[inference][trt] add more unary op and square (#48534)

* add more unary op and square

paddle/fluid/inference/api/analysis_predictor.cc

@@ -2284,13 +2284,9 @@ USE_TRT_CONVERTER(flatten_contiguous_range);
 USE_TRT_CONVERTER(matmul);
 USE_TRT_CONVERTER(matmul_v2);
 USE_TRT_CONVERTER(bmm);
-USE_TRT_CONVERTER(rsqrt);
 USE_TRT_CONVERTER(conv2d);
 USE_TRT_CONVERTER(relu);
-USE_TRT_CONVERTER(exp);
-USE_TRT_CONVERTER(log);
 USE_TRT_CONVERTER(sigmoid);
-USE_TRT_CONVERTER(tanh);
 USE_TRT_CONVERTER(fc);
 USE_TRT_CONVERTER(pool2d);
 USE_TRT_CONVERTER(softmax);
@@ -2346,6 +2342,32 @@ USE_TRT_CONVERTER(conv3d_transpose);
 USE_TRT_CONVERTER(mish);
 USE_TRT_CONVERTER(deformable_conv);
 USE_TRT_CONVERTER(pool3d)
+USE_TRT_CONVERTER(square);
+// unary op
+USE_TRT_CONVERTER(exp);
+USE_TRT_CONVERTER(log);
+USE_TRT_CONVERTER(sqrt);
+USE_TRT_CONVERTER(reciprocal);
+USE_TRT_CONVERTER(abs);
+USE_TRT_CONVERTER(sin);
+USE_TRT_CONVERTER(cos);
+USE_TRT_CONVERTER(tan);
+USE_TRT_CONVERTER(sinh);
+USE_TRT_CONVERTER(cosh);
+USE_TRT_CONVERTER(tanh);
+USE_TRT_CONVERTER(asin);
+USE_TRT_CONVERTER(acos);
+USE_TRT_CONVERTER(atan);
+USE_TRT_CONVERTER(asinh);
+USE_TRT_CONVERTER(acosh);
+USE_TRT_CONVERTER(atanh);
+USE_TRT_CONVERTER(ceil);
+USE_TRT_CONVERTER(floor);
+#if IS_TRT_VERSION_GE(8200)
+USE_TRT_CONVERTER(round);
+USE_TRT_CONVERTER(sign);
+#endif
+USE_TRT_CONVERTER(rsqrt);
 USE_TRT_CONVERTER(fused_preln_embedding_eltwise_layernorm)
 USE_TRT_CONVERTER(fused_embedding_eltwise_layernorm);
 USE_TRT_CONVERTER(preln_skip_layernorm)

paddle/fluid/inference/tensorrt/convert/CMakeLists.txt

@@ -18,6 +18,7 @@ list(
   group_norm_op.cc
   pad_op.cc
   split_op.cc
+  square_op.cc
   prelu_op.cc
   leaky_relu_op.cc
   gelu_op.cc

paddle/fluid/inference/tensorrt/convert/square_op.cc

+/* Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License. */
+
+#include "paddle/fluid/inference/tensorrt/convert/op_converter.h"
+
+namespace paddle {
+namespace inference {
+namespace tensorrt {
+
+class SquareOpConverter : public OpConverter {
+ public:
+  SquareOpConverter() {}
+  void operator()(const framework::proto::OpDesc& op,
+                  const framework::Scope& scope,
+                  bool test_mode) override {
+    framework::OpDesc op_desc(op, nullptr);
+    VLOG(3) << "convert a fluid sqaure op to tensorrt layer ";
+    nvinfer1::ITensor* input_tensor =
+        engine_->GetITensor(op_desc.Input("X")[0]);
+
+    auto* layer = TRT_ENGINE_ADD_LAYER(engine_,
+                                       ElementWise,
+                                       *input_tensor,
+                                       *input_tensor,
+                                       nvinfer1::ElementWiseOperation::kPROD);
+
+    auto output_name = op_desc.Output("Out")[0];
+    RreplenishLayerAndOutput(layer, "square", {output_name}, test_mode);
+  }
+};
+
+}  // namespace tensorrt
+}  // namespace inference
+}  // namespace paddle
+
+REGISTER_TRT_OP_CONVERTER(square, SquareOpConverter);

paddle/fluid/inference/tensorrt/convert/unary_op.cc

@@ -85,6 +85,7 @@ const std::unordered_map<std::string, std::vector<nvinfer1::UnaryOperation>>
         {"acos", {nvinfer1::UnaryOperation::kACOS}},
         {"atan", {nvinfer1::UnaryOperation::kATAN}},
         {"asinh", {nvinfer1::UnaryOperation::kASINH}},
+        {"acosh", {nvinfer1::UnaryOperation::kACOSH}},
         {"atanh", {nvinfer1::UnaryOperation::kATANH}},
         {"ceil", {nvinfer1::UnaryOperation::kCEIL}},
         {"floor", {nvinfer1::UnaryOperation::kFLOOR}},
@@ -92,12 +93,13 @@ const std::unordered_map<std::string, std::vector<nvinfer1::UnaryOperation>>
          {nvinfer1::UnaryOperation::kSQRT, nvinfer1::UnaryOperation::kRECIP}},
         {"logical_not", {nvinfer1::UnaryOperation::kNOT}},
         {"reciprocal", {nvinfer1::UnaryOperation::kRECIP}},
-#if IS_TRT_VERSION_GE(8200)
-        {"sign", {nvinfer1::UnaryOperation::kSIGN}},
-#endif
 #if IS_TRT_VERSION_GE(7000)
         {"erf", {nvinfer1::UnaryOperation::kERF}},
 #endif
+#if IS_TRT_VERSION_GE(8200)
+        {"sign", {nvinfer1::UnaryOperation::kSIGN}},
+        {"round", {nvinfer1::UnaryOperation::kROUND}},
+#endif
 };
 
 class ExpOpConverter : public UnaryOpConverter {
@@ -154,6 +156,10 @@ class AsinhOpConverter : public UnaryOpConverter {
  public:
   AsinhOpConverter() { op_type_ = "asinh"; }
 };
+class AcoshOpConverter : public UnaryOpConverter {
+ public:
+  AcoshOpConverter() { op_type_ = "acosh"; }
+};
 class AtanhOpConverter : public UnaryOpConverter {
  public:
   AtanhOpConverter() { op_type_ = "atanh"; }
@@ -194,6 +200,10 @@ class ErfOpConverter : public UnaryOpConverter {
  public:
   ErfOpConverter() { op_type_ = "erf"; }
 };
+class RoundOpConverter : public UnaryOpConverter {
+ public:
+  RoundOpConverter() { op_type_ = "round"; }
+};
 #endif
 
 }  // namespace tensorrt
@@ -213,15 +223,17 @@ REGISTER_TRT_OP_CONVERTER(asin, AsinOpConverter);
 REGISTER_TRT_OP_CONVERTER(acos, AcosOpConverter);
 REGISTER_TRT_OP_CONVERTER(atan, AtanOpConverter);
 REGISTER_TRT_OP_CONVERTER(asinh, AsinhOpConverter);
+REGISTER_TRT_OP_CONVERTER(acosh, AcoshOpConverter);
 REGISTER_TRT_OP_CONVERTER(atanh, AtanhOpConverter);
 REGISTER_TRT_OP_CONVERTER(ceil, CeilOpConverter);
 REGISTER_TRT_OP_CONVERTER(floor, FloorOpConverter);
 REGISTER_TRT_OP_CONVERTER(rsqrt, RsqrtOpConverter);
 REGISTER_TRT_OP_CONVERTER(logical_not, LogicalNotOpConverter);
 REGISTER_TRT_OP_CONVERTER(reciprocal, ReciprocalOpConverter);
-#if IS_TRT_VERSION_GE(8200)
-REGISTER_TRT_OP_CONVERTER(sign, SignOpConverter);
-#endif
 #if IS_TRT_VERSION_GE(7000)
 REGISTER_TRT_OP_CONVERTER(erf, ErfOpConverter);
 #endif
+#if IS_TRT_VERSION_GE(8200)
+REGISTER_TRT_OP_CONVERTER(sign, SignOpConverter);
+REGISTER_TRT_OP_CONVERTER(round, RoundOpConverter);
+#endif

paddle/fluid/inference/tensorrt/op_teller.cc

@@ -65,6 +65,10 @@ struct SimpleOpTypeSetTeller : public Teller {
     int8_teller_set.insert("sparse_fc");
     teller_set.insert("sparse_multihead_matmul");
     int8_teller_set.insert("sparse_multihead_matmul");
+#endif
+#if IS_TRT_VERSION_GE(8200)
+    teller_set.insert("round");
+    int8_teller_set.insert("round");
 #endif
   }
 
@@ -79,18 +83,18 @@ struct SimpleOpTypeSetTeller : public Teller {
         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",      "atanh",
-        "ceil",        "floor",      "erf",
-        "reciprocal",  "silu",       "celu",
-        "tanh_shrink", "logsigmoid", "sign",
-        "logical_not"};
+        "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"};
     if (act_op_list.find(op_type) != act_op_list.end()) {
       auto* block = desc.Block();
       if (block == nullptr) {
@@ -2456,6 +2460,7 @@ struct SimpleOpTypeSetTeller : public Teller {
       "acos",
       "atan",
       "asinh",
+      "acosh",
       "atanh",
       "ceil",
       "floor",
@@ -2464,6 +2469,7 @@ struct SimpleOpTypeSetTeller : public Teller {
       "reciprocal",
       "logical_not",
       "erf",
+      "square",
       "softmax",
       "sigmoid",
       "hard_swish",
@@ -2599,6 +2605,7 @@ struct SimpleOpTypeSetTeller : public Teller {
       "acos",
       "atan",
       "asinh",
+      "acosh",
       "atanh",
       "ceil",
       "floor",
@@ -2607,6 +2614,7 @@ struct SimpleOpTypeSetTeller : public Teller {
       "reciprocal",
       "logical_not",
       "erf",
+      "square",
       "softmax",
       "sigmoid",
       "hard_swish",

python/paddle/fluid/tests/unittests/ir/inference/test_trt_convert_activation.py

@@ -25,6 +25,10 @@ import paddle.inference as paddle_infer
 
 class TrtConvertActivationTest(TrtLayerAutoScanTest):
     def is_program_valid(self, program_config: ProgramConfig) -> bool:
+        ver = paddle_infer.get_trt_compile_version()
+        if ver[0] * 1000 + ver[1] * 100 + ver[0] * 10 < 8200:
+            if program_config.ops[0].type == "round":
+                return False
         return True
 
     def sample_program_configs(self):

python/paddle/fluid/tests/unittests/ir/inference/test_trt_convert_square.py

+# Copyright (c) 2022 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.
+
+import unittest
+from functools import partial
+from typing import List
+
+import numpy as np
+from program_config import ProgramConfig, TensorConfig
+from trt_layer_auto_scan_test import TrtLayerAutoScanTest
+
+import paddle.inference as paddle_infer
+
+
+class TrtConvertSquareTest(TrtLayerAutoScanTest):
+    def is_program_valid(self, program_config: ProgramConfig) -> bool:
+        return True
+
+    def sample_program_configs(self):
+        def generate_input1(dims):
+            if dims == 1:
+                return np.ones([3]).astype(np.float32)
+            elif dims == 2:
+                return np.ones([3, 64]).astype(np.float32)
+            elif dims == 3:
+                return np.ones([3, 64, 64]).astype(np.float32)
+            else:
+                return np.ones([1, 3, 64, 64]).astype(np.float32)
+
+        for dims in [1, 2, 3, 4]:
+            for alpha in [1.0, 2.0, 3.0]:
+                self.dims = dims
+
+                ops_config = [
+                    {
+                        "op_type": "square",
+                        "op_inputs": {
+                            "X": ["input_data"],
+                        },
+                        "op_outputs": {"Out": ["output_data"]},
+                        "op_attrs": {},
+                    }
+                ]
+                ops = self.generate_op_config(ops_config)
+
+                program_config = ProgramConfig(
+                    ops=ops,
+                    weights={},
+                    inputs={
+                        "input_data": TensorConfig(
+                            data_gen=partial(generate_input1, dims)
+                        )
+                    },
+                    outputs=["output_data"],
+                )
+
+                yield program_config
+
+    def sample_predictor_configs(
+        self, program_config
+    ) -> (paddle_infer.Config, List[int], float):
+        def generate_dynamic_shape(attrs):
+            if self.dims == 1:
+                self.dynamic_shape.min_input_shape = {"input_data": [1]}
+                self.dynamic_shape.max_input_shape = {"input_data": [128]}
+                self.dynamic_shape.opt_input_shape = {"input_data": [64]}
+            elif self.dims == 2:
+                self.dynamic_shape.min_input_shape = {"input_data": [1, 32]}
+                self.dynamic_shape.max_input_shape = {"input_data": [4, 64]}
+                self.dynamic_shape.opt_input_shape = {"input_data": [3, 64]}
+            elif self.dims == 3:
+                self.dynamic_shape.min_input_shape = {"input_data": [1, 32, 32]}
+                self.dynamic_shape.max_input_shape = {
+                    "input_data": [10, 64, 64]
+                }
+                self.dynamic_shape.opt_input_shape = {"input_data": [3, 64, 64]}
+            else:
+                self.dynamic_shape.min_input_shape = {
+                    "input_data": [1, 3, 32, 32]
+                }
+                self.dynamic_shape.max_input_shape = {
+                    "input_data": [4, 3, 64, 64]
+                }
+                self.dynamic_shape.opt_input_shape = {
+                    "input_data": [1, 3, 64, 64]
+                }
+
+        def clear_dynamic_shape():
+            self.dynamic_shape.min_input_shape = {}
+            self.dynamic_shape.max_input_shape = {}
+            self.dynamic_shape.opt_input_shape = {}
+
+        def generate_trt_nodes_num(attrs, dynamic_shape):
+            if not dynamic_shape and self.dims == 1:
+                return 0, 3
+            return 1, 2
+
+        attrs = [
+            program_config.ops[i].attrs for i in range(len(program_config.ops))
+        ]
+
+        # for static_shape
+        clear_dynamic_shape()
+        self.trt_param.precision = paddle_infer.PrecisionType.Float32
+        yield self.create_inference_config(), generate_trt_nodes_num(
+            attrs, False
+        ), 1e-5
+        self.trt_param.precision = paddle_infer.PrecisionType.Half
+        yield self.create_inference_config(), generate_trt_nodes_num(
+            attrs, False
+        ), (1e-3, 1e-3)
+
+        # for dynamic_shape
+        generate_dynamic_shape(attrs)
+        self.trt_param.precision = paddle_infer.PrecisionType.Float32
+        yield self.create_inference_config(), generate_trt_nodes_num(
+            attrs, True
+        ), 1e-5
+        self.trt_param.precision = paddle_infer.PrecisionType.Half
+        yield self.create_inference_config(), generate_trt_nodes_num(
+            attrs, True
+        ), (1e-3, 1e-3)
+
+    def test(self):
+        self.run_test()
+
+
+if __name__ == "__main__":
+    unittest.main()

python/paddle/fluid/tests/unittests/ir/inference/test_trt_convert_unary.py

@@ -25,6 +25,10 @@ import paddle.inference as paddle_infer
 
 class TrtConvertActivationTest(TrtLayerAutoScanTest):
     def is_program_valid(self, program_config: ProgramConfig) -> bool:
+        ver = paddle_infer.get_trt_compile_version()
+        if ver[0] * 1000 + ver[1] * 100 + ver[0] * 10 < 8200:
+            if program_config.ops[0].type == "round":
+                return False
         return True
 
     def sample_program_configs(self):
@@ -54,11 +58,13 @@ class TrtConvertActivationTest(TrtLayerAutoScanTest):
                     "acos",
                     "atan",
                     "asinh",
+                    "acosh",
                     "atanh",
                     "ceil",
                     "floor",
                     "rsqrt",
                     "reciprocal",
+                    "round",
                     "sign",
                 ]:
                     self.dims = dims