v0.9.0: Merge pull request #71 from JDAI-CV/fix_exception_in_ort

Changes:

• e17f11e Merge pull request #71 from JDAI-CV/fix_exception_in_ort
• 4efa112 Merge branch 'master' into fix_exception_in_ort
• a3bfb68 catch exceptions in GetSupportedNodes so that ort nnapi ep will not crash here
• bac3cce Merge pull request #70 from JDAI-CV/add_ops
• e47f9df Add three ops: sub, exp and abs, and update some code
• af84537 Merge pull request #69 from JDAI-CV/ort_ci
• c2dbf71 Revert "Intentionally use a wrong floor op impl to test the ci"
• e59a642 Intentionally use a wrong floor op impl to test the ci
• ac0e086 use daquexian android ort branch
• 656180e Fix ort source dir

See More

• dd96d39 Fix ci yml
• 49ead95 Add ONNX Runtime CI
• ad11c32 Merge pull request #67 from JDAI-CV/update_script_add_new_ops
• 2980add Fix nnapi Bad data error in api < 29 conv_2d
• f64b511 Support nnapi inputs that are only available in some api levels, update conv_2d
• 5aee023 Remove unused file
• 79ac7cb delete copy constructor and copy assignment op of Model according to the rule of three
• 0ff3719 Use strong typed FuseCode all the places
• f779d6c Refactor code about fuse code
• 36c15fb Split fbs param into input param and output param, for better fitting NNAPI
• 5bd1c19 Remove some throw in mbimpl.cpp
• b188392 Change some methods of daq reader to return expected
• a63722c clang-format ex_model_builder.cpp
• 4cd010e Update api names in OnnxConverter
• 512b539 Make AddLayer_* in model builder return expected object instead of throw exceptions on error
• 594ee33 Update the api names in ex_model_builder.cpp
• 3500f1a Add blank lines between generate methods in MBImpl.cpp
• d8b8081 Merge branch 'master' into update_script_add_new_ops
• 4fad420 Rename AddLayer* to AddLayer_* and WriteDaqLayer_*
• eb620ee Rename the misleading attr name in ops.yml
• 3c82a70 introduce expected.hpp
• 66f9a2b Fix the unused buffer bug
• 1c5cee0 Relax the cmake version for android ndk
• a5d5627 Relax the cmake version for android ndk
• e97b9db Make AddLayer*Impl private
• fa28ade ModelBuilder AddLayer* return void now
• 96e4721 Support prelu before android 29
• 19f7f49 Rename Add* to AddLayer* in ModelBuilder
• 8b547c2 Rename AddTensorFromBuffer to ...Persistent..., AddTensorFromBuffer now copies the data
• 37e92bd Add version field and refuse old models
• c7976bb Fix wrong kernel shape of global pooling
• 87466ec make ci log less verbose
• 0895bb4 'converter_simple' should not exist, all related logic should be implemented in Convert()
• f46f008 fix wrong check and re-format onnxconverter.cpp
• 45b0c1b Merge branch 'master' into update_script_add_new_ops
• 4e6078c Make most boilderplate code auto-generated, change many layer names, old daq models will not work any more
• 47caec4 Fix the wrong axis order, try to move all logic into onnxconverter Convert()
• e0dd8ab Add compile_fbs() in generate_code.py, change shaper of min/maximum to eltwise
• e637232 Add the missing auto-generated code
• f498eac Remove "=0" since default value is 0
• 4041d0b Update generate_code.py to generate daq.fbs
• 77520e1 Ignore .clangd and venv/
• b72182e Add some new ops in ops.yml
• 9fdc96f Add a little comment in ops.yml
• dbb8e3f Improve some header include path
• 8370c6f Remove out-dated benchmark in README
• 3c77c30 Merge pull request #65 from JDAI-CV/update_readme
• 0d9e665 Update benchmark in README
• 6612d1e Install cmake 3.14.4 in CI
• 647d4c3 Refuse to build DNNLibrary with cmake 3.15
• 2d48e27 Update PRs welcome badge style
• 2d4b0f0 check api in isnodesupported()
• 52abfa8 Merge pull request #63 from JDAI-CV/update
• c969332 Support sigmoid
• e17f618 Check Android SDK in runtime instead of compile time
• 156eab7 Generate daqreader layer_type_to_str in generate_code.py
• 051b1be The missing rename in generate_code.py
• abcc3b5 not .. in -> .. not in
• 70813c6 format generate_code.py
• 7e70872 Merge pull request #62 from JDAI-CV/update
• 589611c Move template function
• 5f76678 Remove unused function declaration in ModelBuilder.h
• d093f6b Add missing code in DaqReader
• 2719734 Move op implementation of OnnxConverterImpl into OnnxConverterImpl.cpp
• b5dd437 Support two simple op: tanh and floor
• 77fa973 Move op implementation to ModelBuilderImpl.cpp
• b7073db Add comment for cmake 3.15 workaround
• c3cd873 Merge pull request #61 from JDAI-CV/update
• 29a6a99 have no idea why cmake_cxx_standard doesn't work suddenly
• b80ab3c getDevices -> GetDevices
• 812bbe0 Fix std::nullopt
• 4923e9f Add getDevices()
• cf9c566 Remove redundant shaper functions

This list of changes was auto generated.

v0.7.3: Merge pull request #60 from JDAI-CV/update_wrapper

Changes:

• 848b29f Merge pull request #60 from JDAI-CV/update_wrapper
• 4168162 Update nnapi wrapper from tensorflow
• ab22710 Merge pull request #59 from JDAI-CV/check_conv_weight
• a724168 Use shape inference to check concat, softmax and global pooling
• ecc612d Add more check for pooling, batchnorm and identity
• c28e516 getsupportednodes.cpp building -- fix cmake
• 73e4e89 Fix the shape of one-element tensor
• eff13f6 Add rank check in convert func of onnx converter
• 4a4ded8 Merge pull request #58 from JDAI-CV/check_conv_weight
• 85cad29 Add getsupportednodes tool, check for single reshape/dropout

See more

• 79edcc3 Add softmax axis check
• 8501d9a fix depth mulitiplier of dw conv
• 2f08809 conv2d and pooling2d
• b1a25f6 make strkeymap has const, check prelu slope
• 51f2779 check whether onnx convolution weight exists in initializer
• 65b14c9 check api in runtime and make AllowFp16 accessible in api 27
• a804ce1 remove src folder
• efa45e7 make log less verbose
• b5739d5 Update dnn_benchmark to support onnx
• b9584fe Support non-rank-4 input in daqreader
• 1ec5eaa Add the missing throw in dnn_retrieve_result.cpp
• 90cb0f8 Merge pull request #57 from JDAI-CV/support_onnx_model_output
• a75548d Considering identity for output
• 6bbf563 support onnx model output
• 99baa7b Merge pull request #56 from JDAI-CV/3rd_party
• 73f7526 Move flatbuffers to submodule

This list of changes was auto generated.

v0.7.2: Merge pull request #55 from JDAI-CV/support_onnx_input

Changes:

• cf20ec9 Merge pull request #55 from JDAI-CV/support_onnx_input
• 550c68f Fix gemm weight conversion
• 7c1d1b9 Fix eltwise shaper
• 38dd3bd Support onnx input (which can be read from both output and initializer)
• 8cab557 Fix warnings
• b33f375 Merge pull request #54 from JDAI-CV/getsupportednodes
• 869d560 Add GetSupportedNodes()
• 9120998 Add IsNodeSupported()

This list of changes was auto generated.

v0.7.1

Changes:

• 2a705eb Apply clang-format
• 8ded267 Add missing headers
• 3ac1034 Remove android header in onnx2daq
• 1f10295 Update the check of auto_pad
• 4d918b9 Fix bug that input width and height are reversed
• 276f2dd Use dnn::optional instead of std::optional and nonstd::optional
• 59646a6 Update path in generate_code.py
• b5d6707 Fix bug in dnn_retrieve_result.cpp
• 7f5d2f7 Use protobuf-lite in onnx2daq
• 013c6a2 Use protobuf-lite instead of protobuf

See more

• 4fe4b42 Add DNN_CMAKE_INSTALL for disabling cmake install in onnxruntime
• 52ad7de Add guard to not add_subdirectory when the target exists
• 52da80a Add missing DNN_READ_ONNX guard
• 1014a45 Merge pull request #53 from JDAI-CV/export
• 7f794cf Fix alias and destination
• 5ab120f Add alias the same with exported target name
• 2f96e4c Add destination for installing and fix jni
• a6fd498 Add dnn:: namespace
• 6ea82ea Update Config.cmake.in and so on
• 34b5fa2 Reorg the directory structure for exporting, also remove helper.h in
• 51b19b2 Auto set python package version from git tag

This list of changes was auto generated.

v0.7.0

Changes:

• bf69edf Auto generate versionCode in android build.gradle
• 9b428af Avoid name conflict in glog.cmake. BUILD_TESTING is conflicted with eigen in onnxruntime
• 4f12fe1 Merge pull request #52 from JDAI-CV/ci
• 3ca03f7 Update README badge
• a1a295d Skip windows ci
• de62356 Add all-27 configuration
• e1522d5 Fix wrong path of libdaq-jni.so
• 7e86fd8 Fix undefined AllowFp16 in Android 27
• f2fd69a Fix wrong cmake args, delete travis and appveyor ci
• 9a057f3 Fix a bug found by CI: resnetv2 fails

See more

• 748ddd8 Install onnx in ci
• faaf1d2 Disable werror for binaries when bundled protobuf is used
• 2cf973c Fix unused parameter in dnn_benchmark.cpp
• 3d13eab Disable treat warnings as errors for dnnlibrary when reading onnx
• e4ef696 Not build protoc on Android
• c13a6e7 Fix protoc path
• 5182c6d Fix protoc path
• fd53321 Download protoc in ci
• 202aefb fix msvc flag
• db6c783 x86_64 emulator
• e835863 Set github connection
• cdede50 Set android api for dnnlibrary build
• 5376b8a Edit ci files to fit dnnlibrary
• bd9bfd1 rename dabnn to dnnlibrary, onnx2bnn to onnx2daq
• ce1a521 azure pipeline ci init -- move files from dabnn
• a645f98 Merge pull request #47 from JDAI-CV/python_package
• 26af0fb Remove redundant find protobuf
• 3d8fc1e Build python only when DNN_BUILD_PYTHON is on, set PIC to on and misc changes
• cd6249e Update
• cb448e7 Fix bug of cmake
• 845d3e6 Move binary file output of Release/Debug in Windows
• 8cb4433 Set global msvc flag
• 6662958 Update .travis.yml for multiple env var
• 9841b91 Ignore .eggs
• 9288d85 Disable python 3.7 wheel build since onnx doesn't provide 3.7 wheel
• a34c395 Update ONNX submodule
• 5adeeb7 Update ONNX submodule
• 86e4686 Update ONNX submodule
• 06d7e58 Skip 32bit builds, since ONNX Runtime doesn't provide 32-bit package
• 416ea42 Fix ONNX_IMPORT
• db2bc51 Remove empty ONNX_PROTO_SRCS and headers
• c0ce646 Fix ONNX_IMPORT
• 1ef0148 Only set ONNX_BUILD_MAIN_LIB to onnx_proto
• d6a454d Add ONNX_BUILD_MAIN_LIB
• 6a0cae1 Add definitions for protobuf dll
• 6d04b45 Handle the build type
• eaeb3a4 Remove /WX in ONNX
• 262378b Use vs 2017
• eae6eda Alias for protobuf targets, workaround for protobuf 3.5.2 source
• 4f77ab4 Set protobuf git to 3.5.2 for windows ci
• 2aff852 Set conda path
• 54011e2 Move all options together
• a1f4e8b Upgrade protobuf to 3.7.1 release, set ONNX to 1.4.1 release
• fac3159 add cmake to install_require
• 0bc5e7f Only find protobuf when using system protobuf
• c5b0595 Use custom cibuildwheel in .travis.yml
• d1a03a2 Revert "Move python files a same level with c++ files"
• 16abae0 Fix appveyor submodule
• c80d8a5 Update new cibuildwheel
• afb603b Move python files a same level with c++ files
• edb0adc Update .travis.yml and appveyor.yml
• 65ac1a4 Update .travis.yml to test cibuildtools
• 80bd1e2 Update wrapper python files, fix wrong extension module name, add optional table_file argument.
• b51bf28 python package
• b0cad5b Merge pull request #51 from JDAI-CV/unify_onnx2daq
• 82ff1b5 fix wrong naming style in java wrapper
• 1f0c031 Update java wrapper for new get size
• 43e35f1 Fix wrong naming style in dnn_benchmark and apply clang-format
• 765d546 Determine output name automatically in dnn_benchmark
• cced7ed Update dnn_benchmark for to get size
• 3d21350 Disable read onnx by default
• afdec77 Revert some wrong editions in NeuralNetworksWrapper.h
• 8f05911 Fix bug in validate_onnx.py -- support non 4d input
• f2fff56 A more powerful validate_onnx.py. Only dir and dnn_retrieve_result is
• 5c514fc read onnx by default in validate_onnx.py
• 4150702 Fix wrong path in validate_onnx.py
• 896065b Add option to generate NCHW result in dnn_retrieve_result
• 76d1f8b Support multiple inputs in validate_onnx.py
• 7e7346c Import argh to handle cli args in dnn_retrieve_result
• eeeec2d Update dnn_retrieve_result to receive multiple inputs, update
• cbe6128 Update validate_onnx.py to not set output name
• fd17f22 Only delete .daq model
• 39ebfea Not set output explicitly in dnn_retrieve_result
• b0ff867 Automatically find outputs
• 5e9fed5 Apply clang-format to onnxconverter
• 42c8044 Generate code
• 0a93f71 Prepare to generate code about imm_blob_inputs_ and imm_blob_outputs_
• 1ed7bb2 Prepare to generate code about tensor_inputs_ and tensor_outputs_
• 2c00e2e Unify onnx2daq
• 60ad90b Add separated Convert, Save and GetBuf
• 3ec2638 First commit for directly onnx reading, at least it compiles
• 3191182 Update dnn_retrieve_result to receive multiple inputs, update
• 469bd79 Add DNN_ASSERT_EQ
• 3c7e7f8 Add const qualifier in some functions
• a22c6ed Support batch input in model builder
• da2565d Update AppImage
• 329df55 Update LICENSE
• c99b4a0 Upgrade onnx to v1.5.0 and protobuf to v3.7.1
• 78af419 Remove AS header in cmake

This list of changes was auto generated.

0.6.19

Fix bug in publish aar script

0.6.16

Deploy aar to bintray in ci, update usage message in onnx2daq

0.6.15: Move build_aar.sh to scripts/, add build_appimage.sh,

deploy appimage in ci

Please download from jcenter directly

No more extra aar file will be uploaded. Please download from jcenter directly :)

v0.2.2

Update version in README.md

Former-commit-id: 1abc4f339728d56eb5786c5032093a59a5e6ebd1