Release Note 1.6.0

Adapt X2Paddle to PaddlePaddle 3.0, and the test cases of test_benchmark can be verified under:

• python == 3.9
• paddlepaddle == 3.0.0beta1
• tensorflow == 2.16.1
• onnx == 1.17.0
• torch == 2.4.1

What's Changed

• [Bug] Fixed onnx Pad op with length -1. by @megemini in #1058
• [Bug] Fixed version check for PaddlePaddle. by @megemini in #1064
• [Bug] Fixed pytorch op mapper aten_linear. by @megemini in #1079
• [Bug] Fixed onnx unknow dimension. by @megemini in #1082
• [Bug] Fixed the usage of easyocr.Reader for 1.7.2. by @megemini in #1084
• [Bug] Fixed tensorflow manually updates shapes. by @megemini in #1089
• Update Dockerfile. by @megemini in #1061 @luotao1 in #1105
• Update LOD_TENSOR to DENSE_TENSOR. by @co63oc in #1095
• Update x2paddle usage of paddle.fluid for PaddlePaddle 3.0 by @megemini in #1087
• Update x2paddle usage of paddle.jit.to_static with AST mode for PaddlePaddle 3.0. by @megemini in #1093
• Update test_benchmark compatiblity for PaddlePaddle 3.0 by @megemini
  • Remove paddle.fluid and update the usage of paddle.static.load_inference_model in #1058 #1065 #1067 #1068 #1069 #1070 #1071 #1073 #1075 #1076 #1078 #1079 #1080 #1082 #1083 #1084 #1089 #1091
  • Update onnx one_hot type promotion compatiblity for PaddlePaddle 3.0. by @megemini in #1082
  • Update onnx nms3 api compatiblity for PaddlePaddle 3.0. by @megemini in #1082
  • Update onnx roi_align api compatiblity for PaddlePaddle 3.0. by @megemini in #1082
• Update docs for PaddlePaddle 3.0 by @megemini in #1088 #1098 #1101 #1103
• Add new models in test_benchmark by @megemini
  • pytorch GPT2 #1085
  • pytorch MT5_small #1086
  • pytorch YOLOv5 #1099
• Add pytorch aten::pad and fix related test_benchmark test cases. by @megemini in #1073
• Add pytorch aten::list and fix related test_benchmark test cases. by @megemini in #1075
• Add pytorch aten::aten_linalg_vector_norm and fix related test_benchmark test cases. by @megemini in #1076
• Add pytorch aten::scaled_dot_product_attention and fix related test_benchmark test cases. by @megemini in #1078
• Add pytorch aten::unsafe_chunk and fix related test_benchmark test cases. by @megemini in #1080
• Add black.list in test_benchmark to control CI. by @megemini in #1061
• Remove Python 2.0 support for raw_input. by @megemini in #1083

What's Changed

• [Bug]Fixed onnx op_mapper for PNC model. by @wjj19950828 in #962
• [Bug]Fixed some typo. by @co63oc in #976 #977 #978
• [Bug]Fixed pytorch aten::squeeze for CHAN-DST model. by @wjj19950828 in #815
• Add onnx cast op in the LN pattern. by @wjj19950828 in #822
• Add pytorch aten::_convolution_mode. by @wjj19950828 @luotao1 in #1041
• Add pytorch aten::amax and aten::topk by @Ken-reddevil in #1019
• update CI Checks by @luotao (Release 1.5.0)
  • Add 100+ test_benchmark demos (Caffe/TensorFlow/ONNX/Pytorch). in #1014 #1020 #1023 #1025 #1026 #1027 #1028 #1029 #1030 #1031 #1032 #1033 #1034
  • Update test_autoscan unit-tests (Pytorch/ONNX). in #1035 #1036
  • update CodeStyle checks. in #1046 #1047
  • update Dockerfile. in #1022

New Contributors

• @co63oc made their first contribution in #976
• @luotao1 made their first contribution in #1014
• @Ken-reddevil made their first contribution in #1019

What's Changed

• [Bug]Fixed ONNXDecoder by @wjj19950828 in #939
• Add ONNX2Paddle for visualdl by @rainyfly in #954
• [Bug]Support YOLOv8-cls by @wjj19950828 in #952

New Contributors

• @rainyfly made their first contribution in #954

What's Changed

• Add input_shape_dict and rm raw_input by @wjj19950828 in #899
• Fixed ToPILImage && rm SymbolicShapeInference by @wjj19950828 in #903
• Support GFPGAN by @wjj19950828 in #906
• Add FastDeploy YOLOv5/v6/v7 demo in readme by @wjj19950828 in #911
• Updated torch.nn.utils.clip_grad_value_.md by @PommesPeter in #918
• [Bug] Fixed ONNX Gemm bug by @wjj19950828 in #917
• [Bug] Fixed typo error by @wjj19950828 in #920
• Support fmod=1 in Mod OP by @wjj19950828 in #921
• Fixed NonZero bug by @wjj19950828 in #922
• add optim by @AndPuQing in #927
• Support YOLOv8 for dynamic shape by @wjj19950828 in #931

New Contributors

• @PommesPeter made their first contribution in #918
• @AndPuQing made their first contribution in #927

What's Changed

• Support 7-15 ONNX opset version by @wjj19950828 in #838
• Modify opset7~15 inheritance relationship by @wjj19950828 in #843
• Add autoscan test by @wjj19950828 in #836
• Add autoscan test for Elementwise ops by @wjj19950828 in #854
• Add_test_abs by @qqj1130247885 in #859
• Add logical and compare ops by @wjj19950828 in #855
• Fixed torch pad by @wjj19950828 in #863
• Solve the overload problem by @wjj19950828 in #865
• Add AveragePool tests by @wjj19950828 in #872
• Add Reduce_ops and HardSigmoid by @qqj1130247885 in #867
• Add isinf by @qqj1130247885 in #868
• Add IsNaN by @qqj1130247885 in #869
• Add Sum by @qqj1130247885 in #873
• Fixed Readme by @wjj19950828 in #879
• Fixed reduce tests bug by @wjj19950828 in #880
• Set enable_code_optim default to False by @wjj19950828 in #889

What's Changed

• Add HuggingFace Model Conversion Tutorial by @wjj19950828 in #823
• Add aten::upsample_trilinear3d by @wjj19950828 in #817
• Fixed misspell by @wjj19950828 in #821
• Add sysmpy in requirment by @wjj19950828 in #824
• Fixed Misspell by @wjj19950828 in #829
• Fixed rename bug by @wjj19950828 in #820
• Fix ONNX resize op by @qqj1130247885 in #830
• Fix bug in docs directory about code combination of save_image by @ZoruaSama in #833
• Remove scale for add by @wjj19950828 in #835

New Contributors

• @qqj1130247885 made their first contribution in #830
• @ZoruaSama made their first contribution in #833

Full Changelog: v1.3.7...v1.3.8

新增功能

• 全面支持 HuggingFace 模型，包括Bert、Albert等热点模型以及对应不同任务类型，共55种 TorchScript 以及40种 ONNX 模型
• 针对 Bert结构 ONNX模型，新增 Gelu pass，预测速度提升20%以上 #809
• 针对 Bert结构 ONNX模型，新增 div2scale pass，使其命中PaddleInference multihead_matmul_fuse_pass，性能提升30%以上 #813

新增模型支持

PyTorch->Paddle

• MobileNetv3
• HuggingFace T5
• HuggingFace Marian
• HuggingFace Bart
• HuggingFace wav2vec2
• HuggingFace BigBird

ONNX->Paddle

• HuggingFace Roberta
• HuggingFace XLM-Roberta
• HuggingFace T5
• HuggingFace GPT2
• HuggingFace Marian

新增op

• 新增PyTorch op（7个）：aten::rsqrt、aten::replication_pad1d、aten::full、 aten::group_norm、aten::argmax、aten::copy、aten::linear #783 #781 #786 #791 #789 #769
• 新增ONNX op（2个）：CumSum 、LessOrEqual #793 #794

修复op

• 修复PyTorch op（2个）：aten::masked_fill、aten::index #786 #787
• 修复ONNX op（12个）：ConvTranspose 、Unsqueeze 、MatMul 、Gather 、Slice 、Squeeze 、Flatten、Constant 、ConstantOfShape 、Where 、Topk、NonZero #792 #794 #796 #797 #805 #808

修复功能

• 修复LN fuser pass #777
• 修复Linear fuser pass #800

其他

• 在转换完成之后添加问卷链接 #801
• 清除paddle.fluid接口 #806

新增功能

• 新增转换记录统计功能 #766
• 针对 ONNX Transformer 模型，新增 LayerNorm pass，优化预测速度 #761
• HuggingFace 参数迁移功能 #748

新增模型支持

PyTorch->Paddle

• 图像修复算法：Saicinpainting_LaMa

ONNX->Paddle

• SOTA backbone：SwinTransformer
• 关键点检测算法：Blazepose

Caffe->Paddle

• 视频分割算法：SegFlow

TensorFlow->Paddle

• 图像生成算法：Pix2Pix

新增op

• 新增PyTorch op（5个）：aten::complex、aten::real、aten::imag、aten::fft_rfftn、aten::fft_irfftn #729
• 新增ONNX op（5个）：Sin、Cos、Neg、SpaceToDepth、GatherElement #723
• 新增PyTorch-Paddle api（2个）：gather、scatter #726 #730

修复op

• 修复PyTorch op（1个）：aten::_convolution #725
• 修复ONNX op（7个）：pad、resize、Upsample 、Gather 、Slice 、convtranspose 、Split #716 #720 #722 #723 #749 #751
• 修复Caffe op（3个）：scale、crop、Pow #721 #757
• 修复TensorFlow（3个）：FusedBatchNorm、SpaceToBatchND、BatchToSpaceND #738 #741

修复功能

• 优化转换过程中的 log 提示 #764
• 更新外部框架一键转换为 Paddle Lite 模型文档，新增lite_valid_places选择类型 #758
• PyTorch2Paddle 动态shape转换指引 #732

Contributors

• Thanks to external developers for their contributions to open source projects！@geoyee @HUSTLX

新增功能

• 添加enable_code_optim开关，解决部分场景下代码优化可能导致转换出错的问题 #696

新增模型支持

PyTorch->Paddle

• 首次支持声音克隆模型：MockingBird
• 轻量级检测算法：YOLOX

新增op

• 新增PyTorch op（5个）：aten::repeat_interleave、aten::maxpool1d、aten::frobenius_norm、aten::silu、aten::format #691 #696
• 新增ONNX op（1个）：GatherND #645

修复op

• 修复PyTorch op（5个）：aten::div、aten::index_select、aten::select、aten::upsample_nearest2d、aten::_convolution #691 #689
• 修复ONNX op（3个）：Split、Conv、Pad #645 #707 #687
• 修复PyTorch训练代码转换（2个）：ToPILImage、Split #690 #694

修复功能

• Windows环境生成代码问题 #706
• 修复ONNX和PyTorch版本问题 #708
• 去除代码优化部分to_tensor操作，优化预测速度 #696

Contributors

• Thanks to external developers for their contributions to open source projects！@littletomatodonkey @zhoucz97 @firedent @dawn1206

修复功能

• 修复X2Paddle调用ResNet以及Vgg预训练模型bug(#684)