-
Notifications
You must be signed in to change notification settings - Fork 5.6k
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
[Paddle Inference] Support range trt converter and add scalar interfa…
…ce. (#48697) * add_range * add_range
- Loading branch information
1 parent
7507956
commit aee2db0
Showing
7 changed files
with
321 additions
and
5 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,65 @@ | ||
| /* 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 RangeOpConverter : public OpConverter { | ||
| public: | ||
| void operator()(const framework::proto::OpDesc& op, | ||
| const framework::Scope& scope, | ||
| bool test_mode) override { | ||
| VLOG(3) << "convert a range op to tensorrt layer"; | ||
| framework::OpDesc op_desc(op, nullptr); | ||
| nvinfer1::ILayer* layer = nullptr; | ||
| nvinfer1::ITensor* quotient_tensor; | ||
|
|
||
| // Declare inputs | ||
| auto* start = engine_->GetITensor(op_desc.Input("Start")[0]); | ||
| auto* end = engine_->GetITensor(op_desc.Input("End")[0]); | ||
| auto* step = engine_->GetITensor(op_desc.Input("Step")[0]); | ||
| auto output_name = op_desc.Output("Out")[0]; | ||
|
|
||
| auto zero_tensor = Add1DConstantLayer(0, output_name + "_zero_tensor_"); | ||
| auto fquotient_tensor = FloorDiv(Sub(start, end), step); | ||
| if (start->getType() == nvinfer1::DataType::kFLOAT) { | ||
| auto* cast_int32_layer = | ||
| TRT_ENGINE_ADD_LAYER(engine_, Identity, *fquotient_tensor); | ||
| cast_int32_layer->setOutputType(0, nvinfer1::DataType::kINT32); | ||
| cast_int32_layer->getOutput(0)->setType(nvinfer1::DataType::kINT32); | ||
| quotient_tensor = cast_int32_layer->getOutput(0); | ||
| } else { | ||
| quotient_tensor = fquotient_tensor; | ||
| } | ||
| auto number_tensor = Max(Sub(zero_tensor, quotient_tensor), zero_tensor); | ||
| auto* start1 = engine_->GetITensor(op_desc.Input("Start")[0], true); | ||
|
|
||
| layer = TRT_ENGINE_ADD_LAYER( | ||
| engine_, Fill, nvinfer1::Dims{}, nvinfer1::FillOperation::kLINSPACE); | ||
| layer->setInput(0, *number_tensor); | ||
| layer->setInput(1, *start1); | ||
| layer->setInput(2, *step); | ||
|
|
||
| RreplenishLayerAndOutput(layer, "range", {output_name}, test_mode); | ||
| } | ||
| }; | ||
|
|
||
| } // namespace tensorrt | ||
| } // namespace inference | ||
| } // namespace paddle | ||
|
|
||
| REGISTER_TRT_OP_CONVERTER(range, RangeOpConverter); |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
230 changes: 230 additions & 0 deletions
230
python/paddle/fluid/tests/unittests/ir/inference/test_trt_convert_range.py
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,230 @@ | ||
| # 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 TrtConvertRangeDynamicTest(TrtLayerAutoScanTest): | ||
| def is_program_valid(self, program_config: ProgramConfig) -> bool: | ||
| return True | ||
|
|
||
| def sample_program_configs(self): | ||
| def generate_input(): | ||
| return np.array([1]).astype(np.int32) | ||
|
|
||
| for in_dtype in [2]: | ||
| self.in_dtype = in_dtype | ||
| dics = [{}] | ||
| ops_config = [ | ||
| { | ||
| "op_type": "fill_constant", | ||
| "op_inputs": {}, | ||
| "op_outputs": {"Out": ["start_data"]}, | ||
| "op_attrs": { | ||
| "dtype": self.in_dtype, | ||
| "str_value": "7", | ||
| "shape": [1], | ||
| }, | ||
| }, | ||
| { | ||
| "op_type": "fill_constant", | ||
| "op_inputs": {}, | ||
| "op_outputs": {"Out": ["end_data"]}, | ||
| "op_attrs": { | ||
| "dtype": self.in_dtype, | ||
| "str_value": "256", | ||
| "shape": [1], | ||
| }, | ||
| }, | ||
| { | ||
| "op_type": "fill_constant", | ||
| "op_inputs": {}, | ||
| "op_outputs": {"Out": ["step_data"]}, | ||
| "op_attrs": { | ||
| "dtype": self.in_dtype, | ||
| "str_value": "1", | ||
| "shape": [1], | ||
| }, | ||
| }, | ||
| { | ||
| "op_type": "range", | ||
| "op_inputs": { | ||
| "Start": ["start_data"], | ||
| "End": ["end_data"], | ||
| "Step": ["step_data"], | ||
| }, | ||
| "op_outputs": {"Out": ["range_output_data1"]}, | ||
| "op_attrs": dics[0], | ||
| }, | ||
| { | ||
| "op_type": "cast", | ||
| "op_inputs": {"X": ["range_output_data1"]}, | ||
| "op_outputs": {"Out": ["range_output_data"]}, | ||
| "op_attrs": {"in_dtype": self.in_dtype, "out_dtype": 5}, | ||
| }, | ||
| ] | ||
| ops = self.generate_op_config(ops_config) | ||
|
|
||
| program_config = ProgramConfig( | ||
| ops=ops, | ||
| weights={}, | ||
| inputs={ | ||
| "step_data": TensorConfig(data_gen=partial(generate_input)), | ||
| }, | ||
| outputs=["range_output_data"], | ||
| ) | ||
|
|
||
| yield program_config | ||
|
|
||
| def sample_predictor_configs( | ||
| self, program_config | ||
| ) -> (paddle_infer.Config, List[int], float): | ||
| def generate_dynamic_shape(attrs): | ||
| self.dynamic_shape.min_input_shape = { | ||
| "start_data": [1], | ||
| "end_data": [1], | ||
| "step_data": [1], | ||
| } | ||
| self.dynamic_shape.max_input_shape = { | ||
| "start_data": [1], | ||
| "end_data": [1], | ||
| "step_data": [1], | ||
| } | ||
| self.dynamic_shape.opt_input_shape = { | ||
| "start_data": [1], | ||
| "end_data": [1], | ||
| "step_data": [1], | ||
| } | ||
|
|
||
| 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): | ||
| return 1, 2 | ||
|
|
||
| attrs = [ | ||
| program_config.ops[i].attrs for i in range(len(program_config.ops)) | ||
| ] | ||
|
|
||
| # 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-2 | ||
|
|
||
| def test(self): | ||
| self.run_test() | ||
|
|
||
|
|
||
| class TrtConvertRangeStaticTest(TrtLayerAutoScanTest): | ||
| def is_program_valid(self, program_config: ProgramConfig) -> bool: | ||
| return True | ||
|
|
||
| def sample_program_configs(self): | ||
| def generate_input(): | ||
| return np.array([0]).astype(np.int32) | ||
|
|
||
| def generate_input1(): | ||
| return np.array([128]).astype(np.int32) | ||
|
|
||
| def generate_input2(): | ||
| return np.array([1]).astype(np.int32) | ||
|
|
||
| for in_dtype in [2, 5]: | ||
| self.in_dtype = in_dtype | ||
| dics = [{}] | ||
| ops_config = [ | ||
| { | ||
| "op_type": "range", | ||
| "op_inputs": { | ||
| "Start": ["start_data"], | ||
| "End": ["end_data"], | ||
| "Step": ["step_data"], | ||
| }, | ||
| "op_outputs": {"Out": ["range_output_data1"]}, | ||
| "op_attrs": dics[0], | ||
| }, | ||
| { | ||
| "op_type": "cast", | ||
| "op_inputs": {"X": ["range_output_data1"]}, | ||
| "op_outputs": {"Out": ["range_output_data"]}, | ||
| "op_attrs": {"in_dtype": self.in_dtype, "out_dtype": 5}, | ||
| }, | ||
| ] | ||
| ops = self.generate_op_config(ops_config) | ||
|
|
||
| program_config = ProgramConfig( | ||
| ops=ops, | ||
| weights={}, | ||
| inputs={ | ||
| "start_data": TensorConfig( | ||
| data_gen=partial(generate_input) | ||
| ), | ||
| "end_data": TensorConfig(data_gen=partial(generate_input1)), | ||
| "step_data": TensorConfig( | ||
| data_gen=partial(generate_input2) | ||
| ), | ||
| }, | ||
| outputs=["range_output_data"], | ||
| ) | ||
|
|
||
| yield program_config | ||
|
|
||
| def sample_predictor_configs( | ||
| self, program_config | ||
| ) -> (paddle_infer.Config, List[int], float): | ||
| 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): | ||
| return 0, 6 | ||
|
|
||
| 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-2 | ||
|
|
||
| def test(self): | ||
| self.run_test() | ||
|
|
||
|
|
||
| if __name__ == "__main__": | ||
| unittest.main() |