|
bool transpose_x = false;
|
|
bool transpose_y = false;
|
|
|
|
nvinfer1::MatrixOperation matrix_operation_x =
|
|
transpose_x ? nvinfer1::MatrixOperation::kTRANSPOSE
|
|
: nvinfer1::MatrixOperation::kNONE;
|
|
nvinfer1::MatrixOperation matrix_operation_y =
|
|
transpose_y ? nvinfer1::MatrixOperation::kTRANSPOSE
|
|
: nvinfer1::MatrixOperation::kNONE;
|
|
|
|
auto* matrix_layer = TRT_ENGINE_ADD_LAYER(engine_,
|
|
MatrixMultiply,
|
|
*input,
|
|
matrix_operation_x,
|
|
*weight_tensor,
|
|
matrix_operation_y);
|
|
|
|
nvinfer1::Dims trt_dims_bias;
|
|
trt_dims_bias.nbDims = 2;
|
|
trt_dims_bias.d[0] = 1;
|
|
trt_dims_bias.d[1] = n;
|
|
auto* bias_tensor =
|
|
TRT_ENGINE_ADD_LAYER(engine_, Constant, trt_dims_bias, bias)
|
|
->getOutput(0);
|
|
auto* add_layer =
|
|
TRT_ENGINE_ADD_LAYER(engine_,
|
|
ElementWise,
|
|
*matrix_layer->getOutput(0),
|
|
*bias_tensor,
|
|
nvinfer1::ElementWiseOperation::kSUM);
|
|
auto* reshape_before_multihead_layer =
|
|
TRT_ENGINE_ADD_LAYER(engine_, Shuffle, *add_layer->getOutput(0));
|
|
|
|
std::vector<nvinfer1::ITensor*> reshape_tensor;
|
|
reshape_tensor.push_back(
|
|
GetEleTensorOfShape(Shape(matrix_layer->getOutput(0)), 0));
|
|
reshape_tensor.push_back(
|
|
GetEleTensorOfShape(Shape(matrix_layer->getOutput(0)), 1));
|
|
reshape_tensor.push_back(Add1DConstantLayer(1));
|
|
reshape_tensor.push_back(Add1DConstantLayer(1));
|
|
|
|
reshape_before_multihead_layer->setInput(1, *Concat(reshape_tensor));
|
|
reshape_before_multihead_layer->setName(
|
|
("reshape_before_multihead_matmul(Output: " + output_name + ")")
|
|
.c_str());
|
|
|
|
if (op_desc.HasAttr("fc_out_threshold")) {
|
|
PADDLE_ENFORCE_EQ(op_desc.HasAttr("fc_out_threshold"),
|
|
true,
|
|
common::errors::InvalidArgument(
|
|
"must have out threshold in multihead layers "
|
|
"in int8 mode"));
|
|
float out_scale =
|
|
PADDLE_GET_CONST(float, op_desc.GetAttr("fc_out_threshold"));
|
|
|
|
engine_->SetTensorDynamicRange(matrix_layer->getOutput(0),
|
|
out_scale);
|
|
engine_->SetTensorDynamicRange(add_layer->getOutput(0), out_scale);
|
|
engine_->SetTensorDynamicRange(
|
|
reshape_before_multihead_layer->getOutput(0), out_scale);
|
|
|
|
if (qkv2context_plugin_int8) {
|
|
dp_probs =
|
|
PADDLE_GET_CONST(float, op_desc.GetAttr("dp_probs")) / 127.0;
|
|
}
|
|
}
|
|
|
|
auto creator = GetPluginRegistry()->getPluginCreator(
|
|
"CustomQKVToContextPluginDynamic", "2");
|
|
assert(creator != nullptr);
|
|
int type = static_cast<int>(nvinfer1::DataType::kHALF);
|
|
if (qkv2context_plugin_int8 &&
|
|
(engine_->precision() == phi::DataType::INT8)) {
|
|
type = static_cast<int>(nvinfer1::DataType::kINT8);
|
|
}
|
|
bool has_mask = true;
|
|
int var_seqlen = 1;
|
|
std::vector<nvinfer1::PluginField> fields{
|
|
{"type_id", &type, nvinfer1::PluginFieldType::kINT32, 1},
|
|
{"hidden_size",
|
|
&hidden_out,
|
|
nvinfer1::PluginFieldType::kINT32,
|
|
1},
|
|
{"num_heads", &head_number, nvinfer1::PluginFieldType::kINT32, 1},
|
|
{"has_mask", &has_mask, nvinfer1::PluginFieldType::kINT32, 1},
|
|
{"var_seqlen",
|
|
&var_seqlen,
|
|
nvinfer1::PluginFieldType::kINT32,
|
|
1}};
|
|
if (qkv2context_plugin_int8) {
|
|
fields.push_back({"dq_probs",
|
|
&dp_probs,
|
|
nvinfer1::PluginFieldType::kFLOAT32,
|
|
1});
|
|
}
|
|
std::unique_ptr<nvinfer1::PluginFieldCollection> plugin_collection(
|
|
new nvinfer1::PluginFieldCollection);
|
|
plugin_collection->nbFields = static_cast<int>(fields.size());
|
|
plugin_collection->fields = fields.data();
|
|
|
|
auto plugin = creator->createPlugin("CustomQKVToContextPluginDynamic",
|
|
plugin_collection.get());
|
|
plugin_collection.reset();
|
|
|
|
std::vector<nvinfer1::ITensor*> plugin_inputs;
|
|
plugin_inputs.emplace_back(
|
|
reshape_before_multihead_layer->getOutput(0));
|
|
plugin_inputs.emplace_back(mask_tensor);
|
|
plugin_inputs.emplace_back(pos_id_tensor);
|
|
plugin_inputs.emplace_back(
|
|
max_seqlen_tensor); // max_seqlen, eval_placeholder_3
|
|
|
|
auto plugin_layer = engine_->network()->addPluginV2(
|
|
plugin_inputs.data(), plugin_inputs.size(), *plugin);
|
|
plugin_layer->setName(
|
|
("CustomQKVToContextPluginDynamic: " + output_name).c_str());
|
|
// recover no_varlen output
|
|
if (!flag_varseqlen) {
|
|
std::vector<nvinfer1::ITensor*> output_transformer;
|
|
output_transformer.emplace_back(plugin_layer->getOutput(0));
|
|
output_transformer.emplace_back(
|
|
engine_->GetITensor(op_desc.Input("Input").front()));
|
|
output_transformer.emplace_back(pos_id_tensor);
|
|
plugin::TransformerOutputConvertPlugin* plugin =
|
|
new plugin::TransformerOutputConvertPlugin();
|
|
nvinfer1::ILayer* transformer_output_layer =
|
|
engine_->AddDynamicPlugin(output_transformer.data(),
|
|
output_transformer.size(),
|
|
plugin);
|
|
engine_->SetITensor(output_name,
|
|
transformer_output_layer->getOutput(0));
|
|
} else {
|
|
engine_->SetITensor(output_name, plugin_layer->getOutput(0));
|
|
if (op_desc.HasAttr("out_threshold")) {
|
|
float out_scale =
|
|
PADDLE_GET_CONST(float, op_desc.GetAttr("out_threshold"));
|
|
engine_->SetTensorDynamicRange(plugin_layer->getOutput(0),
|
|
out_scale);
|
|
}
|
|
}
|
|
}
|
|
} else {
|
|
auto transpose_weight = [](const float* src, float* dst, int m, int n) {
|
|
for (int i = 0; i < m; i++) {
|
|
for (int j = 0; j < n; j++) {
|
|
dst[j * m + i] = src[i * n + j];
|
|
}
|
|
}
|
|
};
|
|
transpose_weight(weight_data_tmp.data(), weight_data, m, n);
|
|
if (input_dims.d[1] <= 384 && !bias_qk_attr &&
|
|
engine_->precision() != phi::DataType::FLOAT32 &&
|
|
platform::GetGPUComputeCapability(platform::GetCurrentDeviceId()) >=
|
|
75) {
|
|
/*
|
|
* input_dims.d[0]: batch(-1)
|
|
* input_dims.d[1]: length:256
|
|
* input_dims.d[2]: hidden_size:768
|
|
input
|
|
|[b,256,768]
|
|
|
|
|
shuffle weight bias
|
|
|[b,256,768,1,1] | |
|
|
|_____________________|_________|
|
|
|
|
|
fc
|
|
|[b,256,2304,1,1]
|
|
|
|
|
shuffle mask(fake) pos max_length
|
|
|[b*256,2304,1,1] | | |
|
|
| | | |
|
|
|_______________________|_________|________|
|
|
|
|
|
MHA
|
|
|[b*256,768]
|
|
|
|
|
shuffle
|
|
|[b, 256, 768]
|
|
|
|
|
out
|
|
*/
|
|
|
|
nvinfer1::Weights weight{nvinfer1::DataType::kFLOAT,
|
|
static_cast<void*>(weight_data),
|
|
static_cast<int32_t>(weight_t->numel())};
|
|
nvinfer1::Weights bias{nvinfer1::DataType::kFLOAT,
|
|
static_cast<void*>(bias_data),
|
|
static_cast<int32_t>(bias_t->numel())};
|
|
|
|
/*** transpose the weight and bias ***/
|
|
int head_size = hidden_out / head_number;
|
|
// [3, head_number, head_size, hidden_in] -> [head_number, 3,
|
|
// head_size, hidden_in]
|
|
auto transpose_weight_v2 = [](const float* src,
|
|
float* dst,
|
|
int three,
|
|
int head_number,
|
|
int head_size,
|
|
int hidden_in) {
|
|
const int HH = head_size * hidden_in;
|
|
for (int i = 0; i < three; ++i) {
|
|
for (int n = 0; n < head_number; ++n) {
|
|
for (int hh = 0; hh < HH; ++hh) {
|
|
dst[n * three * HH + i * HH + hh] =
|
|
src[i * head_number * HH + n * HH + hh];
|
|
}
|
|
}
|
|
}
|
|
};
|
|
// [3, head_number, head_size] -> [head_number, 3, head_size]
|
|
auto transpose_bias_v2 =
|
|
[](const float* src, float* dst, int N, int H) {
|
|
for (int i = 0; i < 3; ++i) {
|
|
for (int n = 0; n < N; ++n) {
|
|
for (int h = 0; h < H; ++h) {
|
|
dst[n * 3 * H + i * H + h] = src[i * N * H + n * H + h];
|
|
}
|
|
}
|
|
}
|
|
};
|
|
memcpy(weight_data_tmp.data(),
|
|
weight_data,
|
|
weight_t->numel() * sizeof(float));
|
|
transpose_weight_v2(weight_data_tmp.data(),
|
|
weight_data,
|
|
three,
|
|
head_number,
|
|
head_size,
|
|
hidden_in);
|
|
|
|
std::vector<float> bias_data_tmp;
|
|
bias_data_tmp.reserve(bias_t->numel());
|
|
memcpy(
|
|
bias_data_tmp.data(), bias_data, bias_t->numel() * sizeof(float));
|
|
transpose_bias_v2(
|
|
bias_data_tmp.data(), bias_data, head_number, head_size);
|
|
nvinfer1::ITensor* input_shape_tensor = Shape(input);
|
|
#if IS_TRT_VERSION_GE(8600)
|
|
// add matmul and elementwise layer
|
|
auto* fc_weight_layer = TRT_ENGINE_ADD_LAYER(
|
|
engine_, Constant, nvinfer1::Dims3(1, n, hidden_in), weight);
|
|
auto* fc_matmul_layer =
|
|
TRT_ENGINE_ADD_LAYER(engine_,
|
|
MatrixMultiply,
|
|
*input,
|
|
nvinfer1::MatrixOperation::kNONE,
|
|
*fc_weight_layer->getOutput(0),
|
|
nvinfer1::MatrixOperation::kTRANSPOSE);
|
|
fc_matmul_layer->setName(
|
|
("multihead_matmul_fc_matmul(Output: " + output_name + ")")
|
|
.c_str());
|
|
auto* fc_bias_layer = TRT_ENGINE_ADD_LAYER(
|
|
engine_, Constant, nvinfer1::Dims3(1, 1, n), bias);
|
|
auto* fc_add_layer =
|
|
TRT_ENGINE_ADD_LAYER(engine_,
|
|
ElementWise,
|
|
*fc_matmul_layer->getOutput(0),
|
|
*fc_bias_layer->getOutput(0),
|
|
nvinfer1::ElementWiseOperation::kSUM);
|
|
fc_add_layer->setName(
|
|
("multihead_matmul_fc_add(Output: " + output_name + ")").c_str());
|
|
if (op_desc.HasAttr("Input_scale")) {
|
|
PADDLE_ENFORCE_EQ(op_desc.HasAttr("fc_out_threshold"),
|
|
true,
|
|
common::errors::InvalidArgument(
|
|
"must have out threshold in multihead layers "
|
|
"in int8 mode"));
|
|
float out_scale =
|
|
PADDLE_GET_CONST(float, op_desc.GetAttr("fc_out_threshold"));
|
|
engine_->SetTensorDynamicRange(fc_add_layer->getOutput(0),
|
|
out_scale);
|
|
}
|
|
auto* fc_layer = fc_add_layer;
|
|
#else
|
|
// add shuffle for FullyConnected layer
|
|
std::vector<nvinfer1::ITensor*> reshape_before_fc_shape_tensor;
|
|
for (int i = 0; i < 5; i++) {
|
|
reshape_before_fc_shape_tensor.push_back(Add1DConstantLayer(1));
|
|
}
|
|
for (int i = 0; i < 3; i++) {
|
|
reshape_before_fc_shape_tensor[i] =
|
|
GetEleTensorOfShape(input_shape_tensor, i);
|
|
}
|
|
auto* reshape_before_fc_layer =
|
|
TRT_ENGINE_ADD_LAYER(engine_, Shuffle, *input);
|
|
reshape_before_fc_layer->setInput(
|
|
1, *Concat(reshape_before_fc_shape_tensor));
|
|
reshape_before_fc_layer->setName(
|
|
("shuffle_before_fc_multihead_matmul(Output: " + output_name +
|
|
")")
|
|
.c_str());
|
|
|
|
// add fc layer
|
|
nvinfer1::ILayer* fc_layer = nullptr;
|
|
if (op_desc.HasAttr("Input_scale")) {
|
|
engine_->SetTensorDynamicRange(
|
|
reshape_before_fc_layer->getOutput(0), in_scale);
|
|
nvinfer1::DimsHW nv_ksize(1, 1);
|
|
fc_layer =
|
|
TRT_ENGINE_ADD_LAYER(engine_,
|
|
ConvolutionNd,
|
|
*reshape_before_fc_layer->getOutput(0),
|
|
n,
|
|
nv_ksize,
|
|
weight,
|
|
bias);
|
|
PADDLE_ENFORCE_EQ(op_desc.HasAttr("fc_out_threshold"),
|
|
true,
|
|
common::errors::InvalidArgument(
|
|
"must have out threshold in multihead layers "
|
|
"in int8 mode"));
|
|
float out_scale =
|
|
PADDLE_GET_CONST(float, op_desc.GetAttr("fc_out_threshold"));
|
|
engine_->SetTensorDynamicRange(fc_layer->getOutput(0), out_scale);
|
|
} else {
|
|
fc_layer =
|
|
TRT_ENGINE_ADD_LAYER(engine_,
|
|
FullyConnected,
|
|
*reshape_before_fc_layer->getOutput(0),
|
|
n,
|
|
weight,
|
|
bias);
|
|
}
|
|
fc_layer->setName(
|
|
("multihead_matmul_fc(Output: " + output_name + ")").c_str());
|
|
#endif
|
|
|
|
// add shuffle for CustomQKVToContextPluginDynamic layer
|
|
auto* reshape_after_fc_layer =
|
|
TRT_ENGINE_ADD_LAYER(engine_, Shuffle, *fc_layer->getOutput(0));
|
|
std::vector<nvinfer1::ITensor*> mha_input_tensor_shape;
|
|
mha_input_tensor_shape.push_back(Add1DConstantLayer(-1));
|
|
mha_input_tensor_shape.push_back(
|
|
Add1DConstantLayer(hidden_out * 3)); // Q,K,V
|
|
mha_input_tensor_shape.push_back(Add1DConstantLayer(1));
|
|
mha_input_tensor_shape.push_back(Add1DConstantLayer(1));
|
|
reshape_after_fc_layer->setInput(1, *Concat(mha_input_tensor_shape));
|
|
reshape_after_fc_layer->setName(
|
|
("shuffle_after_fc_multihead_matmul(Output: " + output_name + ")")
|
|
.c_str());
|
|
|
|
// add mha_plugin
|
|
auto creator = GetPluginRegistry()->getPluginCreator(
|
|
"CustomQKVToContextPluginDynamic", "2");
|
|
assert(creator != nullptr);
|
|
// set the attributes of mha_plugin
|
|
int type = static_cast<int>(nvinfer1::DataType::kHALF);
|
|
int var_seqlen = 1;
|
|
bool has_mask = true;
|
|
std::vector<nvinfer1::PluginField> fields{
|
|
{"hidden_size",
|
|
&hidden_out,
|
|
nvinfer1::PluginFieldType::kINT32,
|
|
1},
|
|
{"num_heads", &head_number, nvinfer1::PluginFieldType::kINT32, 1},
|
|
{"type_id", &type, nvinfer1::PluginFieldType::kINT32, 1},
|
|
{"has_mask", &has_mask, nvinfer1::PluginFieldType::kINT32, 1},
|
|
{"var_seqlen",
|
|
&var_seqlen,
|
|
nvinfer1::PluginFieldType::kINT32,
|
|
1}};
|
|
std::unique_ptr<nvinfer1::PluginFieldCollection> plugin_collection(
|
|
new nvinfer1::PluginFieldCollection);
|
|
plugin_collection->nbFields = static_cast<int>(fields.size());
|
|
plugin_collection->fields = fields.data();
|
|
auto plugin = creator->createPlugin("CustomQKVToContextPluginDynamic",
|
|
plugin_collection.get());
|
|
plugin_collection.reset();
|
|
// set inputs
|
|
std::vector<nvinfer1::ITensor*> plugin_inputs;
|
|
// input_0 for plugin
|
|
plugin_inputs.emplace_back(reshape_after_fc_layer->getOutput(0));
|
|
// input_1(fake) for plugin
|
|
std::vector<int> mask = {1};
|
|
nvinfer1::ITensor* mask_tensor = Add1DConstantLayer(mask);
|
|
plugin_inputs.emplace_back(mask_tensor);
|
|
// input_2 for plugin
|
|
std::vector<int> pos_id = {0};
|
|
int max_batch = 512;
|
|
int length = (input_dims.d[1] == -1) ? 1 : input_dims.d[1];
|
|
for (int i = 1; i < max_batch; i++) {
|
|
pos_id.push_back(i * length);
|
|
}
|
|
nvinfer1::ITensor* fake_pos_id_tensor = Add1DConstantLayer(pos_id);
|
|
nvinfer1::ITensor* length_tensor =
|
|
GetEleTensorOfShape(input_shape_tensor, 1);
|
|
auto pos_id_layer =
|
|
TRT_ENGINE_ADD_LAYER(engine_,
|
|
ElementWise,
|
|
*fake_pos_id_tensor,
|
|
*length_tensor,
|
|
nvinfer1::ElementWiseOperation::kPROD);
|
|
// size = batch + 1;
|
|
nvinfer1::ITensor* batch_tensor =
|
|
GetEleTensorOfShape(input_shape_tensor, 0);
|
|
std::vector<int> const_data = {1};
|
|
nvinfer1::ITensor* const_tensor = Add1DConstantLayer(const_data);
|
|
auto size_layer =
|
|
TRT_ENGINE_ADD_LAYER(engine_,
|
|
ElementWise,
|
|
*batch_tensor,
|
|
*const_tensor,
|
|
nvinfer1::ElementWiseOperation::kSUM);
|
|
// get size(batch + 1) data from pos_id_tensor
|
|
nvinfer1::Dims start;
|
|
nvinfer1::Dims stride;
|
|
nvinfer1::Dims size;
|
|
|
|
start.nbDims = 1;
|
|
stride.nbDims = 1;
|
|
size.nbDims = 1;
|
|
|
|
start.d[0] = 0;
|
|
stride.d[0] = 1;
|
|
size.d[0] = 1;
|
|
|
|
nvinfer1::ITensor* pos_id_tensor = (input_dims.d[1] == -1)
|
|
? pos_id_layer->getOutput(0)
|
|
: fake_pos_id_tensor;
|
|
|
|
auto* slice_pos_layer = TRT_ENGINE_ADD_LAYER(
|
|
engine_, Slice, *pos_id_tensor, start, size, stride);
|
|
slice_pos_layer->setInput(2, *size_layer->getOutput(0));
|
|
plugin_inputs.emplace_back(slice_pos_layer->getOutput(0));
|
|
|
|
// input_3 for plugin
|
|
int max_length = (input_dims.d[1] == -1) ? 512 : input_dims.d[1];
|
|
std::vector<int> data(max_length, 1);
|
|
nvinfer1::ITensor* fake_max_seqlen_tensor = Add1DConstantLayer(data);
|
|
auto* slice_max_layer = TRT_ENGINE_ADD_LAYER(
|
|
engine_, Slice, *fake_max_seqlen_tensor, start, size, stride);
|
|
slice_max_layer->setInput(2, *length_tensor);
|
|
nvinfer1::ITensor* max_seqlen_tensor =
|
|
(input_dims.d[1] == -1) ? slice_max_layer->getOutput(0)
|
|
: fake_max_seqlen_tensor;
|
|
plugin_inputs.emplace_back(max_seqlen_tensor);
|
|
// plugin_layer
|
|
auto plugin_layer = engine_->network()->addPluginV2(
|
|
plugin_inputs.data(), plugin_inputs.size(), *plugin);
|
|
|
|
// add shuffle
|
|
auto* reshape_after_mha_layer = TRT_ENGINE_ADD_LAYER(
|
|
engine_, Shuffle, *plugin_layer->getOutput(0));
|
|
std::vector<nvinfer1::ITensor*> reshape_tensor;
|
|
reshape_tensor.push_back(batch_tensor);
|
|
reshape_tensor.push_back(length_tensor);
|
|
reshape_tensor.push_back(Add1DConstantLayer(-1));
|
|
reshape_after_mha_layer->setInput(1, *Concat(reshape_tensor));
|
|
reshape_after_mha_layer->setName(
|
|
("shuffle_last_multihead_matmul(Output: " + output_name + ")")
|
|
.c_str());
|
|
|
|
// return
|
|
layer = reshape_after_mha_layer;
|
|
ReplenishLayerAndOutput(
|
|
layer, "multihead_matmul", {output_name}, test_mode);
|
|
} else {
|
|
PADDLE_ENFORCE_EQ(
|
|
input->getDimensions().nbDims,
|
|
3,
|
|
common::errors::InvalidArgument(
|
|
"The Input dim of the MultiheadMatMul should be 3, "
|
|
"but it's (%d) now.",
|
|
input->getDimensions().nbDims));
|
|
// transpose weight_data from m * n to n * m
|
|
|
|
TensorRTEngine::Weight weight{nvinfer1::DataType::kFLOAT,
|
|
static_cast<void*>(weight_data),
|
|
static_cast<size_t>(weight_t->numel())};
|
|
weight.dims.assign({n, m});
|
|
|
|
TensorRTEngine::Weight bias{nvinfer1::DataType::kFLOAT,
|
|
static_cast<void*>(bias_data),
|
|
static_cast<size_t>(bias_t->numel())};
|
|
|
|
#if IS_TRT_VERSION_GE(10000)
|
|
auto* fc_weight_layer =
|
|
TRT_ENGINE_ADD_LAYER(engine_,
|
|
Constant,
|
|
nvinfer1::Dims3(1, n, hidden_in),
|
|
weight.get());
|
|
auto* fc_matmul_layer =
|
|
TRT_ENGINE_ADD_LAYER(engine_,
|
|
MatrixMultiply,
|
|
*input,
|
|
nvinfer1::MatrixOperation::kNONE,
|
|
*fc_weight_layer->getOutput(0),
|
|
nvinfer1::MatrixOperation::kTRANSPOSE);
|
|
fc_matmul_layer->setName(
|
|
("multihead_matmul_fc_matmul(Output: " + output_name + ")")
|
|
.c_str());
|
|
auto* fc_bias_layer = TRT_ENGINE_ADD_LAYER(
|
|
engine_, Constant, nvinfer1::Dims3(1, 1, n), bias.get());
|
|
auto* fc_add_layer =
|
|
TRT_ENGINE_ADD_LAYER(engine_,
|
|
ElementWise,
|
|
*fc_matmul_layer->getOutput(0),
|
|
*fc_bias_layer->getOutput(0),
|
|
nvinfer1::ElementWiseOperation::kSUM);
|
|
fc_add_layer->setName(
|
|
("multihead_matmul_fc_add(Output: " + output_name + ")").c_str());
|
|
if (op_desc.HasAttr("Input_scale")) {
|
|
PADDLE_ENFORCE_EQ(op_desc.HasAttr("fc_out_threshold"),
|
|
true,
|
|
common::errors::InvalidArgument(
|
|
"must have out threshold in multihead layers "
|
|
"in int8 mode"));
|
|
float out_scale =
|
|
PADDLE_GET_CONST(float, op_desc.GetAttr("fc_out_threshold"));
|
|
engine_->SetTensorDynamicRange(fc_add_layer->getOutput(0),
|
|
out_scale);
|
|
}
|
|
auto* reshape_after_fc_layer = TRT_ENGINE_ADD_LAYER(
|
|
engine_, Shuffle, *fc_add_layer->getOutput(0));
|
|
nvinfer1::Dims reshape_after_fc_layer_dim{};
|
|
reshape_after_fc_layer_dim.nbDims = 5;
|
|
reshape_after_fc_layer_dim.d[3] = 1;
|
|
reshape_after_fc_layer_dim.d[4] = 1;
|
|
reshape_after_fc_layer->setReshapeDimensions(
|
|
reshape_after_fc_layer_dim);
|
|
reshape_after_fc_layer->setName(
|
|
("shuffle_after_multihead_matmul(Output: " + output_name + ")")
|
|
.c_str());
|
|
auto* fc_layer = reshape_after_fc_layer;
|
|
#else
|
|
// add shuffle before fc
|
|
std::vector<nvinfer1::ITensor*> reshape_before_fc_shape_tensor;
|
|
nvinfer1::ITensor* input_shape_tensor = Shape(input);
|
|
|
|
for (int i = 0; i < 5; i++) {
|
|
reshape_before_fc_shape_tensor.push_back(Add1DConstantLayer(1));
|
|
}
|
|
for (int i = 0; i < 3; i++) {
|
|
reshape_before_fc_shape_tensor[i] =
|
|
GetEleTensorOfShape(input_shape_tensor, i);
|
|
}
|
|
auto* reshape_before_fc_layer =
|
|
TRT_ENGINE_ADD_LAYER(engine_, Shuffle, *input);
|
|
if (op_desc.HasAttr("Input_scale")) {
|
|
engine_->SetTensorDynamicRange(
|
|
reshape_before_fc_layer->getOutput(0), in_scale);
|
|
}
|
|
reshape_before_fc_layer->setInput(
|
|
1, *Concat(reshape_before_fc_shape_tensor));
|
|
reshape_before_fc_layer->setName(
|
|
("shuffle_before_multihead_matmul(Output: " + output_name + ")")
|
|
.c_str());
|
|
|
|
// add layer fc
|
|
nvinfer1::ILayer* fc_layer = nullptr;
|
|
if (op_desc.HasAttr("Input_scale")) {
|
|
nvinfer1::DimsHW nv_ksize(1, 1);
|
|
fc_layer =
|
|
TRT_ENGINE_ADD_LAYER(engine_,
|
|
ConvolutionNd,
|
|
*reshape_before_fc_layer->getOutput(0),
|
|
n,
|
|
nv_ksize,
|
|
weight.get(),
|
|
bias.get());
|
|
} else {
|
|
fc_layer =
|
|
TRT_ENGINE_ADD_LAYER(engine_,
|
|
FullyConnected,
|
|
*reshape_before_fc_layer->getOutput(0),
|
|
n,
|
|
weight.get(),
|
|
bias.get());
|
|
}
|
|
|
|
if (op_desc.HasAttr("fc_out_threshold")) {
|
|
PADDLE_ENFORCE_EQ(op_desc.HasAttr("fc_out_threshold"),
|
|
true,
|
|
common::errors::InvalidArgument(
|
|
"must have out threshold in multihead layers "
|
|
"in int8 mode"));
|
|
float out_scale =
|
|
PADDLE_GET_CONST(float, op_desc.GetAttr("fc_out_threshold"));
|
|
engine_->SetTensorDynamicRange(fc_layer->getOutput(0), out_scale);
|
|
}
|
|
fc_layer->setName(
|
|
("multihead_matmul_fc(Output: " + output_name + ")").c_str());
|
|
|
|
// no need to add shuffle after fc, just change it in
|
|
#endif
|
|
// QkvToContextPluginDynamic
|
|
|
|
// add qkv to context
|
|
int head_size = hidden_out / head_number;
|
|
float scale = PADDLE_GET_CONST(float, op_desc.GetAttr("alpha"));
|
|
|
|
std::vector<nvinfer1::ITensor*> plugin_inputs;
|
|
plugin_inputs.push_back(fc_layer->getOutput(0));
|
|
auto inputs = op_desc.Inputs();
|
|
bool hasBiasQK =
|
|
(inputs.find("BiasQK") == inputs.end()) ? false : true;
|
|
nvinfer1::ITensor* input_bias_qk = nullptr;
|
|
if (hasBiasQK) {
|
|
input_bias_qk =
|
|
engine_->GetITensor(op_desc.Input("BiasQK").front());
|
|
} else {
|
|
// fake input will be updated in qkv_plugin
|
|
input_bias_qk = fc_layer->getOutput(0);
|
|
}
|
|
plugin_inputs.push_back(input_bias_qk);
|
|
bool with_fp16 =
|
|
engine_->WithFp16() && !engine_->disable_trt_plugin_fp16();
|
|
|
|
if (engine_->precision() == phi::DataType::INT8) {
|
|
with_fp16 = true;
|
|
}
|
|
plugin::DynamicPluginTensorRT* plugin =
|
|
new plugin::QkvToContextPluginDynamic(
|
|
hidden_in, head_number, head_size, scale, with_fp16);
|
|
layer = engine_->AddDynamicPlugin(plugin_inputs.data(), 2, plugin);
|
|
ReplenishLayerAndOutput(
|
|
layer, "multihead_matmul", {output_name}, test_mode);
|
|
}
|
|
}
|
|
} else {
|
|
PADDLE_THROW(common::errors::Fatal(
|
|
"You are running the Ernie(Bert) model in static shape mode, which "
|
|
"is not supported for the time being.\n"
|
|
"You can use the config.SetTRTDynamicShapeInfo(...) interface to set "
|
|
"the shape information to run the dynamic shape mode."));
|
|
}
|
|
}
|
|
};
|
|
|
|
} // namespace paddle::inference::tensorrt
|
|
|
|
REGISTER_TRT_OP_CONVERTER(multihead_matmul, MultiheadMatMulOpConverter);
|