diff --git a/paddle/fluid/inference/tensorrt/op_teller.cc b/paddle/fluid/inference/tensorrt/op_teller.cc index d0e5e346e73290cf98db0bf6628880ed58fc0fdb..d5655c5f87de3e28a98ea8c6312cd9e5de5b5358 100644 --- a/paddle/fluid/inference/tensorrt/op_teller.cc +++ b/paddle/fluid/inference/tensorrt/op_teller.cc @@ -513,7 +513,12 @@ bool OpTeller::Tell(const framework::ir::Node* node, bool use_no_calib_int8, return false; } } - + auto batch_norm_inputs = desc.Inputs(); + if (batch_norm_inputs.find("MomentumTensor") != batch_norm_inputs.end()) { + if (desc.Input("MomentumTensor").size() >= 1) { + return false; + } + } if (desc.Output("Y").size() != 1) { VLOG(3) << "Invalid output Y's size of batch_norm TRT " "converter. Expected 1, received " diff --git a/python/paddle/fluid/tests/unittests/ir/inference/test_trt_convert_batch_norm.py b/python/paddle/fluid/tests/unittests/ir/inference/test_trt_convert_batch_norm.py new file mode 100644 index 0000000000000000000000000000000000000000..ceda10d5d94aa04302ffc1dc1c1ab123b7966487 --- /dev/null +++ b/python/paddle/fluid/tests/unittests/ir/inference/test_trt_convert_batch_norm.py @@ -0,0 +1,220 @@ +# Copyright (c) 2021 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. + +from trt_layer_auto_scan_test import TrtLayerAutoScanTest, SkipReasons +from program_config import TensorConfig, ProgramConfig +import numpy as np +import paddle.inference as paddle_infer +from functools import partial +from typing import Optional, List, Callable, Dict, Any, Set + + +class TrtConvertBatchNormTest(TrtLayerAutoScanTest): + def is_program_valid(self, program_config: ProgramConfig) -> bool: + return True + + def sample_program_configs(self): + def generate_input1(attrs: List[Dict[str, Any]], batch): + if self.dims == 4: + if attrs[0]['data_layout'] == "NCHW": + return np.ones([batch, 3, 24, 24]).astype(np.float32) + elif attrs[0]['data_layout'] == "NHWC": + return np.ones([batch, 24, 24, 3]).astype(np.float32) + elif self.dims == 3: + return np.ones([batch, 3, 24]).astype(np.float32) + elif self.dims == 2: + return np.ones([batch, 3]).astype(np.float32) + + def generate_bias(attrs: List[Dict[str, Any]], batch): + return np.full((3), 0.9).astype("float32") + + def generate_mean(attrs: List[Dict[str, Any]], batch): + return np.full((3), 0.9).astype("float32") + + def generate_scale(attrs: List[Dict[str, Any]], batch): + return np.full((3), 1.1).astype("float32") + + def generate_variance(attrs: List[Dict[str, Any]], batch): + return np.full((3), 1.2).astype("float32") + + def generate_MomentumTensor(attrs: List[Dict[str, Any]], batch): + return np.full((3), 0.9).astype("float32") + + for dims in [2, 3, 4]: + for num_input in [0, 1]: + for batch in [1, 2, 4]: + for epsilon in [1e-6, 1e-5, 1e-4]: + for data_layout in ["NCHW"]: + for momentum in [0.9, 0.8]: + self.num_input = num_input + self.dims = dims + dics = [{ + "epsilon": epsilon, + "data_layout": data_layout, + "momentum": momentum, + "is_test": True, + "trainable_statistics": False + }, {}] + dics_intput = [{ + "X": ["batch_norm_input"], + "Bias": ["Bias"], + "Mean": ["Mean"], + "Scale": ["Scale"], + "Variance": ["Variance"], + "MomentumTensor": ["MomentumTensor"] + }, { + "X": ["batch_norm_input"], + "Bias": ["Bias"], + "Mean": ["Mean"], + "Scale": ["Scale"], + "Variance": ["Variance"] + }] + dics_intputs = [{ + "Bias": TensorConfig(data_gen=partial( + generate_bias, dics, batch)), + "Mean": TensorConfig(data_gen=partial( + generate_mean, dics, batch)), + "Scale": TensorConfig(data_gen=partial( + generate_scale, dics, batch)), + "Variance": TensorConfig(data_gen=partial( + generate_variance, dics, batch)), + "MomentumTensor": + TensorConfig(data_gen=partial( + generate_MomentumTensor, dics, batch)), + }, { + "Bias": TensorConfig(data_gen=partial( + generate_bias, dics, batch)), + "Mean": TensorConfig(data_gen=partial( + generate_mean, dics, batch)), + "Scale": TensorConfig(data_gen=partial( + generate_scale, dics, batch)), + "Variance": TensorConfig(data_gen=partial( + generate_variance, dics, batch)) + }] + ops_config = [{ + "op_type": "batch_norm", + "op_inputs": dics_intput[num_input], + "op_outputs": { + "Y": ["batch_norm_out"], + "MeanOut": ["Mean"], + "VarianceOut": ["Variance"], + "SavedMean": ["SavedMean"], + "SavedVariance": ["SavedVariance"] + }, + "op_attrs": dics[0] + }] + ops = self.generate_op_config(ops_config) + program_config = ProgramConfig( + ops=ops, + weights=dics_intputs[num_input], + inputs={ + "batch_norm_input": TensorConfig( + data_gen=partial(generate_input1, + dics, batch)) + }, + outputs=["batch_norm_out"]) + + yield program_config + + def sample_predictor_configs( + self, program_config) -> (paddle_infer.Config, List[int], float): + def generate_dynamic_shape(attrs): + if self.dims == 4: + if attrs[0]['data_layout'] == "NCHW": + self.dynamic_shape.min_input_shape = { + "batch_norm_input": [1, 3, 24, 24] + } + self.dynamic_shape.max_input_shape = { + "batch_norm_input": [4, 3, 48, 48] + } + self.dynamic_shape.opt_input_shape = { + "batch_norm_input": [1, 3, 24, 48] + } + elif attrs[0]['data_layout'] == "NHWC": + self.dynamic_shape.min_input_shape = { + "batch_norm_input": [1, 24, 24, 3] + } + self.dynamic_shape.max_input_shape = { + "batch_norm_input": [4, 48, 48, 3] + } + self.dynamic_shape.opt_input_shape = { + "batch_norm_input": [1, 24, 48, 3] + } + elif self.dims == 3: + self.dynamic_shape.min_input_shape = { + "batch_norm_input": [1, 3, 24] + } + self.dynamic_shape.max_input_shape = { + "batch_norm_input": [4, 3, 48] + } + self.dynamic_shape.opt_input_shape = { + "batch_norm_input": [1, 3, 48] + } + elif self.dims == 2: + self.dynamic_shape.min_input_shape = { + "batch_norm_input": [1, 3] + } + self.dynamic_shape.max_input_shape = { + "batch_norm_input": [4, 3] + } + self.dynamic_shape.opt_input_shape = { + "batch_norm_input": [1, 3] + } + + 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 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-5 + + # 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-5 + + def add_skip_trt_case(self): + def teller1(program_config, predictor_config): + if len(program_config.weights) == 5: + return True + return False + + self.add_skip_case(teller1, SkipReasons.TRT_NOT_SUPPORT, + "INPUT MomentumTensor NOT SUPPORT") + + def test(self): + self.add_skip_trt_case() + self.run_test() + + +if __name__ == "__main__": + unittest.main()