| 
									
										
											  
											
												- build:
	- unify schema building in core and converter;
	- add more build script for android;
	- add linux build script for python;
- ops impl:
	- add floor mod support in binary;
	- use eltwise impl in add/max/sub/mul binary for optimization;
	- remove fake double support in cast;
	- fix 5d support for concat;
	- add adjX and adjY support for batch matmul;
	- optimize conv2d back prop filter;
	- add pad mode support for conv3d;
	- fix bug in conv2d & conv depthwise with very small feature map;
	- optimize binary without broacast;
	- add data types support for gather;
	- add gather ND support;
	- use uint8 data type in gather v2;
	- add transpose support for matmul;
	- add matrix band part;
	- add dim != 4 support for padding, reshape & tensor convert;
	- add pad type support for pool3d;
	- make ops based on TensorFlow Lite quantization optional;
	- add all & any support for reduction;
	- use type in parameter as output type in reduction;
	- add int support for unary;
	- add variable weight support for conv2d;
	- fix conv2d depthwise weights initialization;
	- fix type support for transpose;
	- fix grad outputs count for  reduce grad and reshape grad;
	- fix priorbox & detection output;
	- fix metal softmax error;
- python:
	- add runSessionWithCallBackInfo interface;
	- add max nodes limit (1400) for visualization tool;
	- fix save error in python3;
	- align default dim;
- convert:
	- add extra design for optimization;
	- add more post converting optimizers;
	- add caffe v1 weights blob support;
	- add cast, unary, conv transpose support for onnx model;
	- optimize batchnorm, conv with variable weights, prelu, reshape, slice, upsample for onnx model;
	- add cos/sin/atan/tan support for unary for tensorflow model;
	- add any/all support for reduction for tensorflow model;
	- add elu, conv3d, pool3d support for tensorflow model;
	- optimize argmax, batchnorm, concat, batch to space, conv with variable weights, prelu, slice for tensorflow model;
- others:
	- fix size computer lock;
	- fix thread pool deadlock;
	- add express & parameters in express;
	- rewrite blitter chooser without static map;
	- add tests for expr;
											
										 
											2019-10-29 13:37:26 +08:00
										 |  |  | //
 | 
					
						
							|  |  |  | //  TemplateMerge.hpp
 | 
					
						
							|  |  |  | //  MNNConverter
 | 
					
						
							|  |  |  | //
 | 
					
						
							|  |  |  | //  Created by MNN on 2019/09/16.
 | 
					
						
							|  |  |  | //  Copyright © 2018, Alibaba Group Holding Limited
 | 
					
						
							|  |  |  | //
 | 
					
						
							|  |  |  | 
 | 
					
						
							| 
									
										
										
										
											2019-12-27 22:16:57 +08:00
										 |  |  | #include <MNN/expr/Optimizer.hpp>
 | 
					
						
							| 
									
										
										
										
											2021-06-11 17:17:13 +08:00
										 |  |  | #include "Global.hpp"
 | 
					
						
							|  |  |  | #include "config.hpp"
 | 
					
						
							| 
									
										
										
										
											2020-11-25 19:03:07 +08:00
										 |  |  | 
 | 
					
						
							|  |  |  | #define MNN_THROW_CHECK(success, log) \
 | 
					
						
							|  |  |  | if(!(success)){ \ | 
					
						
							|  |  |  | MNN_ERROR("Check failed: %s ==> %s\n", #success, #log); \ | 
					
						
							|  |  |  | } | 
					
						
							| 
									
										
										
										
											2020-11-05 16:41:56 +08:00
										 |  |  | 
 | 
					
						
							| 
									
										
											  
											
												- build:
	- unify schema building in core and converter;
	- add more build script for android;
	- add linux build script for python;
- ops impl:
	- add floor mod support in binary;
	- use eltwise impl in add/max/sub/mul binary for optimization;
	- remove fake double support in cast;
	- fix 5d support for concat;
	- add adjX and adjY support for batch matmul;
	- optimize conv2d back prop filter;
	- add pad mode support for conv3d;
	- fix bug in conv2d & conv depthwise with very small feature map;
	- optimize binary without broacast;
	- add data types support for gather;
	- add gather ND support;
	- use uint8 data type in gather v2;
	- add transpose support for matmul;
	- add matrix band part;
	- add dim != 4 support for padding, reshape & tensor convert;
	- add pad type support for pool3d;
	- make ops based on TensorFlow Lite quantization optional;
	- add all & any support for reduction;
	- use type in parameter as output type in reduction;
	- add int support for unary;
	- add variable weight support for conv2d;
	- fix conv2d depthwise weights initialization;
	- fix type support for transpose;
	- fix grad outputs count for  reduce grad and reshape grad;
	- fix priorbox & detection output;
	- fix metal softmax error;
- python:
	- add runSessionWithCallBackInfo interface;
	- add max nodes limit (1400) for visualization tool;
	- fix save error in python3;
	- align default dim;
- convert:
	- add extra design for optimization;
	- add more post converting optimizers;
	- add caffe v1 weights blob support;
	- add cast, unary, conv transpose support for onnx model;
	- optimize batchnorm, conv with variable weights, prelu, reshape, slice, upsample for onnx model;
	- add cos/sin/atan/tan support for unary for tensorflow model;
	- add any/all support for reduction for tensorflow model;
	- add elu, conv3d, pool3d support for tensorflow model;
	- optimize argmax, batchnorm, concat, batch to space, conv with variable weights, prelu, slice for tensorflow model;
- others:
	- fix size computer lock;
	- fix thread pool deadlock;
	- add express & parameters in express;
	- rewrite blitter chooser without static map;
	- add tests for expr;
											
										 
											2019-10-29 13:37:26 +08:00
										 |  |  | namespace MNN { | 
					
						
							|  |  |  | namespace Express { | 
					
						
							| 
									
										
										
										
											2020-11-05 16:41:56 +08:00
										 |  |  | 
 | 
					
						
							|  |  |  | enum PassPriority : int { | 
					
						
							|  |  |  |    PASS_PRIORITY_FRONT = 0, | 
					
						
							|  |  |  |    PASS_PRIORITY_HIGH = 1, | 
					
						
							|  |  |  |    PASS_PRIORITY_MIDDLE = 2, | 
					
						
							|  |  |  |    PASS_PRIORITY_LOW = 3, | 
					
						
							|  |  |  |    PASS_PRIORITY_FINAL = 4, | 
					
						
							|  |  |  | }; | 
					
						
							|  |  |  | 
 | 
					
						
							| 
									
										
											  
											
												- build:
	- unify schema building in core and converter;
	- add more build script for android;
	- add linux build script for python;
- ops impl:
	- add floor mod support in binary;
	- use eltwise impl in add/max/sub/mul binary for optimization;
	- remove fake double support in cast;
	- fix 5d support for concat;
	- add adjX and adjY support for batch matmul;
	- optimize conv2d back prop filter;
	- add pad mode support for conv3d;
	- fix bug in conv2d & conv depthwise with very small feature map;
	- optimize binary without broacast;
	- add data types support for gather;
	- add gather ND support;
	- use uint8 data type in gather v2;
	- add transpose support for matmul;
	- add matrix band part;
	- add dim != 4 support for padding, reshape & tensor convert;
	- add pad type support for pool3d;
	- make ops based on TensorFlow Lite quantization optional;
	- add all & any support for reduction;
	- use type in parameter as output type in reduction;
	- add int support for unary;
	- add variable weight support for conv2d;
	- fix conv2d depthwise weights initialization;
	- fix type support for transpose;
	- fix grad outputs count for  reduce grad and reshape grad;
	- fix priorbox & detection output;
	- fix metal softmax error;
- python:
	- add runSessionWithCallBackInfo interface;
	- add max nodes limit (1400) for visualization tool;
	- fix save error in python3;
	- align default dim;
- convert:
	- add extra design for optimization;
	- add more post converting optimizers;
	- add caffe v1 weights blob support;
	- add cast, unary, conv transpose support for onnx model;
	- optimize batchnorm, conv with variable weights, prelu, reshape, slice, upsample for onnx model;
	- add cos/sin/atan/tan support for unary for tensorflow model;
	- add any/all support for reduction for tensorflow model;
	- add elu, conv3d, pool3d support for tensorflow model;
	- optimize argmax, batchnorm, concat, batch to space, conv with variable weights, prelu, slice for tensorflow model;
- others:
	- fix size computer lock;
	- fix thread pool deadlock;
	- add express & parameters in express;
	- rewrite blitter chooser without static map;
	- add tests for expr;
											
										 
											2019-10-29 13:37:26 +08:00
										 |  |  | class TemplateMerge : public Optimizer { | 
					
						
							|  |  |  | public: | 
					
						
							|  |  |  |     virtual Cost onMeasure(const std::vector<VARP>& outputs, | 
					
						
							|  |  |  |                            std::shared_ptr<Parameters> parameters = nullptr) override { | 
					
						
							|  |  |  |         return Cost(); | 
					
						
							|  |  |  |     } | 
					
						
							| 
									
										
										
										
											2020-11-05 16:41:56 +08:00
										 |  |  |     bool onExecute(const std::vector<VARP>& outputs, | 
					
						
							|  |  |  |                    std::shared_ptr<Parameters> parameters = nullptr) override { | 
					
						
							|  |  |  |         return onExecute(outputs, PASS_PRIORITY_HIGH); | 
					
						
							|  |  |  |     } | 
					
						
							|  |  |  |     bool onExecute(const std::vector<VARP>& outputs, PassPriority priority); | 
					
						
							| 
									
										
											  
											
												- build:
	- unify schema building in core and converter;
	- add more build script for android;
	- add linux build script for python;
- ops impl:
	- add floor mod support in binary;
	- use eltwise impl in add/max/sub/mul binary for optimization;
	- remove fake double support in cast;
	- fix 5d support for concat;
	- add adjX and adjY support for batch matmul;
	- optimize conv2d back prop filter;
	- add pad mode support for conv3d;
	- fix bug in conv2d & conv depthwise with very small feature map;
	- optimize binary without broacast;
	- add data types support for gather;
	- add gather ND support;
	- use uint8 data type in gather v2;
	- add transpose support for matmul;
	- add matrix band part;
	- add dim != 4 support for padding, reshape & tensor convert;
	- add pad type support for pool3d;
	- make ops based on TensorFlow Lite quantization optional;
	- add all & any support for reduction;
	- use type in parameter as output type in reduction;
	- add int support for unary;
	- add variable weight support for conv2d;
	- fix conv2d depthwise weights initialization;
	- fix type support for transpose;
	- fix grad outputs count for  reduce grad and reshape grad;
	- fix priorbox & detection output;
	- fix metal softmax error;
- python:
	- add runSessionWithCallBackInfo interface;
	- add max nodes limit (1400) for visualization tool;
	- fix save error in python3;
	- align default dim;
- convert:
	- add extra design for optimization;
	- add more post converting optimizers;
	- add caffe v1 weights blob support;
	- add cast, unary, conv transpose support for onnx model;
	- optimize batchnorm, conv with variable weights, prelu, reshape, slice, upsample for onnx model;
	- add cos/sin/atan/tan support for unary for tensorflow model;
	- add any/all support for reduction for tensorflow model;
	- add elu, conv3d, pool3d support for tensorflow model;
	- optimize argmax, batchnorm, concat, batch to space, conv with variable weights, prelu, slice for tensorflow model;
- others:
	- fix size computer lock;
	- fix thread pool deadlock;
	- add express & parameters in express;
	- rewrite blitter chooser without static map;
	- add tests for expr;
											
										 
											2019-10-29 13:37:26 +08:00
										 |  |  | 
 | 
					
						
							|  |  |  |     static TemplateMerge& getInstance(const std::string& pass); | 
					
						
							|  |  |  | 
 | 
					
						
							| 
									
										
										
										
											2020-11-05 16:41:56 +08:00
										 |  |  |     void insertTemplate(std::string key, std::function<bool(EXPRP)> compare, std::function<bool(EXPRP)> transform, | 
					
						
							|  |  |  |                         PassPriority priority = PASS_PRIORITY_HIGH); | 
					
						
							| 
									
										
										
										
											2021-06-11 17:17:13 +08:00
										 |  |  |     void insertTemplateV2(std::string key, std::function<bool(EXPRP)> transform, PassPriority priority = PASS_PRIORITY_HIGH); | 
					
						
							| 
									
										
											  
											
												- build:
	- unify schema building in core and converter;
	- add more build script for android;
	- add linux build script for python;
- ops impl:
	- add floor mod support in binary;
	- use eltwise impl in add/max/sub/mul binary for optimization;
	- remove fake double support in cast;
	- fix 5d support for concat;
	- add adjX and adjY support for batch matmul;
	- optimize conv2d back prop filter;
	- add pad mode support for conv3d;
	- fix bug in conv2d & conv depthwise with very small feature map;
	- optimize binary without broacast;
	- add data types support for gather;
	- add gather ND support;
	- use uint8 data type in gather v2;
	- add transpose support for matmul;
	- add matrix band part;
	- add dim != 4 support for padding, reshape & tensor convert;
	- add pad type support for pool3d;
	- make ops based on TensorFlow Lite quantization optional;
	- add all & any support for reduction;
	- use type in parameter as output type in reduction;
	- add int support for unary;
	- add variable weight support for conv2d;
	- fix conv2d depthwise weights initialization;
	- fix type support for transpose;
	- fix grad outputs count for  reduce grad and reshape grad;
	- fix priorbox & detection output;
	- fix metal softmax error;
- python:
	- add runSessionWithCallBackInfo interface;
	- add max nodes limit (1400) for visualization tool;
	- fix save error in python3;
	- align default dim;
- convert:
	- add extra design for optimization;
	- add more post converting optimizers;
	- add caffe v1 weights blob support;
	- add cast, unary, conv transpose support for onnx model;
	- optimize batchnorm, conv with variable weights, prelu, reshape, slice, upsample for onnx model;
	- add cos/sin/atan/tan support for unary for tensorflow model;
	- add any/all support for reduction for tensorflow model;
	- add elu, conv3d, pool3d support for tensorflow model;
	- optimize argmax, batchnorm, concat, batch to space, conv with variable weights, prelu, slice for tensorflow model;
- others:
	- fix size computer lock;
	- fix thread pool deadlock;
	- add express & parameters in express;
	- rewrite blitter chooser without static map;
	- add tests for expr;
											
										 
											2019-10-29 13:37:26 +08:00
										 |  |  | 
 | 
					
						
							|  |  |  | private: | 
					
						
							|  |  |  |     TemplateMerge() { | 
					
						
							|  |  |  |     } | 
					
						
							| 
									
										
										
										
											2020-11-05 16:41:56 +08:00
										 |  |  |     std::vector<std::vector<std::string>> mPriorities; | 
					
						
							| 
									
										
										
										
											2021-06-11 17:17:13 +08:00
										 |  |  |     std::map<std::string, std::function<bool(EXPRP)>> mTemplates; | 
					
						
							| 
									
										
											  
											
												- build:
	- unify schema building in core and converter;
	- add more build script for android;
	- add linux build script for python;
- ops impl:
	- add floor mod support in binary;
	- use eltwise impl in add/max/sub/mul binary for optimization;
	- remove fake double support in cast;
	- fix 5d support for concat;
	- add adjX and adjY support for batch matmul;
	- optimize conv2d back prop filter;
	- add pad mode support for conv3d;
	- fix bug in conv2d & conv depthwise with very small feature map;
	- optimize binary without broacast;
	- add data types support for gather;
	- add gather ND support;
	- use uint8 data type in gather v2;
	- add transpose support for matmul;
	- add matrix band part;
	- add dim != 4 support for padding, reshape & tensor convert;
	- add pad type support for pool3d;
	- make ops based on TensorFlow Lite quantization optional;
	- add all & any support for reduction;
	- use type in parameter as output type in reduction;
	- add int support for unary;
	- add variable weight support for conv2d;
	- fix conv2d depthwise weights initialization;
	- fix type support for transpose;
	- fix grad outputs count for  reduce grad and reshape grad;
	- fix priorbox & detection output;
	- fix metal softmax error;
- python:
	- add runSessionWithCallBackInfo interface;
	- add max nodes limit (1400) for visualization tool;
	- fix save error in python3;
	- align default dim;
- convert:
	- add extra design for optimization;
	- add more post converting optimizers;
	- add caffe v1 weights blob support;
	- add cast, unary, conv transpose support for onnx model;
	- optimize batchnorm, conv with variable weights, prelu, reshape, slice, upsample for onnx model;
	- add cos/sin/atan/tan support for unary for tensorflow model;
	- add any/all support for reduction for tensorflow model;
	- add elu, conv3d, pool3d support for tensorflow model;
	- optimize argmax, batchnorm, concat, batch to space, conv with variable weights, prelu, slice for tensorflow model;
- others:
	- fix size computer lock;
	- fix thread pool deadlock;
	- add express & parameters in express;
	- rewrite blitter chooser without static map;
	- add tests for expr;
											
										 
											2019-10-29 13:37:26 +08:00
										 |  |  | }; | 
					
						
							|  |  |  | class TemplateMergeRegister { | 
					
						
							|  |  |  | public: | 
					
						
							| 
									
										
										
										
											2020-11-05 16:41:56 +08:00
										 |  |  |     TemplateMergeRegister(const std::string& pass, std::string key, std::function<bool(EXPRP)> compare, std::function<bool(EXPRP)> transform, | 
					
						
							|  |  |  |                           PassPriority priority) { | 
					
						
							|  |  |  |         TemplateMerge::getInstance(pass).insertTemplate(key, compare, transform, priority); | 
					
						
							| 
									
										
											  
											
												- build:
	- unify schema building in core and converter;
	- add more build script for android;
	- add linux build script for python;
- ops impl:
	- add floor mod support in binary;
	- use eltwise impl in add/max/sub/mul binary for optimization;
	- remove fake double support in cast;
	- fix 5d support for concat;
	- add adjX and adjY support for batch matmul;
	- optimize conv2d back prop filter;
	- add pad mode support for conv3d;
	- fix bug in conv2d & conv depthwise with very small feature map;
	- optimize binary without broacast;
	- add data types support for gather;
	- add gather ND support;
	- use uint8 data type in gather v2;
	- add transpose support for matmul;
	- add matrix band part;
	- add dim != 4 support for padding, reshape & tensor convert;
	- add pad type support for pool3d;
	- make ops based on TensorFlow Lite quantization optional;
	- add all & any support for reduction;
	- use type in parameter as output type in reduction;
	- add int support for unary;
	- add variable weight support for conv2d;
	- fix conv2d depthwise weights initialization;
	- fix type support for transpose;
	- fix grad outputs count for  reduce grad and reshape grad;
	- fix priorbox & detection output;
	- fix metal softmax error;
- python:
	- add runSessionWithCallBackInfo interface;
	- add max nodes limit (1400) for visualization tool;
	- fix save error in python3;
	- align default dim;
- convert:
	- add extra design for optimization;
	- add more post converting optimizers;
	- add caffe v1 weights blob support;
	- add cast, unary, conv transpose support for onnx model;
	- optimize batchnorm, conv with variable weights, prelu, reshape, slice, upsample for onnx model;
	- add cos/sin/atan/tan support for unary for tensorflow model;
	- add any/all support for reduction for tensorflow model;
	- add elu, conv3d, pool3d support for tensorflow model;
	- optimize argmax, batchnorm, concat, batch to space, conv with variable weights, prelu, slice for tensorflow model;
- others:
	- fix size computer lock;
	- fix thread pool deadlock;
	- add express & parameters in express;
	- rewrite blitter chooser without static map;
	- add tests for expr;
											
										 
											2019-10-29 13:37:26 +08:00
										 |  |  |     } | 
					
						
							|  |  |  | }; | 
					
						
							|  |  |  | } // namespace Express
 | 
					
						
							|  |  |  | } // namespace MNN
 |