wehub/paddlepaddle--paddle
1
0
Fork 0
Code Issues Pull Requests Actions 2 Deployments Models Agent Notes Packages Projects Releases Wiki Activity
Files
ff97fb6869c2eeecb7b4c09af41ce69f43f2c574
paddlepaddle--paddle/paddle/fluid/inference/tensorrt/op_teller.cc
T
wehub-resource-sync e25996e7db chore: import upstream snapshot with attribution
2026-07-13 12:40:42 +08:00

3501 lines
119 KiB
C++
Raw Blame History

// Copyright (c) 2019 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/op_teller.h"
#include <bitset>
#include "paddle/fluid/framework/block_desc.h"
#include "paddle/fluid/framework/data_layout.h"
#include "paddle/fluid/framework/phi_utils.h"
#include "paddle/fluid/inference/tensorrt/dynamic_shape_infermeta_factory.h"
#include "paddle/phi/api/ext/op_meta_info.h"
#include "paddle/phi/core/compat/op_utils.h"
#include "paddle/phi/core/kernel_factory.h"
namespace paddle::framework {
class OpDesc;
} // namespace paddle::framework
namespace paddle::inference::tensorrt {
// Check if it is a dynamic shape. If it is a dynamic shape, return true;
// otherwise, return false
bool IsDynamicShapeOp(const framework::OpDesc& desc) {
VLOG(3) << "forbid_dynamic_op_enter_into_trt is open";
auto* block = desc.Block();
auto inputs = desc.Inputs();
for (auto iter : inputs) {
for (auto var_name : iter.second) {
if (block) {
auto* var_desc = block->FindVar(var_name);
const auto shape = var_desc->GetShape();
for (auto ele : shape) {
if (ele < 0) {
return true;
}
}
}
}
}
auto outputs = desc.Outputs();
for (auto iter : outputs) {
for (auto var_name : iter.second) {
if (block) {
auto* var_desc = block->FindVar(var_name);
const auto shape = var_desc->GetShape();
for (auto ele : shape) {
if (ele < 0) {
return true;
}
}
}
}
}
return false;
}
// Just tell by the op_types.
struct SimpleOpTypeSetTeller : public Teller {
SimpleOpTypeSetTeller() { // NOLINT
// use TensorRT plugin
teller_set.insert("group_norm");
teller_set.insert("multiclass_nms3");
teller_set.insert("multiclass_nms");
int8_teller_set.insert("multiclass_nms3");
int8_teller_set.insert("multiclass_nms");
teller_set.insert("tile");
int8_teller_set.insert("tile");
teller_set.insert("flatten_contiguous_range");
int8_teller_set.insert("flatten_contiguous_range");
teller_set.insert("rnn");
int8_teller_set.insert("rnn");
teller_set.insert("fill_constant_batch_size_like");
int8_teller_set.insert("fill_constant_batch_size_like");
teller_set.insert("reshape");
teller_set.insert("reshape2");
int8_teller_set.insert("reshape");
int8_teller_set.insert("reshape2");
teller_set.insert("sparse_fc");
int8_teller_set.insert("sparse_fc");
teller_set.insert("sparse_multihead_matmul");
int8_teller_set.insert("sparse_multihead_matmul");
#if IS_TRT_VERSION_GE(8522)
teller_set.insert("flash_multihead_matmul");
int8_teller_set.insert("flash_multihead_matmul");
teller_set.insert("cross_multihead_matmul");
int8_teller_set.insert("cross_multihead_matmul");
teller_set.insert("qk_multihead_matmul");
int8_teller_set.insert("qk_multihead_matmul");
#endif
#if IS_TRT_VERSION_GE(8200)
teller_set.insert("round");
int8_teller_set.insert("round");
teller_set.insert("set_value");
teller_set.insert("index_select");
int8_teller_set.insert("index_select");
int8_teller_set.insert("einsum");
teller_set.insert("einsum");
#endif
}
bool operator()(const framework::OpDesc& desc,
bool use_no_calib_int8 = false,
bool with_dynamic_shape = false,
bool forbid_dynamic_op_enter_into_trt = false,
bool use_explicit_quantization = false) override {
const std::string op_type = desc.Type();
std::unordered_set<std::string> control_set = {"conditional_block",
"while"};
std::unordered_set<std::string> feed_fetch_set = {"feed", "fetch"};
if (control_set.find(op_type) != control_set.end()) {
return false;
}
if (feed_fetch_set.find(op_type) != feed_fetch_set.end()) {
return false;
}
if (forbid_dynamic_op_enter_into_trt && IsDynamicShapeOp(desc)) {
return false;
}
// do not support the op which is labeled the `skip_quant`
if ((desc.HasAttr("namescope") &&
PADDLE_GET_CONST(std::string, desc.GetAttr("op_namescope")) ==
"/skip_quant_2/") ||
desc.HasAttr("skip_quant"))
return false;
std::unordered_set<std::string> act_op_list = {
"relu", "relu6", "sigmoid",
"elu", "selu", "softsign",
"softplus", "stanh", "thresholded_relu",
"exp", "log", "sqrt",
"abs", "sin", "cos",
"tan", "tanh", "sinh",
"cosh", "asin", "acos",
"atan", "asinh", "acosh",
"atanh", "ceil", "celu",
"erf", "floor", "round",
"sign", "silu", "logical_not",
"reciprocal", "tanh_shrink", "logsigmoid",
"rsqrt", "swish", "hard_sigmoid",
"hard_swish", "leaky_relu"};
std::unordered_set<std::string> unary_list = {
"exp", "log", "sqrt", "abs", "sin",
"cos", "tan", "tanh", "sinh", "cosh",
"asin", "acos", "atan", "asinh", "acosh",
"atanh", "ceil", "celu", "floor", "round",
"sign", "logical_not", "reciprocal", "tanh_shrink", "logsigmoid",
"erf", "bitwise_not", "equal", "not_equal", "rsqrt"};
// Static shape does not support 0 or 1 dim's input.
if (!with_dynamic_shape) {
auto inputs = desc.Inputs();
for (auto iter : inputs) {
for (auto var_name : iter.second) {
auto* block = desc.Block();
if (block) {
auto* var_desc = block->FindVarRecursive(var_name);
// Can't get feed op's TensorDesc
if (op_type != "feed" && var_desc && !var_desc->Persistable()) {
const auto shape = var_desc->GetShape();
if (shape.size() == 1 || shape.empty()) return false;
}
}
}
}
}
if (act_op_list.find(op_type) != act_op_list.end()) {
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3) << "The block desc is nullptr, we can't continue to analyze. "
"Developers need to check whether block_desc is passed in "
"the pass.";
return false;
}
auto x_var_name = desc.Input("X")[0];
auto* x_var_desc = block->FindVarRecursive(x_var_name);
auto x_dtype = x_var_desc->GetDataType();
if (x_dtype == framework::proto::VarType::COMPLEX64 ||
x_dtype == framework::proto::VarType::COMPLEX128) {
VLOG(3) << op_type
<< " op does not support COMPLEX64 or COMPLEX128 input";
return false;
}
#if !IS_TRT_VERSION_GE(8600)
const auto x_shape = x_var_desc->GetShape();
if (x_shape.empty() && unary_list.find(op_type) != unary_list.end()) {
VLOG(3) << op_type
<< " op does not support 0 dim input when TensorRT < 8.6.";
return false;
}
#endif
}
if (op_type == "dropout") {
/*
* Some OpDescs Attribute support both constant value and dynamic
* runtime value (which is a Variable(s) type). But TensorRT maybe
* only support constant value Attribute, so we shall distinguish
* this case in time and return False in OpTeller.Tell().
* If Attribute is Variable(s), HasAttr() will return False
*/
if (!desc.HasAttr("dropout_prob", /*with_attr_var=*/false)) {
VLOG(3)
<< "Skip to convert into TRT while found Attribute('dropout_prob') "
"is Variable type in dropout.";
return false;
}
}
if (op_type == "pool2d") {
// If Attribute is Variable(s), HasAttr() will return False
if (!desc.HasAttr("ksize", /*with_attr_var=*/false)) {
VLOG(3) << "Skip to convert into TRT while found Attribute('ksize') is "
"Variable type in pool2d.";
return false;
}
std::vector<int> paddings =
PADDLE_GET_CONST(std::vector<int>, desc.GetAttr("paddings"));
if (paddings.size() > 2) {
return false;
}
if (desc.Input("X").size() != 1) {
VLOG(3) << "TRT Pool2d expect 1 input, but got "
<< desc.Input("X").size();
return false;
}
if (desc.Output("Out").size() != 1) {
VLOG(3) << "TRT Pool2d has only 1 output, but got "
<< desc.Output("Out").size();
return false;
}
if (desc.HasAttr("data_format")) {
std::string data_format =
PADDLE_GET_CONST(std::string, desc.GetAttr("data_format"));
if (data_format == "NHWC" || data_format == "NDHWC") {
return false;
}
}
if (!desc.HasAttr("pooling_type")) {
return false;
} else {
std::string pool_type =
PADDLE_GET_CONST(std::string, desc.GetAttr("pooling_type"));
if (pool_type != "max" && pool_type != "avg") {
VLOG(3) << "Wrong pool op type, the trt do not support the "
<< pool_type << " pool type.";
return false;
}
if (pool_type == "avg") {
if (desc.HasAttr("global_pooling")) {
if (!PADDLE_GET_CONST(bool, desc.GetAttr("global_pooling"))) {
if (desc.HasAttr("exclusive")) {
if (PADDLE_GET_CONST(bool, desc.GetAttr("exclusive"))) {
std::vector<int> ksize =
PADDLE_GET_CONST(std::vector<int>, desc.GetAttr("ksize"));
for (size_t i = 0; i < ksize.size(); i++) {
if (ksize[i] <= paddings[i]) {
VLOG(3) << "the padding size should be less than the "
"filter size "
"for exclusive-counting pooling.";
return false;
}
}
}
}
}
}
}
}
}
if (op_type == "conv2d" || op_type == "conv2d_transpose" ||
op_type == "fused_conv2d_add_act" || op_type == "depthwise_conv2d" ||
op_type == "depthwise_conv2d_transpose") {
if (desc.Input("Input").size() != 1) {
VLOG(3) << "TRT Conv2d expect 1 input, but got "
<< desc.Input("Input").size() << " input.";
return false;
}
if (desc.Input("Filter").size() != 1) {
VLOG(3) << "TRT Conv2d expect 1 filter, but got "
<< desc.Input("Filter").size() << " filter.";
return false;
}
if (desc.HasAttr("enable_int8")) {
if (op_type == "conv2d" || op_type == "fused_conv2d_add_act") {
if (!desc.HasAttr("Input_scale")) {
VLOG(3) << "Input scale not found. TRT int8"
" requires conv/deconv to have "
"input quantization scales.";
return false;
}
}
}
if (op_type == "conv2d_transpose" ||
op_type == "depthwise_conv2d_transpose") {
if (!desc.HasAttr("dilations")) {
return false;
} else {
const std::vector<int> dilations =
PADDLE_GET_CONST(std::vector<int>, desc.GetAttr("dilations"));
if (dilations[0] != 1 || dilations[1] != 1) {
VLOG(3) << "In conv2d_transpose, Dilations must be (1, 1) for "
"tensorRT, but given ("
<< dilations[0] << ", " << dilations[1] << ")";
return false;
}
}
}
if (desc.Output("Output").size() != 1) {
VLOG(3) << "TRT Conv2d expect 1 output, but got "
<< desc.Output("Output").size() << " output.";
return false;
}
auto* block = desc.Block();
if (block) {
auto* filter_var_desc =
block->FindVarRecursive(desc.Input("Filter")[0]);
if (!filter_var_desc->Persistable()) {
#if IS_TRT_VERSION_GE(8600)
#else
LOG(INFO)
<< "Trt below 8.6 not support conv2d's filter is a intermediate "
"tensor in conv2d op, please upgrade your TensorRT.";
return false;
#endif
}
}
}
if (op_type == "deformable_conv") {
if (!desc.HasAttr("groups") || !desc.HasAttr("strides") ||
!desc.HasAttr("paddings"))
return false;
auto* block = desc.Block();
auto input_name = desc.Input("Input")[0];
auto* input_desc = block->FindVarRecursive(input_name);
const auto input_shape = input_desc->GetShape();
if (input_shape.size() != 4) {
VLOG(3) << "Input of deformable conv should be 4-D Tensor, but got "
<< input_shape.size();
return false;
}
auto filter_name = desc.Input("Filter")[0];
auto* filter_desc = block->FindVarRecursive(filter_name);
const auto filter_shape = filter_desc->GetShape();
int groups = PADDLE_GET_CONST(int, desc.GetAttr("groups"));
if (input_shape[1] != filter_shape[1] * groups) {
VLOG(3) << "The number of input channels should be equal to filter "
<< "channels * groups. But got input channels "
<< input_shape[1] << "filter channels " << filter_shape[1];
return false;
}
const std::vector<int> strides =
PADDLE_GET_CONST(std::vector<int>, desc.GetAttr("strides"));
if (strides.size() != 2) {
VLOG(3) << "The size of strides should be 2, but got "
<< strides.size();
return false;
}
const std::vector<int> paddings =
PADDLE_GET_CONST(std::vector<int>, desc.GetAttr("paddings"));
if (paddings.size() != 2) {
VLOG(3) << "The size of paddings should be 2, but got "
<< paddings.size();
return false;
}
}
if (op_type == "bmm") {
if (!with_dynamic_shape) {
return false;
}
}
if (op_type == "range") {
if (!with_dynamic_shape) {
return false;
}
#if IS_TRT_VERSION_LT(8400)
auto* block = desc.Block();
auto start_var_name = desc.Input("Start")[0];
auto* start_var_desc = block->FindVarRecursive(start_var_name);
auto start_dtype = start_var_desc->GetDataType();
if (start_dtype == framework::proto::VarType::FP32 ||
start_dtype == framework::proto::VarType::FP64) {
return false;
}
#endif
}
if (op_type == "sign") {
#if IS_TRT_VERSION_GE(8200)
if (!with_dynamic_shape) {
return false;
}
#else
VLOG(3) << "sign op is only supported by trt8.2 above ";
return false;
#endif
}
if (op_type == "logical_not") {
#if IS_TRT_VERSION_GE(8400)
if (!with_dynamic_shape) {
return false;
}
#else
VLOG(3) << "logical_not op is only supported by trt8.4 above because of "
"cast op";
return false;
#endif
}
if (op_type == "softmax") {
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3) << "The block desc is nullptr, we can't continue to analyze. "
"Developers need to check whether block_desc is passed in "
"the pass.";
return false;
}
auto x_var_name = desc.Input("X")[0];
auto* x_var_desc = block->FindVarRecursive(x_var_name);
const auto x_shape = x_var_desc->GetShape();
if (with_dynamic_shape && (x_shape.size() == 1 || x_shape.empty())) {
int axis = desc.HasAttr("axis")
? PADDLE_GET_CONST(int, desc.GetAttr("axis"))
: -1;
if (axis > 0) {
return false;
}
}
}
if (op_type == "group_norm") {
if (!desc.HasAttr("epsilon") || !desc.HasAttr("groups") ||
!desc.HasAttr("data_layout"))
return false;
auto registry = GetPluginRegistry();
if (registry == nullptr) return false;
std::string layout_str =
PADDLE_GET_CONST(std::string, desc.GetAttr("data_layout"));
if (layout_str != "NCHW") {
VLOG(3) << "Group norm trt plugin only support NCHW layout, but got "
<< layout_str;
return false;
}
}
if (op_type == "concat") {
if (!desc.HasAttr("axis")) {
return false;
}
int axis = PADDLE_GET_CONST(int, desc.GetAttr("axis"));
if (!with_dynamic_shape) {
if (axis == 0) return false;
}
auto concat_inputs = desc.Inputs();
if (concat_inputs.find("AxisTensor") != concat_inputs.end()) {
if (!desc.Input("AxisTensor").empty()) {
return false;
}
}
}
if (op_type == "transpose2" || op_type == "transpose") {
if (!desc.HasAttr("axis")) {
return false;
}
std::vector<int> axis =
PADDLE_GET_CONST(std::vector<int>, desc.GetAttr("axis"));
if (!with_dynamic_shape && axis[0] != 0) return false;
if (axis.size() >= nvinfer1::Dims::MAX_DIMS) return false;
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3) << "The block desc is nullptr, we can't continue to analyze. "
"Developers need to check whether block_desc is passed in "
"the pass.";
return false;
}
auto x_var_name = desc.Input("X")[0];
auto* x_var_desc = block->FindVarRecursive(x_var_name);
const auto x_shape = x_var_desc->GetShape();
if (axis.size() != x_shape.size()) return false;
int dims = x_shape.size();
std::vector<int> perm(nvinfer1::Dims::MAX_DIMS);
for (int i = 0; i < dims; i++) {
perm[i] = axis[i];
}
auto is_valid_permutation = [&](int dims,
const std::vector<int>& permutation) {
std::bitset<nvinfer1::Dims::MAX_DIMS> found;
for (int i = 0; i < dims; ++i) {
const int x = permutation[i];
if ((x < 0) || (x >= dims) || found[x])
return false; // Out of bounds or duplicate
found.set(x);
}
return true;
};
if (!is_valid_permutation(dims, perm)) {
VLOG(3) << "Invalid permutation dimensions for trt transpose op "
"converter: duplicate or out of bound.";
return false;
}
}
if (op_type == "flatten2" || op_type == "flatten") {
if (!desc.HasAttr("axis")) {
return false;
} else {
int axis = PADDLE_GET_CONST(int, desc.GetAttr("axis"));
if (axis != 1) return false;
}
}
if (op_type == "flatten_contiguous_range") {
if (!with_dynamic_shape) {
if (!desc.HasAttr("start_axis") || !desc.HasAttr("stop_axis")) {
return false;
}
int start_axis = PADDLE_GET_CONST(int, desc.GetAttr("start_axis"));
int stop_axis = PADDLE_GET_CONST(int, desc.GetAttr("stop_axis"));
auto x_var_name = desc.Input("X")[0];
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3) << "The block desc is nullptr, we can't continue to analyze. "
"Developers need to check whether block_desc is passed in "
"the pass.";
return false;
}
auto* x_var_desc = block->FindVarRecursive(x_var_name);
const auto x_shape = x_var_desc->GetShape();
int dims = x_shape.size();
if (dims == 0) {
VLOG(3) << op_type
<< " op does not support input's dim is 0 in tensorrt "
"static shape mode.";
return false;
}
if (start_axis < 0) start_axis += dims;
if (start_axis == 0) {
VLOG(3) << "TRT flatten_contiguous_range not support the "
"batch-dimension being changed";
return false;
}
if (stop_axis < 0) stop_axis += dims;
for (int i = start_axis; i <= stop_axis; ++i) {
if (x_shape[i] < 0) {
VLOG(3) << "On TRT static shape,flatten_contiguous_range input dim "
"should be > 0";
return false;
}
}
}
}
if (op_type == "gather") {
auto gather_inputs = desc.Inputs();
if (gather_inputs.find("Axis") != gather_inputs.end()) {
if (!desc.Input("Axis").empty()) {
return false;
}
}
if (!with_dynamic_shape) {
return false;
} else {
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3) << "The block desc is nullptr, we can't continue to analyze. "
"Developers need to check whether block_desc is passed in "
"the pass.";
return false;
}
}
}
if (op_type == "gather_nd") {
if (!with_dynamic_shape) return false;
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3) << "The block desc is nullptr, we can't continue to analyze. "
"Developers need to check whether block_desc is passed in "
"the pass.";
return false;
}
#if IS_TRT_VERSION_LT(8200)
auto index_var_name = desc.Input("Index")[0];
auto* index_var_desc = block->FindVarRecursive(index_var_name);
auto x_var_name = desc.Input("X")[0];
auto* x_var_desc = block->FindVarRecursive(x_var_name);
const auto index_shape = index_var_desc->GetShape();
const auto x_shape = x_var_desc->GetShape();
if (x_shape.size() <= 2) {
VLOG(3) << "gather_nd op requires the input's dimension to be greater "
"than 2";
return false;
}
if (x_shape.size() != index_shape.size()) {
VLOG(3) << "gather_nd op Index input dims size [" << index_shape.size()
<< " ] not equal to x dims size [" << x_shape.size() << "]";
return false;
}
#endif
}
if (op_type == "index_select") {
#if !IS_TRT_VERSION_GE(8200)
return false;
#endif
auto gather_inputs = desc.Inputs();
if (!with_dynamic_shape) {
return false;
} else {
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3) << "The block desc is nullptr, we can't continue to analyze. "
"Developers need to check whether block_desc is passed in "
"the pass.";
return false;
}
auto index_var_name = desc.Input("Index")[0];
auto* index_var_desc = block->FindVarRecursive(index_var_name);
// The index input must be int32 or int64 datatype.
if (index_var_desc->GetDataType() !=
paddle::framework::proto::VarType_Type::VarType_Type_INT32 &&
index_var_desc->GetDataType() !=
paddle::framework::proto::VarType_Type::VarType_Type_INT64) {
VLOG(3)
<< "Index select op Index input data type must be int32 or int64";
return false;
}
}
}
if (op_type == "take_along_axis") {
#if IS_TRT_VERSION_GE(8200)
if (!with_dynamic_shape) return false;
auto* block = desc.Block();
auto input_var_name = desc.Input("Input")[0];
auto index_var_name = desc.Input("Index")[0];
auto* input_var_desc = block->FindVarRecursive(input_var_name);
auto* index_var_desc = block->FindVarRecursive(index_var_name);
const auto input_shape = input_var_desc->GetShape();
const auto index_shape = index_var_desc->GetShape();
if (input_shape.size() != index_shape.size()) {
VLOG(3) << "take_along_axis op Index input dims size ["
<< index_shape.size() << " ] not equal to input dims size ["
<< input_shape.size() << "]";
return false;
}
#else
VLOG(3) << "take_along_axis op is only supported by trt8.2 above ";
return false;
#endif
}
if (op_type == "anchor_generator") {
if (!with_dynamic_shape) return false;
}
if (op_type == "yolo_box") {
return false;
}
if (op_type == "yolo_box_head") {
return false;
}
if (op_type == "arg_max" || op_type == "arg_min") {
if (!desc.HasAttr("axis", /*with_attr_var=*/false)) {
VLOG(3) << "Skip to convert into TRT while found Attribute('axis') is "
"Variable type in arg_max.";
return false;
}
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3) << "The block desc is nullptr, we can't continue to analyze. "
"Developers need to check whether block_desc is passed in "
"the pass.";
return false;
}
auto x_var_name = desc.Input("X")[0];
auto* x_var_desc = block->FindVarRecursive(x_var_name);
auto x_dtype = x_var_desc->GetDataType();
if (!(x_dtype == framework::proto::VarType::FP32 ||
x_dtype == framework::proto::VarType::FP16 ||
x_dtype == framework::proto::VarType::FP64)) {
return false;
}
int axis = desc.HasAttr("axis")
? PADDLE_GET_CONST(int64_t, desc.GetAttr("axis"))
: -1;
bool flatten = desc.HasAttr("flatten")
? PADDLE_GET_CONST(bool, desc.GetAttr("flatten"))
: false;
int dtype = desc.HasAttr("dtype")
? PADDLE_GET_CONST(int, desc.GetAttr("dtype"))
: 3;
if (axis == 0 || flatten || (dtype != 2 && dtype != 3)) return false;
}
if (op_type == "affine_channel") {
if (!desc.HasAttr("data_layout")) return false;
auto data_layout = common::StringToDataLayout(
PADDLE_GET_CONST(std::string, desc.GetAttr("data_layout")));
if (data_layout != phi::DataLayout::NCHW) return false;
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3) << "The block desc is nullptr, we can't continue to analyze. "
"Developers need to check whether block_desc is passed in "
"the pass.";
return false;
}
auto x_var_name = desc.Input("X")[0];
auto* x_var_desc = block->FindVarRecursive(x_var_name);
const auto x_shape = x_var_desc->GetShape();
if (x_shape.size() == 2) {
return false;
}
}
if (op_type == "multiclass_nms" || op_type == "multiclass_nms3") {
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3) << "The block desc is nullptr, we can't continue to analyze. "
"Developers need to check whether block_desc is passed in "
"the pass.";
return false;
}
auto multiclass_nms_inputs = desc.Inputs();
if (multiclass_nms_inputs.find("RoisNum") !=
multiclass_nms_inputs.end()) {
if (!desc.Input("RoisNum").empty()) {
return false;
}
}
for (auto& param_name : multiclass_nms_inputs) {
for (auto& var_name : param_name.second) {
auto* var_desc = block->FindVarRecursive(var_name);
const auto shape = var_desc->GetShape();
if (shape.size() != 3) {
VLOG(3) << "multiclass_nms op dims != 3 not supported in tensorrt, "
"but got dims "
<< shape.size() << ", so jump it.";
return false;
}
}
}
bool has_attrs =
(desc.HasAttr("background_label") &&
desc.HasAttr("score_threshold") && desc.HasAttr("nms_top_k") &&
desc.HasAttr("keep_top_k") && desc.HasAttr("normalized"));
if (has_attrs == false) return false;
// TODO(wangxinxin08): tricky solution because the outputs of batchedNMS
// plugin are not constient with those of multiclass_nms3
if (desc.HasAttr("nms_eta") == false) return false;
auto nms_eta = PADDLE_GET_CONST(float, desc.GetAttr("nms_eta"));
if (nms_eta <= 1.0) return false;
auto nms_top_k = PADDLE_GET_CONST(int, desc.GetAttr("nms_top_k"));
if (nms_top_k < 0) return false;
auto keep_top_k = PADDLE_GET_CONST(int, desc.GetAttr("keep_top_k"));
if (keep_top_k < 0) return false;
auto registry = GetPluginRegistry();
if (registry == nullptr) return false;
}
if (op_type == "nearest_interp") {
std::vector<std::string> attrs{
"interp_method", "align_corners", "scale", "out_h", "out_w"};
for (auto const& attr : attrs) {
if (!desc.HasAttr(attr)) return false;
}
if (desc.HasAttr("data_layout")) {
auto data_layout = common::StringToDataLayout(
PADDLE_GET_CONST(std::string, desc.GetAttr("data_layout")));
if (data_layout != phi::DataLayout::NCHW &&
data_layout != phi::DataLayout::NHWC)
return false;
}
auto interp_method =
PADDLE_GET_CONST(std::string, desc.GetAttr("interp_method"));
if (interp_method != "nearest") return false;
auto scale = PADDLE_GET_CONST(float, desc.GetAttr("scale"));
auto out_h = PADDLE_GET_CONST(int, desc.GetAttr("out_h"));
auto out_w = PADDLE_GET_CONST(int, desc.GetAttr("out_w"));
auto align_corners =
PADDLE_GET_CONST(bool, desc.GetAttr("align_corners"));
if (!(scale > 0.f && (out_h <= 0 && out_w <= 0))) {
if (out_h <= 0) {
VLOG(3) << "out_h must be greater than 0 if scale is not set.";
return false;
}
if (out_w <= 0) {
VLOG(3) << "out_w must be greater than 0 if scale is not set.";
return false;
}
}
if ((scale <= 0.f) && with_dynamic_shape) {
VLOG(3) << "dynamic shape not support scale not set.";
return false;
}
// When align_corners = true, the paddle's and trt_layer's results has
// diff
if (align_corners && scale != 1) {
return false;
}
}
if (op_type == "nearest_interp_v2") {
std::vector<std::string> attrs{"data_layout",
"interp_method",
"align_corners",
"scale",
"out_h",
"out_w"};
for (auto const& attr : attrs) {
if (!desc.HasAttr(attr)) return false;
}
auto data_layout = common::StringToDataLayout(
PADDLE_GET_CONST(std::string, desc.GetAttr("data_layout")));
if (data_layout != phi::DataLayout::NCHW &&
data_layout != phi::DataLayout::NHWC)
return false;
auto interp_method =
PADDLE_GET_CONST(std::string, desc.GetAttr("interp_method"));
if (interp_method != "nearest") return false;
#if IS_TRT_VERSION_GE(8200)
auto resize_inputs = desc.Inputs();
if (with_dynamic_shape &&
resize_inputs.find("SizeTensor") != resize_inputs.end() &&
desc.Input("SizeTensor").size() == 2) {
return true;
}
#endif
auto scale = PADDLE_GET_CONST(std::vector<float>, desc.GetAttr("scale"));
auto out_h = PADDLE_GET_CONST(int, desc.GetAttr("out_h"));
auto out_w = PADDLE_GET_CONST(int, desc.GetAttr("out_w"));
if (!(out_h > 0 && out_w > 0)) {
if (scale.size() < 2) return false;
if (scale[0] <= 0.f || scale[1] <= 0.f) {
VLOG(3) << "scale factor must be greater than 0 if out_h or out_w is "
"not set.";
return false;
}
}
}
if (op_type == "bilinear_interp_v2") {
std::vector<std::string> attrs{"data_layout",
"interp_method",
"align_corners",
"scale",
"out_h",
"out_w"};
for (auto const& attr : attrs) {
if (!desc.HasAttr(attr)) {
VLOG(3) << "The op_type " << op_type << " doesn't have the attr "
<< attr << " and return false";
return false;
}
}
auto resize_inputs = desc.Inputs();
if (resize_inputs.find("SizeTensor") != resize_inputs.end()) {
#if IS_TRT_VERSION_GE(8200)
if (desc.Input("SizeTensor").size() == 2) {
// TODO(lizexu123): When SizeTensor exists, at least one of the input
// variable names must contain 'shape' in order for TRT conversion to
// proceed; otherwise, TRT conversion will be disallowed."
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3)
<< "The block desc is nullptr,we can't continue to analyze.";
return false;
}
bool valid_source = false;
//
std::vector<std::string> size_tensor_names = desc.Input("SizeTensor");
for (const auto& tensor_name : size_tensor_names) {
auto* var_desc = block->FindVarRecursive(tensor_name);
if (!var_desc) continue;
if (tensor_name.find("shape") != std::string::npos) {
valid_source = true;
break;
}
}
if (!valid_source) {
VLOG(3) << "The SizeTensor for bilinear_interp_v2 doesn't come "
"from a valid source.";
return false;
}
return true;
}
#else
if (!desc.Input("SizeTensor").empty()) {
VLOG(3)
<< "The Paddle-TRT doesn't support the SizeTensor for op_type "
<< op_type;
return false;
}
#endif
}
if (resize_inputs.find("OutSize") != resize_inputs.end()) {
if (!with_dynamic_shape) {
VLOG(3) << "Static shape don't support the OutSize for op_type "
<< op_type;
return false;
}
}
auto data_layout = common::StringToDataLayout(
PADDLE_GET_CONST(std::string, desc.GetAttr("data_layout")));
if (data_layout != phi::DataLayout::NCHW &&
data_layout != phi::DataLayout::NHWC) {
VLOG(3) << "The op_type " << op_type
<< " is not NCHW or NHWC return false";
return false;
}
auto interp_method =
PADDLE_GET_CONST(std::string, desc.GetAttr("interp_method"));
if (interp_method != "bilinear") {
VLOG(3) << "The interp_method of op_type " << op_type
<< " is not bilinear";
return false;
}
bool has_scale_input_size =
(resize_inputs.find("Scale") != resize_inputs.end());
if (!has_scale_input_size ||
(has_scale_input_size && desc.Input("Scale").size() != 1)) {
const std::vector<float> scale =
PADDLE_GET_CONST(std::vector<float>, desc.GetAttr("scale"));
if (scale.size() <= 1) {
if (!desc.HasAttr("out_h") || !desc.HasAttr("out_w")) {
VLOG(3) << "The op_type " << op_type
<< " doesn't have Scale and the scale size <=1 and without "
"out_h / out_w, it will return false";
return false;
}
auto out_h = PADDLE_GET_CONST(int, desc.GetAttr("out_h"));
auto out_w = PADDLE_GET_CONST(int, desc.GetAttr("out_w"));
if (!(out_h <= 0 && out_w <= 0)) {
if (out_h <= 0) {
VLOG(3) << "The op_type " << op_type
<< "'s out_h must be greater than 0 if scale is not set.";
return false;
}
if (out_w <= 0) {
VLOG(3) << "The op_type " << op_type
<< "'s out_w must be greater than 0 if scale is not set.";
return false;
}
}
} else {
for (size_t i = 0; i < scale.size(); i++) {
if (scale[i] <= 0 && with_dynamic_shape) {
VLOG(3) << "dynamic shape not support Attr(scale[" << i << "]) "
<< scale[i]
<< " less than 1 and Input(Scale) vector not set.";
return false;
}
}
}
}
}
if (op_type == "linear_interp_v2") {
std::vector<std::string> attrs{"data_layout",
"interp_method",
"align_corners",
"scale",
"out_h",
"out_w"};
for (auto const& attr : attrs) {
if (!desc.HasAttr(attr)) {
VLOG(3) << "The op_type " << op_type << " doesn't have the attr "
<< attr << " and return false";
return false;
}
}
auto resize_inputs = desc.Inputs();
if (resize_inputs.find("SizeTensor") != resize_inputs.end()) {
#if IS_TRT_VERSION_GE(8200)
if (desc.Input("SizeTensor").size() == 1) {
return true;
}
#else
if (!desc.Input("SizeTensor").empty()) {
VLOG(3)
<< "The Paddle-TRT doesn't support the SizeTensor for op_type "
<< op_type;
return false;
}
#endif
}
if (resize_inputs.find("OutSize") != resize_inputs.end()) {
if (!with_dynamic_shape) {
VLOG(3) << "Static shape don't support the OutSize for op_type "
<< op_type;
return false;
}
}
auto data_layout = common::StringToDataLayout(
PADDLE_GET_CONST(std::string, desc.GetAttr("data_layout")));
if (data_layout != phi::DataLayout::NCHW &&
data_layout != phi::DataLayout::NHWC) {
VLOG(3) << "The op_type " << op_type
<< " is not NCHW or NHWC return false";
return false;
}
auto interp_method =
PADDLE_GET_CONST(std::string, desc.GetAttr("interp_method"));
if (interp_method != "linear") {
VLOG(3) << "The interp_method of op_type " << op_type
<< " is not linear";
return false;
}
bool has_scale_input_size =
(resize_inputs.find("Scale") != resize_inputs.end());
if (!has_scale_input_size ||
(has_scale_input_size && desc.Input("Scale").size() != 1)) {
const std::vector<float> scale =
PADDLE_GET_CONST(std::vector<float>, desc.GetAttr("scale"));
if (scale.size() == 0) {
if (!desc.HasAttr("out_w")) {
VLOG(3) << "The op_type " << op_type
<< " doesn't have Scale and the scale size <=1 and without "
" out_w, it will return false";
return false;
}
auto out_w = PADDLE_GET_CONST(int, desc.GetAttr("out_w"));
if (out_w <= 0) {
VLOG(3) << "The op_type " << op_type
<< "'s out_w must be greater than 0 if scale is not set.";
return false;
}
} else {
for (size_t i = 0; i < scale.size(); i++) {
if (scale[i] <= 0 && with_dynamic_shape) {
VLOG(3) << "dynamic shape not support Attr(scale[" << i << "]) "
<< scale[i]
<< " less than 1 and Input(Scale) vector not set.";
return false;
}
}
}
}
}
if (op_type == "unsqueeze2") {
std::vector<int> axes;
if (desc.HasAttr("axes")) {
axes = PADDLE_GET_CONST(std::vector<int>, desc.GetAttr("axes"));
}
if (axes.empty()) {
VLOG(3) << "The necessary attributes of the squeeze2 operator axes is "
"missing.";
return false;
}
if (!with_dynamic_shape) {
if (std::find(axes.begin(), axes.end(), 0) != axes.end()) {
VLOG(3) << "Invalid squeeze axes. Axes having batch axis is not "
"supported in static shape";
return false;
}
}
}
if (op_type == "batch_norm") {
const std::vector<std::string> bn_inputs = {
"X", "Bias", "Mean", "Scale", "Variance"};
for (unsigned int i = 0; i < bn_inputs.size(); i++) {
if (desc.Input(bn_inputs[i]).size() != 1) {
VLOG(3) << "Invalid " << bn_inputs[i]
<< "'s size of batch_norm TRT "
"converter. Expected 1, received "
<< desc.Input(bn_inputs[i]).size() << ".";
return false;
}
}
auto batch_norm_inputs = desc.Inputs();
if (batch_norm_inputs.find("MomentumTensor") != batch_norm_inputs.end()) {
if (!desc.Input("MomentumTensor").empty()) {
return false;
}
}
if (desc.Output("Y").size() != 1) {
VLOG(3) << "Invalid output Y's size of batch_norm TRT "
"converter. Expected 1, received "
<< desc.Output("Y").size() << ".";
return false;
}
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3) << "The block desc is nullptr, we can't continue to analyze. "
"Developers need to check whether block_desc is passed in "
"the pass.";
return false;
}
auto x_var_name = desc.Input("X")[0];
auto* x_var_desc = block->FindVarRecursive(x_var_name);
const auto x_shape = x_var_desc->GetShape();
}
if (op_type == "split") {
if (desc.Input("X").size() != 1) {
VLOG(3) << "Invalid input X's size of split TRT converter. "
"Expected 1, received "
<< desc.Input("X").size() << ".";
return false;
}
auto split_inputs = desc.Inputs();
if (split_inputs.find("AxisTensor") != split_inputs.end()) {
if (!desc.Input("AxisTensor").empty()) {
return false;
}
}
if (split_inputs.find("SectionsTensorList") != split_inputs.end()) {
if (!desc.Input("SectionsTensorList").empty()) {
if (!with_dynamic_shape) {
return false;
}
}
}
if (!desc.HasAttr("axis")) {
return false;
}
int axis = PADDLE_GET_CONST(int, desc.GetAttr("axis"));
if (!with_dynamic_shape && axis == 0) {
VLOG(3) << "Invalid split axis. Split on batch is not supported in "
"TensorRT with static shape";
return false;
}
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3) << "The block desc is nullptr, we can't continue to analyze. "
"Developers need to check whether block_desc is passed in "
"the pass.";
return false;
}
auto x_var_name = desc.Input("X")[0];
auto* x_var_desc = block->FindVarRecursive(x_var_name);
const auto x_shape = x_var_desc->GetShape();
size_t output_num = desc.Output("Out").size();
std::vector<int> output_lengths;
int num = 0;
if (desc.HasAttr("num")) {
num = PADDLE_GET_CONST(int, desc.GetAttr("num"));
}
if (desc.HasAttr("sections")) {
output_lengths =
PADDLE_GET_CONST(std::vector<int>, desc.GetAttr("sections"));
}
if (output_lengths.empty() && num == 0) {
VLOG(3) << "sections and num cannot be equal to 0 at the same time";
return false;
}
axis += (axis < 0) ? x_shape.size() : 0;
if (x_shape[axis] == -1) {
VLOG(3) << "The (" << axis << ") dim of input should not be -1";
return false;
}
if (output_lengths.empty()) {
if (num > 0) {
int64_t in_axis_dim = x_shape[axis];
if (in_axis_dim % num != 0) {
VLOG(3) << "Invalid number to split. Tensor split does not result"
" in an equal division of dimensions. Axis dim = "
<< in_axis_dim << " num = " << num << "!= 0";
return false;
}
size_t out_axis_dim = in_axis_dim / num;
for (int i = 0; i < num; ++i) {
output_lengths.push_back(out_axis_dim);
}
}
}
if (output_lengths.size() != output_num) {
VLOG(3) << "The output_length should be equal to the output size.";
return false;
}
}
if (op_type == "scale") {
auto scale_inputs = desc.Inputs();
if (scale_inputs.find("ScaleTensor") != scale_inputs.end()) {
if (!desc.Input("ScaleTensor").empty()) {
return false;
}
}
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3) << "The block desc is nullptr, we can't continue to analyze. "
"Developers need to check whether block_desc is passed in "
"the pass.";
return false;
}
auto x_var_name = desc.Input("X")[0];
auto* x_var_desc = block->FindVarRecursive(x_var_name);
const auto x_shape = x_var_desc->GetShape();
auto dtype = x_var_desc->GetDataType();
if (!with_dynamic_shape) {
// At present, only support float32 or float16 or float64 into trt.
if (!(dtype == framework::proto::VarType::FP32 ||
dtype == framework::proto::VarType::FP64 ||
dtype == framework::proto::VarType::FP16)) {
return false;
}
} else {
// At present, only support float32 or float16 or float64 or int32 or
// int64 into trt.
if (!(dtype == framework::proto::VarType::FP32 ||
dtype == framework::proto::VarType::FP16 ||
dtype == framework::proto::VarType::FP64 ||
dtype == framework::proto::VarType::INT32 ||
dtype == framework::proto::VarType::INT64)) {
return false;
}
}
}
if (op_type == "roll") {
if (!with_dynamic_shape) {
return false;
}
}
if (op_type == "strided_slice") {
if (!desc.HasAttr("axes") || !desc.HasAttr("starts") ||
!desc.HasAttr("ends") || !desc.HasAttr("strides")) {
VLOG(3)
<< "The necessary attributes of the strided_slice operator miss ";
return false;
}
}
if (op_type == "rnn") {
if (!with_dynamic_shape) {
return false;
}
if (desc.HasAttr("mode")) {
std::string mode = PADDLE_GET_CONST(std::string, desc.GetAttr("mode"));
if (mode != "LSTM") return false;
}
if (desc.HasAttr("dropout_prob")) {
float dropout_prob =
PADDLE_GET_CONST(float, desc.GetAttr("dropout_prob"));
if (dropout_prob > 1e-5) return false;
}
// not support following four inputs for rnn in paddle-trt
auto rnn_inputs = desc.Inputs();
if (rnn_inputs.find("SequenceLength") != rnn_inputs.end()) {
if (!desc.Input("SequenceLength").empty()) {
return false;
}
}
}
if (op_type == "fill_constant_batch_size_like") {
if (!with_dynamic_shape) {
return false;
}
if (!desc.HasAttr("input_dim_idx")) {
return false;
}
if (!desc.HasAttr("output_dim_idx")) {
return false;
}
if (!desc.HasAttr("shape")) {
return false;
}
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3) << "The block desc is nullptr, we can't continue to analyze. "
"Developers need to check whether block_desc is passed in "
"the pass.";
return false;
}
auto x_var_name = desc.Input("Input")[0];
auto* x_var_desc = block->FindVarRecursive(x_var_name);
auto dtype = x_var_desc->GetDataType();
// At present, only support float32 into trt.
if (dtype != 5) {
return false;
}
}
if (op_type == "fill_any_like") {
if (!with_dynamic_shape) {
VLOG(3) << "the fill_any_like does not support static shape yet";
return false;
}
int dtype = desc.HasAttr("dtype")
? PADDLE_GET_CONST(int, desc.GetAttr("dtype"))
: -1;
auto* block = desc.Block();
auto* x_var_desc = block->FindVarRecursive(desc.Input("X")[0]);
auto input_type = x_var_desc->GetDataType();
#if IS_TRT_VERSION_GE(8400)
if (dtype == 0 ||
(dtype == -1 && input_type == framework::proto::VarType::BOOL)) {
VLOG(3) << "the fill_any_like supports input of BOOL by trt8.4 above";
return true;
}
#endif
if (dtype != -1 && dtype != 2 && dtype != 3 && dtype != 5 && dtype != 6) {
VLOG(3)
<< "the fill_any_like only supports int32/int64/float32/float64 by "
"trt8.4 below";
return false;
}
if (dtype == -1) {
if (input_type != framework::proto::VarType::INT32 &&
input_type != framework::proto::VarType::INT64 &&
input_type != framework::proto::VarType::FP32 &&
input_type != framework::proto::VarType::FP64) {
VLOG(3) << "the fill_any_like only supports "
"int32/int64/float32/float64 by "
"trt8.4 below";
return false;
}
}
}
if (op_type == "slice") {
if (desc.HasAttr("decrease_axis")) {
std::vector<int> decrease_axis =
PADDLE_GET_CONST(std::vector<int>, desc.GetAttr("decrease_axis"));
if (!with_dynamic_shape) {
if (decrease_axis.end() !=
std::find(decrease_axis.begin(), decrease_axis.end(), 0)) {
return false;
}
}
}
std::vector<int> axes;
if (!desc.HasAttr("axes")) {
VLOG(3) << "The necessary attributes of the slice operator axes "
" are missing.";
return false;
} else {
axes = PADDLE_GET_CONST(std::vector<int>, desc.GetAttr("axes"));
if (!with_dynamic_shape) {
for (size_t i = 0; i < axes.size(); i++) {
if (axes[i] == 0) {
VLOG(3) << "Invalid slice axis. Slice on batch axis is not "
"supported in TensorRT";
return false;
}
}
}
}
// not support following four inputs for slice in paddle-trt
auto slice_inputs = desc.Inputs(); // its size == 5
if (slice_inputs.find("StartsTensor") != slice_inputs.end() &&
!desc.Input("StartsTensor").empty()) {
VLOG(3) << "The Slice has StartsTensor input.";
} else {
if (!desc.HasAttr("starts")) {
VLOG(3) << "The necessary attributes of the slice operator starts or "
"StartsTensor"
" are missing.";
return false;
} else {
std::vector<int> starts =
PADDLE_GET_CONST(std::vector<int>, desc.GetAttr("starts"));
if (axes.size() != starts.size()) {
VLOG(3) << "The shape of attributes of the slice operator axes "
"and starts are not equal.";
return false;
}
}
}
if (slice_inputs.find("EndsTensor") != slice_inputs.end() &&
!desc.Input("EndsTensor").empty()) {
VLOG(3) << "The Slice has EndsTensor input.";
} else {
if (!desc.HasAttr("ends")) {
VLOG(3) << "The necessary attributes of the slice operator ends or "
"EndsTensor"
" are missing.";
return false;
} else {
std::vector<int> ends =
PADDLE_GET_CONST(std::vector<int>, desc.GetAttr("ends"));
if (axes.size() != ends.size()) {
VLOG(3) << "The shape of attributes of the slice operator axes "
"and ends are not equal.";
return false;
}
}
}
if (slice_inputs.find("StartsTensorList") != slice_inputs.end()) {
VLOG(3) << "The Slice has StartsTensorList input.";
}
if (slice_inputs.find("EndsTensorList") != slice_inputs.end()) {
VLOG(3) << "The Slice has EndsTensorList input.";
}
}
if (op_type == "less_than" || op_type == "greater_than" ||
op_type == "logical_or" || op_type == "logical_xor" ||
op_type == "logical_and" || op_type == "less_equal" ||
op_type == "greater_equal") {
#if IS_TRT_VERSION_GE(8400)
// TRT does not support kEQUAL/kGREATER/kLESS work with implicit batch
if (!with_dynamic_shape) {
VLOG(3) << "Ops(" << op_type << ") do not support static shape yet.";
return false;
}
auto* block = desc.Block();
auto* x_var_desc = block->FindVarRecursive(desc.Input("X")[0]);
auto* y_var_desc = block->FindVarRecursive(desc.Input("Y")[0]);
auto x_dtype = x_var_desc->GetDataType();
auto y_dtype = y_var_desc->GetDataType();
if (op_type == "logical_or" || op_type == "logical_xor" ||
op_type == "logical_and") {
if (x_dtype != framework::proto::VarType::BOOL ||
y_dtype != framework::proto::VarType::BOOL) {
VLOG(3) << "the op (" << op_type << ") only support input of BOOL.";
return false;
}
}
if (op_type == "less_than" || op_type == "greater_than" ||
op_type == "less_equal" || op_type == "greater_equal") {
if (x_dtype == framework::proto::VarType::BOOL ||
y_dtype == framework::proto::VarType::BOOL) {
VLOG(3)
<< "ElementWiseOperation::kLESS/ElementWiseOperation::kGREATER "
"do not support boolean datatype.";
return false;
}
}
#else
VLOG(3) << "these are not supported when TensorRT < 8.4";
return false;
#endif
}
if (op_type == "elementwise_add" || op_type == "elementwise_mul" ||
op_type == "elementwise_sub" || op_type == "elementwise_div" ||
op_type == "elementwise_pow" || op_type == "elementwise_min" ||
op_type == "elementwise_max" || op_type == "elementwise_floordiv" ||
op_type == "elementwise_mod") {
if (desc.Input("X").size() != 1) {
VLOG(3) << "The input op's Input(\"X\").size() "
"should equal to 1, but received Input(\"X\").size() = "
<< desc.Input("X").size() << ".";
return false;
}
if (desc.Input("Y").size() != 1) {
VLOG(3) << "The input op's Input(\"Y\").size() "
"should equal to 1, but received Input(\"Y\").size() = "
<< desc.Input("Y").size() << ".";
return false;
}
if (desc.Output("Out").size() != 1) {
VLOG(3) << "The input op's Output(\"Out\").size() "
"should equal to 1, but received Output(\"Out\").size() = "
<< desc.Output("Out").size() << ".";
return false;
}
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3) << "The block desc is nullptr, we can't continue to analyze. "
"Developers need to check whether block_desc is passed in "
"the pass.";
return false;
}
auto* x_var_desc = block->FindVarRecursive(desc.Input("X")[0]);
auto* y_var_desc = block->FindVarRecursive(desc.Input("Y")[0]);
const auto x_shape = x_var_desc->GetShape();
const auto y_shape = y_var_desc->GetShape();
// These operations do not support boolean datatype.
if (op_type == "elementwise_add" || op_type == "elementwise_mul" ||
op_type == "elementwise_sub" || op_type == "elementwise_div" ||
op_type == "elementwise_pow" || op_type == "elementwise_min" ||
op_type == "elementwise_max" || op_type == "elementwise_floordiv" ||
op_type == "elementwise_mod") {
if (x_var_desc->GetDataType() ==
paddle::framework::proto::VarType_Type::VarType_Type_BOOL) {
VLOG(3)
<< "These operations "
"(elementwise_add/mul/sub/div/pow/min/max/floordiv/mod) do "
"not support boolean datatype.";
return false;
}
}
// These operations input do not support int32 datatype.
if (op_type == "elementwise_pow") {
if (x_var_desc->GetDataType() ==
paddle::framework::proto::VarType_Type::VarType_Type_INT32) {
VLOG(3) << "These operations (elementwise_pow) do not support int32 "
"datatype.";
return false;
}
}
// The case when x_shape.size() == 1 is dealt with in common case
if (!with_dynamic_shape && (!y_var_desc->Persistable()) &&
y_shape.size() == 1) {
VLOG(3) << "Static shape in trt not support y is a 1D intermediate "
"tensor in "
"elementwise op.";
return false;
}
if (x_var_desc->Persistable() && !with_dynamic_shape) {
VLOG(3)
<< "Input X is a parameter which is not supported for "
"elementwise in tensorrt's static shape, swap x and y will work";
return false;
}
}
if (op_type == "pow") {
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3) << "The block desc is nullptr, we can't continue to analyze. "
"Developers need to check whether block_desc is passed in "
"the pass.";
return false;
}
auto* x_var_desc = block->FindVarRecursive(desc.Input("X")[0]);
// the same as `elementwise_pow`.
if (x_var_desc->GetDataType() ==
paddle::framework::proto::VarType_Type::VarType_Type_INT32) {
VLOG(3) << "These operations (pow) do not support int32 "
"datatype.";
return false;
}
}
if (op_type == "stack") {
if (!with_dynamic_shape) {
VLOG(3)
<< "static shape mode is not supported for TRT stack.\n"
"You can use the config.SetTRTDynamicShapeInfo(...) interface"
" to set the shape information to run the dynamic shape "
"mode.";
return false;
}
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3) << "The block desc is nullptr, we can't continue to analyze. "
"Developers need to check whether block_desc is passed in "
"the pass.";
return false;
}
auto x_var_name = desc.Input("X")[0];
auto* x_var_desc = block->FindVarRecursive(x_var_name);
const auto x_shape = x_var_desc->GetShape();
int rank = x_shape.size();
int axis = desc.HasAttr("axis")
? PADDLE_GET_CONST(int, desc.GetAttr("axis"))
: -1;
if (axis > rank || axis < -(rank + 1)) {
return false;
}
}
if (op_type == "shape" && !with_dynamic_shape) {
return false;
}
if (op_type == "fused_embedding_eltwise_layernorm") {
if (!with_dynamic_shape) {
VLOG(3) << "fused_embedding_eltwise_layernorm should run on dynamic "
"shape mode.";
return false;
}
if (desc.Input("Ids").size() != desc.Input("Embs").size()) {
return false;
}
}
if (op_type == "fused_bias_dropout_residual_layer_norm") {
if (!with_dynamic_shape) {
VLOG(3) << "fused_bias_dropout_residual_layer_norm should run on "
"dynamic shape mode.";
return false;
}
float dropout_rate =
PADDLE_GET_CONST(float, desc.GetAttr("dropout_rate"));
if (dropout_rate != 0.0f) {
VLOG(4) << "preln_residual_bias trt layer can not work with "
"fused_bias_dropout_residual_layer_norm op in which the "
"dropout_rate != 0, stop convert";
return false;
}
}
if (op_type == "fused_preln_embedding_eltwise_layernorm") {
if (!with_dynamic_shape) {
VLOG(3) << "fused_preln_embedding_eltwise_layernorm should run on "
"dynamic "
"shape mode.";
return false;
}
if (desc.Input("Ids").size() != desc.Input("Embs").size()) {
VLOG(3) << "The id and emb size of fused PrelnEmbEltwiseLayerNormOp "
"should be same ";
return false;
}
if (!desc.HasAttr("enable_int8")) {
VLOG(3) << "PrelnEmbEltwiseLayerNormOp must use int8 mode.";
return false;
}
}
if (op_type == "gelu") {
if (desc.Input("X").size() != 1) {
VLOG(3) << "gelu op has only 1 input, but got "
<< desc.Input("X").size();
return false;
}
if (desc.Output("Out").size() != 1) {
VLOG(3) << "gelu op has only 1 output, but got "
<< desc.Output("Out").size();
return false;
}
}
if (op_type == "layer_norm") {
if (desc.Input("X").size() != 1) {
VLOG(3) << "input of layer_norm op converter should be 1, got "
<< desc.Input("X").size();
return false;
}
if (desc.Input("Bias").size() != 1) {
VLOG(3) << "Bias of layer_norm op converter should be 1, got "
<< desc.Input("Bias").size();
return false;
}
if (desc.Input("Scale").size() != 1) {
VLOG(3) << "Scale of layer_norm op converter should be 1, got "
<< desc.Input("Scale").size();
return false;
}
if (desc.Output("Y").size() != 1) {
VLOG(3) << "output of layer_norm op converter should be 1, got "
<< desc.Output("Y").size();
return false;
}
}
if (op_type == "fill_constant") {
auto fill_constant_inputs = desc.Inputs();
if (fill_constant_inputs.find("ValueTensor") !=
fill_constant_inputs.end()) {
if (!desc.Input("ValueTensor").empty()) return false;
}
if (desc.HasInput("ShapeTensor")) {
if (desc.Input("ShapeTensor").size() > 1) return false;
if (desc.Input("ShapeTensor").size() == 1) {
#if IS_TRT_VERSION_LT(8500)
VLOG(3) << "fill_constant ShapeTensor is not supported when TensorRT "
"< 8.5.0";
return false;
#endif
}
}
#if IS_TRT_VERSION_LT(8500)
if (desc.HasInput("ShapeTensorList")) {
if (desc.Input("ShapeTensorList").size() >= 1) {
VLOG(3) << "fill_constant ShapeTensorList is not supported when "
"TensorRT < 8.5.0";
return false;
}
}
#endif
int dtype = desc.HasAttr("dtype")
? PADDLE_GET_CONST(int, desc.GetAttr("dtype"))
: 5;
// only support int32, int64, float32
if (!(dtype == 2 || dtype == 3 || dtype == 5)) {
return false;
}
}
if (op_type == "instance_norm") {
if (desc.Input("X").size() != 1) {
VLOG(3) << "input of instance_norm op converter should be 1, got "
<< desc.Input("X").size();
return false;
}
if (desc.Input("Bias").size() != 1) {
VLOG(3) << "Bias of instance_norm op converter should be 1, got "
<< desc.Input("Bias").size();
return false;
}
if (desc.Input("Scale").size() != 1) {
VLOG(3) << "Scale of instance_norm op converter should be 1, got "
<< desc.Input("Scale").size();
return false;
}
if (desc.Output("Y").size() != 1) {
VLOG(3) << "output of layer_norm op converter should be 1, got "
<< desc.Output("Y").size();
return false;
}
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3) << "The block desc is nullptr, we can't continue to analyze. "
"Developers need to check whether block_desc is passed in "
"the pass.";
return false;
}
auto x_var_name = desc.Input("X")[0];
auto* x_var_desc = block->FindVarRecursive(x_var_name);
const auto x_shape = x_var_desc->GetShape();
if (x_shape.size() != 4) {
VLOG(3) << "The instance_norm op only support 4-dimensional input in "
"tensorrt.";
return false;
}
}
if (op_type == "pad") {
if (!desc.HasAttr("pad_value") || !desc.HasAttr("paddings")) return false;
const float pad_value =
PADDLE_GET_CONST(float, desc.GetAttr("pad_value"));
if (pad_value != 0.0f) {
VLOG(3) << "The pad layer of TRT only support zero.";
return false;
}
std::vector<int64_t> shape;
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3) << "The block desc is nullptr, we can't continue to analyze. "
"Developers need to check whether block_desc is passed in "
"the pass.";
return false;
}
for (auto& param_name : desc.Inputs()) {
for (auto& var_name : param_name.second) {
auto* var_desc = block->FindVarRecursive(var_name);
shape = var_desc->GetShape();
}
}
int nbDims = shape.size();
std::vector<int> paddings =
PADDLE_GET_CONST(std::vector<int>, desc.GetAttr("paddings"));
int pad_size = paddings.size();
if (nbDims < 2) {
return false;
}
if (nbDims * 2 != pad_size) {
return false;
}
for (int i = 0; i < pad_size - 4; i++) {
if (paddings[i] != 0) {
return false;
}
}
}
if (op_type == "bitwise_and") {
#if IS_TRT_VERSION_LT(8400)
VLOG(3) << "bitwise_and is not supported when TensorRT < 8.4";
return false;
#endif
if (!with_dynamic_shape) {
VLOG(3) << "Ops(" << op_type << ") do not support static shape yet.";
return false;
}
auto x_var_name = desc.Input("X")[0];
auto y_var_name = desc.Input("Y")[0];
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3) << "The block desc is nullptr, we can't continue to analyze. "
"Developers need to check whether block_desc is passed in "
"the pass.";
return false;
}
auto* x_var_desc = block->FindVarRecursive(x_var_name);
auto* y_var_desc = block->FindVarRecursive(y_var_name);
auto x_dtype = x_var_desc->GetDataType();
auto y_dtype = y_var_desc->GetDataType();
if (x_dtype != framework::proto::VarType::BOOL ||
y_dtype != framework::proto::VarType::BOOL) {
VLOG(3) << "the bitwise_and only support input of BOOL.";
return false;
}
}
if (op_type == "bitwise_or") {
#if IS_TRT_VERSION_LT(8400)
VLOG(3) << "bitwise_or is not supported when TensorRT < 8.4";
return false;
#endif
if (!with_dynamic_shape) {
VLOG(3) << "Ops(" << op_type << ") do not support static shape yet.";
return false;
}
auto x_var_name = desc.Input("X")[0];
auto y_var_name = desc.Input("Y")[0];
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3) << "The block desc is nullptr, we can't continue to analyze. "
"Developers need to check whether block_desc is passed in "
"the pass.";
return false;
}
auto* x_var_desc = block->FindVarRecursive(x_var_name);
auto* y_var_desc = block->FindVarRecursive(y_var_name);
auto x_dtype = x_var_desc->GetDataType();
auto y_dtype = y_var_desc->GetDataType();
if (x_dtype != framework::proto::VarType::BOOL ||
y_dtype != framework::proto::VarType::BOOL) {
VLOG(3) << "the bitwise_or only support input of BOOL.";
return false;
}
}
if (op_type == "size") {
if (!with_dynamic_shape) {
VLOG(3) << "Ops(" << op_type << ") do not support static shape yet.";
return false;
}
}
if (op_type == "pad3d") {
#if !IS_TRT_VERSION_GE(8200)
VLOG(3) << "pad3d is not supported when TensorRT < 8.2";
return false;
#endif
if (!with_dynamic_shape) {
VLOG(3) << "pad3d is not supported static shape";
return false;
}
if (!desc.HasAttr("paddings") && !desc.HasInput("Paddings")) {
return false;
}
if (desc.HasAttr("mode")) {
std::string mode = PADDLE_GET_CONST(std::string, desc.GetAttr("mode"));
if (mode != "constant" && mode != "reflect" && mode != "replicate") {
VLOG(3) << "The pad3d layer of TRT only support "
"constant/reflect/replicate mode.";
return false;
}
}
if (desc.HasAttr("data_format")) {
std::string data_format =
PADDLE_GET_CONST(std::string, desc.GetAttr("data_format"));
if (data_format != "NCDHW") {
VLOG(3) << "The pad3d layer of TRT only support NCDHW data format.";
return false;
}
}
}
if (op_type == "prelu") {
if (desc.Input("X").size() != 1) {
VLOG(3) << "Invalid input X's size of prelu TRT converter. "
"Expected 1, received "
<< desc.Input("X").size() << ".";
return false;
}
if (desc.Output("Out").size() != 1) {
VLOG(3) << "Invalid output Out's size of prelu TRT converter. "
"Expected 1, received "
<< desc.Output("Out").size() << ".";
return false;
}
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3) << "The block desc is nullptr, we can't continue to analyze. "
"Developers need to check whether block_desc is passed in "
"the pass.";
return false;
}
auto* alpha_var = block->FindVarRecursive(desc.Input("Alpha")[0]);
if (!alpha_var) {
VLOG(3) << "Variable Alpha of prelu TRT converter not found.";
return false;
}
auto alpha_shape = alpha_var->GetShape();
if (!with_dynamic_shape && alpha_shape.empty()) {
VLOG(3) << op_type
<< " op does not support alpha's dim is 0 in tensorrt "
"static shape mode.";
return false;
}
}
if (op_type == "mish") {
if (desc.Input("X").size() != 1) {
VLOG(3) << "Invalid input X's size of mish TRT converter. "
"Expected 1, received "
<< desc.Input("X").size() << ".";
return false;
}
if (desc.Output("Out").size() != 1) {
VLOG(3) << "Invalid output Out's size of mish TRT converter. "
"Expected 1, received "
<< desc.Output("Out").size() << ".";
return false;
}
}
if (op_type == "roi_align") {
if (!with_dynamic_shape) {
VLOG(3) << "TRT roi align plugin only accept the dynamic shape, "
"because that "
"the roi_align will change the batch size.";
return false;
}
std::vector<std::string> attrs{"pooled_height",
"pooled_width",
"spatial_scale",
"sampling_ratio",
"aligned"};
for (auto const& attr : attrs) {
if (!desc.HasAttr(attr)) return false;
}
const auto pooled_height =
PADDLE_GET_CONST(int, desc.GetAttr("pooled_height"));
if (pooled_height <= 0) return false;
const auto pooled_width =
PADDLE_GET_CONST(int, desc.GetAttr("pooled_width"));
if (pooled_width <= 0) return false;
const auto spatial_scale =
PADDLE_GET_CONST(float, desc.GetAttr("spatial_scale"));
if (spatial_scale <= 0.f) return false;
auto roi_align_inputs = desc.Inputs();
if (roi_align_inputs.find("RoisNum") != roi_align_inputs.end()) {
if (!desc.Input("RoisNum").empty()) {
return false;
}
}
}
if (op_type == "shuffle_channel") {
#if !IS_TRT_VERSION_GE(8000)
if (with_dynamic_shape) {
VLOG(3) << "You are running the TRT Dynamic Shape mode, "
"the shuffle_channel op does not support dynamic shape "
"trt versions below 8.0 yet";
return false;
}
#endif
}
if (op_type == "where") {
#if !IS_TRT_VERSION_GE(8400)
VLOG(3) << "where is not supported when TensorRT < 8.4";
return false;
#endif
if (!with_dynamic_shape) {
VLOG(3) << "the where op does not support static shape yet";
return false;
}
}
if (op_type == "bitwise_not") {
auto* block = desc.Block();
auto x_var_name = desc.Input("X")[0];
auto* x_var_desc = block->FindVarRecursive(x_var_name);
auto dtype = x_var_desc->GetDataType();
if (dtype == framework::proto::VarType::INT8 ||
dtype == framework::proto::VarType::UINT8) {
VLOG(3) << "INT8 / UINT8 type convert to trt is not supported";
return false;
}
if (dtype == framework::proto::VarType::BOOL) {
#if !IS_TRT_VERSION_GE(8400)
VLOG(3) << "BOOL type support requires TensorRT 8.4";
return false;
#elif !IS_TRT_VERSION_GE(8600)
const auto x_shape = x_var_desc->GetShape();
if (x_shape.empty()) {
VLOG(3) << "BOOL type does not support 0 dim input when TensorRT < "
"8.6.";
return false;
}
#endif
}
}
if (op_type == "one_hot" || op_type == "one_hot_v2") {
#if IS_TRT_VERSION_LT(8510)
VLOG(3) << "one_hot/one_hot_v2 is not supported when TensorRT < 8.5.1";
return false;
#endif
if (!with_dynamic_shape) {
VLOG(3)
<< "the one_hot/one_hot_v2 op does not support static shape yet";
return false;
}
if (desc.HasAttr("allow_out_of_range")) {
VLOG(3) << "allow_out_of_range one_hot/one_hot_v2 op is not "
"supported now.";
if (PADDLE_GET_CONST(bool, desc.GetAttr("allow_out_of_range")))
return false;
}
if (desc.HasAttr("dtype")) {
const int dtype = PADDLE_GET_CONST(int, desc.GetAttr("dtype"));
if (dtype != 2 && dtype != 3 && dtype != 5) {
VLOG(3) << "one_hot/one_hot_v2 op only support int32, int64, float.";
return false;
}
}
auto one_hot_inputs = desc.Inputs();
if (one_hot_inputs.find("depth_tensor") != one_hot_inputs.end()) {
if (!desc.Input("depth_tensor").empty()) {
return true;
}
}
if (desc.HasAttr("depth")) {
const int depth = PADDLE_GET_CONST(int, desc.GetAttr("depth"));
if (depth <= 0) {
VLOG(3) << "depth only support positive in one_hot/one_hot_v2 op.";
return false;
}
}
}
if (op_type == "skip_layernorm") {
if (!with_dynamic_shape) {
VLOG(3) << "the skip_layernorm does not support static shape yet";
return false;
}
}
if (op_type == "preln_skip_layernorm") {
if (!with_dynamic_shape) {
VLOG(3) << "the preln_skip_layernorm does not support static shape yet";
return false;
}
if (!desc.HasAttr("enable_int8")) {
VLOG(3) << "PrelnEmbEltwiseLayerNormOp must use int8 mode.";
return false;
}
}
if (op_type == "multihead_matmul") {
if (!with_dynamic_shape) {
VLOG(3) << "the multihead_matmul does not support static shape yet";
return false;
}
if (desc.HasAttr("enable_int8") && !desc.HasAttr("Input_scale")) {
VLOG(3) << "Multihead layers must have input scale in int8 mode.";
return false;
}
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3) << "The block desc is nullptr, we can't continue to analyze. "
"Developers need to check whether block_desc is passed in "
"the pass.";
return false;
}
auto* input_desc = block->FindVarRecursive(desc.Input("Input").front());
const auto input_shape = input_desc->GetShape();
const auto head_number =
PADDLE_GET_CONST(int, desc.GetAttr("head_number"));
auto inputs = desc.Inputs();
bool has_bias_qk = (inputs.find("BiasQK") == inputs.end()) ? false : true;
if (has_bias_qk) {
auto* bias_qk_desc =
block->FindVarRecursive(desc.Input("BiasQK").front());
const auto bias_qk_shape = bias_qk_desc->GetShape();
// The BiasQK's shape requires to be
// [batch, 1, 1, length] or [batch, head, length, length].
bool has_same_shape = head_number == bias_qk_shape[1] &&
input_shape[1] == bias_qk_shape[2] &&
input_shape[1] == bias_qk_shape[3];
bool is_broadcastable = bias_qk_shape[1] == 1 &&
bias_qk_shape[2] == 1 &&
input_shape[1] == bias_qk_shape[3];
is_broadcastable = is_broadcastable ||
(bias_qk_shape[0] == 1 && bias_qk_shape[1] == 1 &&
input_shape[1] == bias_qk_shape[2] &&
input_shape[1] == bias_qk_shape[3]);
if (!(has_same_shape || is_broadcastable)) {
VLOG(3) << "The BiasQK's shape is invalid, expect [" << input_shape[0]
<< ", 1, 1, " << input_shape[1] << "] "
<< "or [" << input_shape[0] << ", " << head_number << ", "
<< input_shape[1] << ", " << input_shape[1] << "] "
<< "or [" << input_shape[0] << "/1, " << 1 << ", "
<< input_shape[1] << ", " << input_shape[1] << "] "
<< "but got [" << bias_qk_shape[0] << ", " << bias_qk_shape[1]
<< ", " << bias_qk_shape[2] << ", " << bias_qk_shape[3]
<< "].";
return false;
}
} else {
#if (IS_TRT_VERSION_GE(8000) && IS_TRT_VERSION_LT(8100))
VLOG(3) << "There are some bugs with trt 8.0";
return false;
#endif
}
}
if (op_type == "multihead_matmul_roformer") {
if (!with_dynamic_shape) {
VLOG(3) << "the multihead_matmul_roformer does not support static "
"shape yet";
return false;
}
if (desc.HasAttr("enable_int8") && !desc.HasAttr("Input_scale")) {
VLOG(3) << "Multihead layers must have input scale in int8 mode.";
return false;
}
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3) << "The block desc is nullptr, we can't continue to analyze. "
"Developers need to check whether block_desc is passed in "
"the pass.";
return false;
}
auto* input_desc = block->FindVarRecursive(desc.Input("Input").front());
const auto input_shape = input_desc->GetShape();
const auto head_number =
PADDLE_GET_CONST(int, desc.GetAttr("head_number"));
auto inputs = desc.Inputs();
bool has_bias_qk = (inputs.find("BiasQK") == inputs.end()) ? false : true;
if (has_bias_qk) {
auto* bias_qk_desc =
block->FindVarRecursive(desc.Input("BiasQK").front());
const auto bias_qk_shape = bias_qk_desc->GetShape();
// The BiasQK's shape requires to be
// [batch, 1, 1, length] or [batch, head, length, length].
bool has_same_shape = head_number == bias_qk_shape[1] &&
input_shape[1] == bias_qk_shape[2] &&
input_shape[1] == bias_qk_shape[3];
bool is_broadcastable = bias_qk_shape[1] == 1 &&
bias_qk_shape[2] == 1 &&
input_shape[1] == bias_qk_shape[3];
if (!(has_same_shape || is_broadcastable)) {
VLOG(3) << "The BiasQK's shape is invalid, expect [" << input_shape[0]
<< ", 1, 1, " << input_shape[1] << "] or [" << input_shape[0]
<< ", " << head_number << ", " << input_shape[1] << ", "
<< input_shape[1] << "] but [" << bias_qk_shape[0] << ", "
<< bias_qk_shape[1] << ", " << bias_qk_shape[2] << ", "
<< bias_qk_shape[3] << "].";
return false;
}
} else {
#if !IS_TRT_VERSION_GE(8000)
VLOG(3) << "The version of TRT must be greater than 8000";
return false;
#endif
}
}
if (op_type == "reshape" || op_type == "reshape2") {
if (!desc.HasAttr("shape")) {
return false;
}
if (with_dynamic_shape) {
return true;
}
// Static shape does not support the input tensors: Shape and
// ShapeTensor
auto reshape_inputs = desc.Inputs();
if (reshape_inputs.find("Shape") != reshape_inputs.end()) {
if (!desc.Input("Shape").empty()) {
return false;
}
}
if (reshape_inputs.find("ShapeTensor") != reshape_inputs.end()) {
if (!desc.Input("ShapeTensor").empty()) {
return false;
}
}
std::vector<int> shape =
PADDLE_GET_CONST(std::vector<int>, desc.GetAttr("shape"));
if (shape.size() >= nvinfer1::Dims::MAX_DIMS) return false;
if (!with_dynamic_shape) {
if (shape.size() == 1) {
return false;
}
if (shape[0] == 0) {
return true;
} else {
auto* block = desc.Block();
auto x_var_name = desc.Input("X")[0];
auto* x_var_desc = block->FindVarRecursive(x_var_name);
const auto x_shape = x_var_desc->GetShape();
int input_num = std::accumulate(
x_shape.begin() + 1, x_shape.end(), 1, std::multiplies<int>());
int shape_num = std::accumulate(
shape.begin() + 1, shape.end(), 1, std::multiplies<int>());
if (input_num == shape_num) {
return true;
}
}
return false;
}
}
if (op_type == "clip") {
if (!with_dynamic_shape) {
VLOG(3) << "the clip does not support static "
"shape yet";
return false;
}
// Paddle-TRT does not support the input tensors: Min and Max
auto clip_inputs = desc.Inputs();
if (clip_inputs.find("Min") != clip_inputs.end()) {
if (!desc.Input("Min").empty()) {
return false;
}
}
if (clip_inputs.find("Max") != clip_inputs.end()) {
if (!desc.Input("Max").empty()) {
return false;
}
}
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3) << "The block desc is nullptr, we can't continue to analyze. "
"Developers need to check whether block_desc is passed in "
"the pass.";
return false;
}
auto x_var_name = desc.Input("X")[0];
auto* x_var_desc = block->FindVarRecursive(x_var_name);
const auto x_shape = x_var_desc->GetShape();
if (x_shape.empty()) {
VLOG(3) << op_type
<< " op does not support input's dim is 0 in tensorrt.";
return false;
}
}
if (op_type == "reduce_sum" || op_type == "reduce_mean" ||
op_type == "reduce_max" || op_type == "reduce_min" ||
op_type == "reduce_prod" || op_type == "reduce_any" ||
op_type == "reduce_all") {
if (!desc.HasAttr("dim", /*with_attr_var=*/false)) {
VLOG(3) << "Skip to convert into TRT while found Attribute('dim') is "
"Variable type in "
<< desc.Type();
return false;
}
if (!(desc.HasAttr("keep_dim") && desc.HasAttr("dim") &&
desc.HasAttr("reduce_all"))) {
VLOG(3) << "the " << op_type
<< " does not have attr (keep_dim or dim or "
"reduce_all)";
return false;
}
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3) << "The block desc is nullptr, we can't continue to analyze. "
"Developers need to check whether block_desc is passed in "
"the pass.";
return false;
}
// The batch size dimension cannot be reduced if it's not dynamic shape.
auto* x_var_desc = block->FindVarRecursive(desc.Input("X")[0]);
if (!with_dynamic_shape) {
if (PADDLE_GET_CONST(bool, desc.GetAttr("reduce_all"))) return false;
std::vector<int32_t> dim =
PADDLE_GET_CONST(std::vector<int32_t>, desc.GetAttr("dim"));
const auto input_shape = x_var_desc->GetShape();
for (auto x : dim) {
if (x == 0 || (x + input_shape.size() == 0)) return false;
}
}
auto dtype = x_var_desc->GetDataType();
if (op_type == "reduce_all" || op_type == "reduce_any") {
if (dtype != framework::proto::VarType::BOOL) {
VLOG(3)
<< "reduce_all and reduce_any op input data type must be bool";
return false;
}
} else {
if (dtype != framework::proto::VarType::INT32 &&
dtype != framework::proto::VarType::INT64 &&
dtype != framework::proto::VarType::FP32 &&
dtype != framework::proto::VarType::FP64) {
VLOG(3) << "reduce op input data type must be int32 or int64 or "
"float32 or "
"float64";
return false;
}
}
}
if (op_type == "tile") {
// Paddle-TRT does not support the input tensors.
auto tile_inputs = desc.Inputs();
if (!with_dynamic_shape) {
if (tile_inputs.find("repeat_times_tensor") != tile_inputs.end()) {
if (!desc.Input("repeat_times_tensor").empty()) {
VLOG(3) << "Tile op: repeat_times_tensor is not empty.";
return false;
}
}
if (tile_inputs.find("RepeatTimes") != tile_inputs.end()) {
if (!desc.Input("RepeatTimes").empty()) {
VLOG(3) << "Tile op: RepeatTimes is not empty.";
return false;
}
}
if (!desc.HasAttr("repeat_times")) {
VLOG(3) << "Tile op:`repeat_times` is not set.";
return false;
}
}
}
// conv3d_transpose
if (op_type == "conv3d_transpose") {
// trt doesn't support output_padding when < 8406
// output_padding is usually set when stride > 1
#if !IS_TRT_VERSION_GE(8400)
if (desc.HasAttr("output_padding")) {
const std::vector<int> output_padding =
PADDLE_GET_CONST(std::vector<int>, desc.GetAttr("output_padding"));
if (output_padding.size() > 0) {
int max_padding =
*std::max_element(output_padding.begin(), output_padding.end());
if (max_padding > 0) return false;
}
}
#endif
}
if (op_type == "conv3d" || op_type == "conv3d_transpose") {
if (desc.HasAttr("padding_algorithm")) {
std::string padding_algorithm =
PADDLE_GET_CONST(std::string, desc.GetAttr("padding_algorithm"));
// trt error is raised if conv3d_transpose and SAME
if (op_type == "conv3d_transpose" && padding_algorithm == "SAME" &&
!with_dynamic_shape) {
return false;
}
}
std::vector<int> paddings =
PADDLE_GET_CONST(std::vector<int>, desc.GetAttr("paddings"));
// conv3d and conv3d_transpose need padding check
if (paddings.size() > 3) return false;
if (desc.Input("Input").size() != 1) {
VLOG(3) << "TRT Conv3d expect 1 input, but got "
<< desc.Input("Input").size() << " input.";
return false;
}
if (desc.Input("Filter").size() != 1) {
VLOG(3) << "TRT Conv3d expect 1 filter, but got "
<< desc.Input("Filter").size() << " filter.";
return false;
}
if (op_type == "conv3d_transpose") {
if (!desc.HasAttr("dilations")) {
return false;
} else {
const std::vector<int> dilations =
PADDLE_GET_CONST(std::vector<int>, desc.GetAttr("dilations"));
if (dilations[0] != 1 || dilations[1] != 1 || dilations[2] != 1) {
VLOG(3) << "In conv3d_transpose, Dilations must be (1, 1, 1) for "
"tensorRT, but given ("
<< dilations[0] << ", " << dilations[1] << ", "
<< dilations[2] << ")";
return false;
}
}
}
if (desc.Output("Output").size() != 1) {
VLOG(3) << "TRT Conv3d expect 1 output, but got "
<< desc.Output("Output").size() << " output.";
return false;
}
}
if (op_type == "cast") {
if (!(desc.HasAttr("in_dtype") && desc.HasAttr("out_dtype"))) {
VLOG(3) << "the " << op_type
<< " does not have attr (in_dtype or "
"out_dtype)";
return false;
}
int in_dtype = PADDLE_GET_CONST(int, desc.GetAttr("in_dtype"));
int out_dtype = PADDLE_GET_CONST(int, desc.GetAttr("out_dtype"));
if (in_dtype == 0 || out_dtype == 0) {
#if IS_TRT_VERSION_GE(8400)
if (with_dynamic_shape) {
VLOG(3) << "the cast op supports inputs and outputs of BOOL by "
"trt8.4 above ";
return true;
}
#endif
return false;
}
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3) << "The block desc is nullptr, we can't continue to analyze. "
"Developers need to check whether block_desc is passed in "
"the pass.";
return false;
}
auto x_var_name = desc.Input("X")[0];
auto* x_var_desc = block->FindVarRecursive(x_var_name);
const auto x_shape = x_var_desc->GetShape();
if (!with_dynamic_shape && (x_shape.size() == 1 || x_shape.empty())) {
VLOG(3) << op_type
<< " op does not support input's dim is 1 or 0 in tensorrt "
"static shape mode.";
return false;
}
}
if (op_type == "set_value") {
#if !IS_TRT_VERSION_GE(8200)
return false;
#endif
auto inputs = desc.Inputs();
if (inputs.find("StartsTensorList") != inputs.end()) {
if (!desc.Input("StartsTensorList").empty()) {
return false;
}
}
if (inputs.find("EndsTensorList") != inputs.end()) {
if (!desc.Input("EndsTensorList").empty()) {
return false;
}
}
if (inputs.find("StepsTensorList") != inputs.end()) {
if (!desc.Input("StepsTensorList").empty()) {
return false;
}
}
if (!(desc.HasAttr("axes") && desc.HasAttr("starts") &&
desc.HasAttr("steps"))) {
VLOG(3) << "the " << op_type
<< " does not have attr (axes or "
"starts or steps)";
return false;
}
if (desc.HasAttr("axes")) {
auto axes =
PADDLE_GET_CONST(std::vector<int64_t>, desc.GetAttr("axes"));
if (axes.size() != 1UL) {
VLOG(3) << "the set_value op"
<< "has more than one element in attribute axes, it can not "
"enter into trt.";
return false;
}
}
}
if (op_type == "top_k_v2" || op_type == "top_k") {
if (desc.HasAttr("axis")) {
int axis = PADDLE_GET_CONST(int, desc.GetAttr("axis"));
if (!with_dynamic_shape && axis == 0) {
VLOG(3) << "top_k_v2 does not support axis == 0 in "
"tensorrt static shape.";
return false;
}
}
if (desc.HasAttr("sorted")) {
bool sorted = PADDLE_GET_CONST(bool, desc.GetAttr("sorted"));
if (!sorted) {
VLOG(3) << op_type
<< " does not support results not sorted in "
"tensorrt";
return false;
}
}
}
#if IS_TRT_VERSION_GE(8000)
if (op_type == "sparse_fc" || op_type == "sparse_multihead_matmul") {
if (!with_dynamic_shape) {
VLOG(3) << "the sparse_fc and sparse_multihead_matmul does not support "
"static shape yet";
return false;
}
}
#endif
if (op_type == "equal" || op_type == "not_equal") {
#if !IS_TRT_VERSION_GE(8000)
VLOG(3) << "equal is not supported when TensorRT < 8.0";
return false;
#else
// TRT does not support kEQUAL/kGREATER/kLESS work with implicit batch
if (!with_dynamic_shape) {
VLOG(3) << "the equal does not support "
"static shape yet";
return false;
}
if (!desc.HasAttr("axis")) {
return false;
}
int axis = PADDLE_GET_CONST(int, desc.GetAttr("axis"));
if (axis == 0) {
return false;
}
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3) << "The block desc is nullptr, we can't continue to analyze. "
"Developers need to check whether block_desc is passed in "
"the pass.";
return false;
}
#endif
}
if (op_type == "layernorm_shift_partition") {
if (!with_dynamic_shape) {
VLOG(3) << "the layernorm_shift_partition does not support "
"static shape yet";
return false;
}
}
if (op_type == "preln_layernorm_shift_partition") {
if (!with_dynamic_shape) {
VLOG(3) << "the layernorm_shift_partition does not support "
"static shape yet";
return false;
}
}
if (op_type == "merge_layernorm") {
if (!with_dynamic_shape) {
VLOG(3) << "The merge_layernorm op does not support "
"static shape yet";
return false;
}
}
if (op_type == "reverse_roll") {
if (!with_dynamic_shape) {
VLOG(3) << "The reverse roll fused op does not support static shape "
"mode yet.";
return false;
}
}
if (op_type == "skip_merge_layernorm") {
if (!with_dynamic_shape) {
VLOG(3) << "The merge_layernorm op does not support "
"static shape yet";
return false;
}
}
if (op_type == "skip_groupnorm_act") {
if (!with_dynamic_shape) {
VLOG(3) << "The skip_groupnorm_act op does not support "
"static shape yet";
return false;
}
}
if (op_type == "preln_groupnorm_act") {
if (!with_dynamic_shape) {
VLOG(3) << "The preln_groupnorm_act op does not support "
"static shape yet";
return false;
}
}
if (op_type == "trans_layernorm") {
if (!with_dynamic_shape) {
VLOG(3) << "The trans_layernorm op does not support "
"static shape yet";
return false;
}
}
if (op_type == "fuse_eleadd_transpose") {
if (!with_dynamic_shape) {
VLOG(3) << "The fuse_eleadd_transpose op does not support "
"static shape yet";
return false;
}
}
if (op_type == "lookup_table" || op_type == "lookup_table_v2") {
if (!with_dynamic_shape) {
VLOG(3) << "the lookup_table does not support "
"static shape yet";
return false;
}
}
if (op_type == "expand_as_v2" || op_type == "expand_v2") {
if (!with_dynamic_shape) {
VLOG(3) << "the " << op_type
<< "does not support "
"static shape yet";
return false;
}
auto inputs = desc.Inputs();
if (op_type == "expand_as_v2") {
if (!desc.HasAttr("target_shape") && inputs.find("Y") == inputs.end()) {
VLOG(3)
<< "expand_as_v2 op need have input(Y) or attr(target_shape). ";
return false;
}
} else if (op_type == "expand_v2") {
if (!desc.HasAttr("shape") && inputs.find("Shape") == inputs.end() &&
inputs.find("expand_shapes_tensor") == inputs.end()) {
VLOG(3) << "expand_v2 op need have input(Shape) or "
"input(expand_shapes_tensor) or attr(shape) . ";
return false;
}
}
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3) << "The block desc is nullptr, we can't continue to analyze. "
"Developers need to check whether block_desc is passed in "
"the pass.";
return false;
}
}
if (op_type == "grid_sampler") {
#if !IS_TRT_VERSION_GE(8510)
VLOG(3) << "grid_sampler is not supported when TensorRT < 8.5.1";
return false;
#else
if (!with_dynamic_shape) {
VLOG(3) << "the grid_sampler does not support "
"static shape yet";
return false;
}
if (!desc.HasAttr("mode") || !desc.HasAttr("padding_mode") ||
!desc.HasAttr("align_corners")) {
VLOG(3) << "grid_sampler need attributes : mode, padding_mode, "
"align_corners";
return false;
}
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3) << "The block desc is nullptr, we can't continue to analyze. "
"Developers need to check whether block_desc is passed in "
"the pass.";
return false;
}
auto input_name = desc.Input("X")[0];
auto* input_desc = block->FindVarRecursive(input_name);
const auto input_shape = input_desc->GetShape();
auto grid_name = desc.Input("Grid")[0];
auto* grid_desc = block->FindVarRecursive(grid_name);
const auto grid_shape = grid_desc->GetShape();
if (input_shape.size() != 4 || grid_shape.size() != 4) {
VLOG(3) << "The input and grid tensors must be shape tensors of rank 4 "
"using TRT GridSample layer.";
return false;
}
#endif
}
if (op_type == "cumsum") {
if (!with_dynamic_shape) {
VLOG(3) << "the cumsum does not support "
"static shape yet";
return false;
}
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3) << "The block desc is nullptr, we can't continue to analyze. "
"Developers need to check whether block_desc is passed in "
"the pass.";
return false;
}
}
if (op_type == "argsort") {
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3) << "The block desc is nullptr, we can't continue to analyze. "
"Developers need to check whether block_desc is passed in "
"the pass.";
return false;
}
if (!desc.HasAttr("descending") || !desc.HasAttr("axis")) {
VLOG(3) << op_type << " needs attributes: descending and axis.";
}
auto x_var_name = desc.Input("X")[0];
auto* x_var_desc = block->FindVarRecursive(x_var_name);
std::vector<int64_t> shape = x_var_desc->GetShape();
int axis = PADDLE_GET_CONST(int, desc.GetAttr("axis"));
if (axis < 0) {
axis += shape.size();
}
if (shape[axis] > 3840 || shape[axis] < 0) {
VLOG(3) << op_type << " shape[" << axis << "] = " << shape[axis]
<< " is invalid, it should less than 3840 and greater than "
"zero in TensorRT.";
return false;
}
}
if (op_type == "unbind") {
if (!with_dynamic_shape) {
VLOG(3) << "the unbind does not support "
"static shape yet";
return false;
}
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3) << "The block desc is nullptr, we can't continue to analyze. "
"Developers need to check whether block_desc is passed in "
"the pass.";
return false;
}
}
if (op_type == "isnan_v2") {
if (!with_dynamic_shape) {
VLOG(3) << "the isnan_v2 does not support "
"static shape yet";
return false;
}
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3) << "The block desc is nullptr, we can't continue to analyze. "
"Developers need to check whether block_desc is passed in "
"the pass.";
return false;
}
}
if (op_type == "p_norm") {
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3) << "The block desc is nullptr, we can't continue to analyze. "
"Developers need to check whether block_desc is passed in "
"the pass.";
return false;
}
if (!(desc.HasAttr("asvector") && desc.HasAttr("axis") &&
desc.HasAttr("porder") && desc.HasAttr("keepdim"))) {
VLOG(3) << op_type << " op need attrs asvector, porder, axis, keepdim.";
return false;
}
bool asvector = PADDLE_GET_CONST(bool, desc.GetAttr("asvector"));
int axis = PADDLE_GET_CONST(int, desc.GetAttr("axis"));
double porder = PADDLE_GET_CONST(double, desc.GetAttr("porder"));
if (asvector || porder != 2.0 || axis != -1) {
VLOG(3) << op_type
<< " op only support asvector=False, porder=2, axis = -1.";
return false;
}
}
if (op_type == "index_put") {
#if IS_TRT_VERSION_LT(8510)
VLOG(3) << "index_put is not supported when TensorRT < 8.5.1";
return false;
#endif
if (!with_dynamic_shape) {
VLOG(3) << "the index_put does not support "
"static shape yet";
return false;
}
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3) << "The block desc is nullptr, we can't continue to analyze. "
"Developers need to check whether block_desc is passed in "
"the pass.";
return false;
}
auto value_var_name = desc.Input("value")[0];
auto* value_var_desc = block->FindVarRecursive(value_var_name);
const auto value_shape = value_var_desc->GetShape();
int value_num = std::accumulate(
value_shape.begin(), value_shape.end(), 1, std::multiplies<int>());
if (value_num != 1) {
VLOG(3) << op_type << " op only support value_num = 1 in tensorrt.";
return false;
}
auto indices_var_name = desc.Input("indices")[0];
auto* indices_var_desc = block->FindVarRecursive(indices_var_name);
auto dtype = indices_var_desc->GetDataType();
if (dtype != framework::proto::VarType::BOOL) {
VLOG(3) << op_type << " op only support bool indices in tensorrt.";
return false;
}
}
if (op_type == "temporal_shift") {
#if !IS_TRT_VERSION_GE(8200)
VLOG(3) << "temporal_shift is not supported when TensorRT < 8.2";
return false;
#endif
if (!with_dynamic_shape) {
VLOG(3) << "the temporal shift does not support "
"static shape yet";
return false;
}
if (!desc.HasAttr("shift_ratio") || !desc.HasAttr("seg_num")) {
VLOG(3) << "temporal shift need attributes : shift_ratio and seg_num";
return false;
}
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3) << "The block desc is nullptr, we can't continue to analyze. "
"Developers need to check whether block_desc is passed in "
"the pass.";
return false;
}
auto input_name = desc.Input("X")[0];
auto* input_desc = block->FindVarRecursive(input_name);
const auto input_shape = input_desc->GetShape();
if (input_shape.size() != 4) {
VLOG(3) << "The input and grid tensors must be shape tensors of rank 4 "
"using TRT TemporalShift layer.";
return false;
}
}
if (op_type == "einsum") {
#if !IS_TRT_VERSION_GE(8200)
VLOG(3) << "einsum is not supported when TensorRT < 8.2";
return false;
#else
if (!with_dynamic_shape) {
VLOG(3) << "the einsum does not support "
"static shape yet";
return false;
}
auto operand_inputs = desc.Input("Operands");
if (operand_inputs.size() > 2) {
VLOG(3) << "TensorRT currently supports up to 2 input tensors"
<< "to einsum but operation had" << operand_inputs.size()
<< "input tensors !";
return false;
}
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3) << "The block desc is nullptr, we can't continue to analyze. "
"Developers need to check whether block_desc is passed in "
"the pass.";
return false;
}
auto equation = PADDLE_GET_CONST(std::string, desc.GetAttr("equation"));
if (equation.find("...") != std::string::npos) {
VLOG(3) << "TensorRT currently does not support ellipses !";
return false;
}
#endif
}
if (op_type == "quantize_linear" || op_type == "dequantize_linear") {
#if !IS_TRT_VERSION_GE(8000)
VLOG(3) << "quantize / dequantize linear is not supported when TensorRT "
"< 8.0";
return false;
#else
return true;
#endif
}
if (op_type == "flip") {
if (!with_dynamic_shape) {
VLOG(3) << "the flip does not support "
"static shape yet";
return false;
}
}
if (use_no_calib_int8) {
return int8_teller_set.count(op_type);
} else {
return teller_set.count(op_type);
}
}
private:
// use this set for no calib int8.
std::unordered_set<std::string> int8_teller_set{
"matrix_multiply",
"bmm",
"range",
"conv2d",
"fused_conv2d_add_act",
"pool2d",
"relu",
"elu",
"selu",
"softsign",
"softplus",
"stanh",
"thresholded_relu",
"exp",
"log",
"sqrt",
"abs",
"sin",
"cos",
"tan",
"sinh",
"cosh",
"asin",
"acos",
"atan",
"asinh",
"acosh",
"atanh",
"ceil",
"floor",
"rsqrt",
"sign",
"reciprocal",
"logical_not",
"erf",
"square",
"softmax",
"sigmoid",
"hard_swish",
"depthwise_conv2d",
"batch_norm",
"concat",
"tanh",
"pad3d",
"pad",
"elementwise_add",
"elementwise_sub",
"elementwise_mul",
"elementwise_div",
"elementwise_pow",
"elementwise_min",
"elementwise_max",
"elementwise_floordiv",
"elementwise_mod",
"equal",
"not_equal",
"less_than",
"greater_than",
"logical_or",
"logical_xor",
"logical_and",
"less_equal",
"greater_equal",
"dropout",
"fill_any_like",
"prelu",
"conv2d_transpose",
"depthwise_conv2d_transpose",
"leaky_relu",
"shuffle_channel",
"where",
"bitwise_not",
"one_hot",
"one_hot_v2",
"swish",
"silu",
"celu",
"split",
"instance_norm",
"gelu",
"layer_norm",
"scale",
"stack",
"transpose2",
"transpose",
"top_k",
"top_k_v2",
"flatten2",
"flatten",
"gather",
"gather_nd",
"group_norm",
"yolo_box",
"yolo_box_head",
"arg_max",
"arg_min",
"roi_align",
"affine_channel",
"nearest_interp",
"anchor_generator",
"reduce_max",
"reduce_min",
"reduce_mean",
"reduce_sum",
"reduce_prod",
"reduce_any",
"reduce_all",
"conv3d",
"conv3d_transpose",
"mish",
"nearest_interp_v2",
"bilinear_interp_v2",
"linear_interp_v2",
"pool3d",
"deformable_conv",
"relu6",
"hard_sigmoid",
"clip",
"prompt_tuning_emb_eltwise_layernorm",
"fused_embedding_eltwise_layernorm",
"multihead_matmul",
"multihead_matmul_roformer",
"skip_layernorm",
"slice",
"strided_slice",
"fused_preln_embedding_eltwise_layernorm",
"fused_bias_dropout_residual_layer_norm",
"c_allreduce_sum",
"roll",
"cast",
"preln_skip_layernorm",
"transformer_input_convert",
"recover_padding",
"remove_padding",
"fill_constant",
"sum",
"shape",
"squeeze2",
"unsqueeze2",
"index_put",
"layernorm_shift_partition",
"reverse_roll",
"take_along_axis",
"tanh_shrink",
"logsigmoid",
"preln_layernorm_shift_partition",
"lookup_table",
"lookup_table_v2",
"trans_layernorm",
"merge_layernorm",
"skip_merge_layernorm",
"expand_v2",
"expand_as_v2",
"fuse_eleadd_transpose",
"skip_groupnorm_act",
"preln_groupnorm_act",
"temporal_shift",
"grid_sampler",
"cumsum",
"unbind",
"isnan_v2",
"p_norm",
"assign",
"flip",
"quantize_linear",
"dequantize_linear",
"share_data",
"argsort",
"bitwise_and",
"bitwise_or",
"size"};
std::unordered_set<std::string> teller_set{
"matrix_multiply",
"bmm",
"range",
"conv2d",
"fused_conv2d_add_act",
"pool2d",
"relu",
"elu",
"selu",
"softsign",
"softplus",
"stanh",
"thresholded_relu",
"exp",
"log",
"sqrt",
"abs",
"sin",
"cos",
"tan",
"sinh",
"cosh",
"asin",
"acos",
"atan",
"asinh",
"acosh",
"atanh",
"ceil",
"floor",
"rsqrt",
"sign",
"reciprocal",
"logical_not",
"erf",
"square",
"softmax",
"sigmoid",
"hard_swish",
"depthwise_conv2d",
"batch_norm",
"concat",
"tanh",
"pad3d",
"pad",
"elementwise_add",
"elementwise_sub",
"elementwise_mul",
"elementwise_div",
"elementwise_pow",
"pow",
"elementwise_min",
"elementwise_max",
"elementwise_floordiv",
"elementwise_mod",
"equal",
"not_equal",
"less_than",
"greater_than",
"logical_or",
"logical_xor",
"logical_and",
"less_equal",
"greater_equal",
"dropout",
"fill_any_like",
"prelu",
"conv2d_transpose",
"depthwise_conv2d_transpose",
"leaky_relu",
"shuffle_channel",
"where",
"bitwise_not",
"one_hot",
"one_hot_v2",
"swish",
"silu",
"celu",
"split",
"instance_norm",
"gelu",
"layer_norm",
"scale",
"stack",
"transpose2",
"transpose",
"top_k",
"top_k_v2",
"flatten2",
"flatten",
"gather",
"gather_nd",
"yolo_box",
"yolo_box_head",
"arg_max",
"arg_min",
"roi_align",
"affine_channel",
"nearest_interp",
"anchor_generator",
"reduce_max",
"reduce_min",
"reduce_mean",
"reduce_sum",
"reduce_prod",
"reduce_any",
"reduce_all",
"conv3d",
"conv3d_transpose",
"mish",
"bilinear_interp_v2",
"linear_interp_v2",
"nearest_interp_v2",
"pool3d",
"deformable_conv",
"relu6",
"hard_sigmoid",
"clip",
"prompt_tuning_emb_eltwise_layernorm",
"fused_embedding_eltwise_layernorm",
"multihead_matmul",
"multihead_matmul_roformer",
"skip_layernorm",
"slice",
"strided_slice",
"fused_preln_embedding_eltwise_layernorm",
"preln_skip_layernorm",
"fused_bias_dropout_residual_layer_norm",
"c_allreduce_sum",
"roll",
"cast",
"transformer_input_convert",
"recover_padding",
"remove_padding",
"fill_constant",
"sum",
"shape",
"squeeze2",
"unsqueeze2",
"fused_token_prune",
"layernorm_shift_partition",
"reverse_roll",
"tanh_shrink",
"index_put",
"take_along_axis",
"logsigmoid",
"preln_layernorm_shift_partition",
"trans_layernorm",
"merge_layernorm",
"skip_merge_layernorm",
"lookup_table",
"lookup_table_v2",
"expand_v2",
"expand_as_v2",
"fuse_eleadd_transpose",
"skip_groupnorm_act",
"preln_groupnorm_act",
"temporal_shift",
"grid_sampler",
"cumsum",
"unbind",
"isnan_v2",
"p_norm",
"assign",
"flip",
"quantize_linear",
"dequantize_linear",
"share_data",
"argsort",
"bitwise_and",
"bitwise_or",
"size"};
};
struct GenericPluginTeller : public Teller {
public:
GenericPluginTeller() = default;
bool operator()(const framework::OpDesc& desc,
bool use_no_calib_int8 = false,
bool with_dynamic_shape = false,
bool forbid_dynamic_op_enter_into_trt = false,
bool use_explicit_quantization = false) override {
const std::string op_type = desc.Type();
// only consider dynamic_shape mode
if (!with_dynamic_shape) {
return false;
}
if (op_type == "yolo_box") {
if (!desc.HasAttr("iou_aware") && !desc.HasAttr("iou_aware_factor"))
return false;
} else if (op_type == "solve") {
auto x_var_name = desc.Input("X")[0];
auto y_var_name = desc.Input("Y")[0];
auto* block = desc.Block();
if (block == nullptr) {
VLOG(3) << "The block desc is nullptr, we can't continue to analyze. "
"Developers need to check whether block_desc is passed in "
"the pass.";
return false;
}
auto* x_var_desc = block->FindVar(x_var_name);
auto* y_var_desc = block->FindVar(y_var_name);
auto x_dtype = x_var_desc->GetDataType();
auto y_dtype = y_var_desc->GetDataType();
if (x_dtype == framework::proto::VarType::FP64 ||
y_dtype == framework::proto::VarType::FP64) {
VLOG(3) << op_type << " not support input of FP64.";
return false;
}
}
// TODO(lizexu123): the tensorrt version on Windows supports low-level
// and inconsistent supportformation
if (op_type == "argsort") {
if (!with_dynamic_shape) {
VLOG(3) << "Ops(" << op_type << ") do not support static shape yet.";
return false;
}
}
if (use_no_calib_int8) {
return false;
} else {
framework::InitDefaultKernelSignatureMap();
bool res = phi::OpUtilsMap::Instance().HasArgumentMappingFn(op_type) ||
phi::DefaultKernelSignatureMap::Instance().Has(op_type);
if (!res) {
VLOG(3) << op_type << " has no KernelSignature";
return false;
}
res = phi::KernelFactory::Instance().HasCompatiblePhiKernel(op_type);
if (!res) {
VLOG(3) << op_type << " has no CompatiblePhiKernel in phi.";
return false;
}
auto& dynamic_infermeta_factory =
tensorrt::DynamicMetaFnFactory::Instance();
res = dynamic_infermeta_factory.Contains(op_type);
if (!res) {
VLOG(3) << op_type << " has no DynamicMetaFn.";
return false;
}
if (forbid_dynamic_op_enter_into_trt && IsDynamicShapeOp(desc)) {
return false;
}
return true;
}
}
};
struct CustomPluginTeller : public Teller {
public:
CustomPluginTeller() = default;
bool operator()(const framework::OpDesc& desc,
bool use_no_calib_int8 = false,
bool with_dynamic_shape = false,
bool forbid_dynamic_op_enter_into_trt = false,
bool use_explicit_quantization = false) override {
const std::string op_type = desc.Type();
std::string expect_plugin_name;
if (with_dynamic_shape) {
expect_plugin_name = op_type + "_paddle_trt_dynamic_plugin";
} else {
expect_plugin_name = op_type + "_paddle_trt_plugin";
}
int num = 0;
auto creators = GetPluginRegistry()->getPluginCreatorList(&num);
for (int i = 0; i < num; i++) {
if (std::string(creators[i]->getPluginName()) == expect_plugin_name)
return true;
}
return false;
if (forbid_dynamic_op_enter_into_trt && IsDynamicShapeOp(desc)) {
return false;
}
}
};
struct CustomGenericPluginTeller : public Teller {
CustomGenericPluginTeller() = default;
bool operator()(const framework::OpDesc& desc,
bool use_no_calib_int8 = false,
bool with_dynamic_shape = false,
bool forbid_dynamic_op_enter_into_trt = false,
bool use_explicit_quantization = false) override {
const std::string op_type = desc.Type();
auto& op_meta_info_map = OpMetaInfoMap::Instance();
const auto& meta_info_map = op_meta_info_map.GetMap();
if (meta_info_map.count(op_type) > 0) {
auto& op_info = meta_info_map.at(op_type).front();
auto& trt_infer_shape_fn = OpMetaInfoHelper::GetTrtInferShapeFn(op_info);
if (trt_infer_shape_fn == nullptr) {
VLOG(3) << op_type
<< " has no trt getOutputDimensions function. Please set by "
"SetTrtInferShapeFn.";
return false;
}
auto& trt_supports_format_config =
OpMetaInfoHelper::GetTrtSupportsFormatConfig(op_info);
if (trt_supports_format_config.empty()) {
VLOG(3)
<< op_type
<< " has no trt supportsFormatCombination config. Please set by "
"SetTrtSupportsFormatConfig.";
return false;
}
return true;
}
VLOG(3) << op_type << " has no meta info";
return false;
if (forbid_dynamic_op_enter_into_trt && IsDynamicShapeOp(desc)) {
return false;
}
}
};
bool OpTeller::Tell(const framework::ir::Node* node,
bool use_no_calib_int8,
bool with_dynamic_shape,
bool forbid_dynamic_op_enter_into_trt,
bool use_explicit_quantization) {
const std::string op_type = node->Op()->Type();
const framework::OpDesc desc = *node->Op();
// do not support the op which is labeled the `skip_quant`
if ((desc.HasAttr("namescope") &&
PADDLE_GET_CONST(std::string, desc.GetAttr("op_namescope")) ==
"/skip_quant_2/") ||
desc.HasAttr("skip_quant"))
return false;
auto& default_teller = GetDefaultTeller();
if ((*default_teller)(desc,
use_no_calib_int8,
with_dynamic_shape,
forbid_dynamic_op_enter_into_trt,
use_explicit_quantization)) {
SetOpConverterType(node->Op(), OpConverterType::Default);
return true;
}
auto& generic_plugin_teller = GetGenericPluginTeller();
if ((*generic_plugin_teller)(desc,
use_no_calib_int8,
with_dynamic_shape,
forbid_dynamic_op_enter_into_trt,
use_explicit_quantization)) {
SetOpConverterType(node->Op(), OpConverterType::GenericPluginCreator);
return true;
}
auto& custom_plugin_teller = GetCustomPluginTeller();
if ((*custom_plugin_teller)(desc,
use_no_calib_int8,
with_dynamic_shape,
forbid_dynamic_op_enter_into_trt,
use_explicit_quantization)) {
SetOpConverterType(node->Op(), OpConverterType::CustomPluginCreator);
return true;
}
auto& custom_generic_plugin_teller = GetCustomGenericPluginTeller();
if ((*custom_generic_plugin_teller)(desc,
use_no_calib_int8,
with_dynamic_shape,
forbid_dynamic_op_enter_into_trt,
use_explicit_quantization)) {
SetOpConverterType(node->Op(), OpConverterType::CustomGenericPluginCreator);
return true;
}
return false;
}
OpTeller::OpTeller() { // NOLINT
tellers_.emplace_back(new tensorrt::SimpleOpTypeSetTeller);
tellers_.emplace_back(new tensorrt::GenericPluginTeller);
tellers_.emplace_back(new tensorrt::CustomPluginTeller);
tellers_.emplace_back(new tensorrt::CustomGenericPluginTeller);
}
} // namespace paddle::inference::tensorrt
Reference in New Issue View Git Blame Copy Permalink