What's Changed

• [Feature] Add the support of BallQuery op for Ascend device by @Ginray in #2963
• [Feature] Add the support of three_interpolate op for Ascend device by @lihao7212148 in #2962
• Update version info by @zhouzaida in #2966
• [Docs] Add api reference table by @fanqiNO1 in #2971
• [Fix] Fix deform_conv.py for torch_npu v2.1 by @shun001 in #2967
• DIOPI adapt codes for npu device by @CokeDong in #2973
• [Fix] Fix modulated_deform_conv.py for torch_npu v2.1 by @shun001 in #2983
• [Feature] Add NPU operator RotatedFeatureAlign by @huaweiZJX in #2994
• torch_npu support aclnn and add op by @momo609 in #2998
• [Feature] Add the support for stack_group_points ops with Ascend by @673958639 in #3023
• Fix ut of stack_grouping_points by @673958639 in #3026
• Add multi_scale_deform_attn op adapter for NPU. by @DaGaiBa in #3032
• add multi_scale_deform_attn_grad op adapter for NPU by @RRaoyzee in #3042
• Bump version to 2.2.0 by @HAOCHENYE in #3087
• add multi npu op by @momo609 in #3085

New Contributors

• @lihao7212148 made their first contribution in #2962
• @fanqiNO1 made their first contribution in #2971
• @shun001 made their first contribution in #2967
• @huaweiZJX made their first contribution in #2994

Full Changelog: v2.1.0...v2.2.0

😀 We are glad to announce that we are providing the mmcv-full pre-built packages for PyTorch 2.1.0 and 2.0.0 starting from MMCV v1.7.2.

pip install -U openmim
mim install mmcv-full

Features

• Support DCNv1 on Ascend device by @MiniTIckW in #2480
• Support PSAMask op for Ascend device by @DaGaiBa in #2487
• Support Voxelization with cambricon MLU device by @ZShaopeng in #2500
• Add Ascend support for RoIPool op by @momo609 in #2483
• Add MLU support for DCN by @mengpenghui in #2540
• Add support for Ascend devices with gather_points by @zcc-zjut in #2555
• Add ops bbox_overlaps by @enemy1205 in #2477
• Support ball_query with cambricon MLU backend and mlu-ops library. by @DanieeelLiu in #2520
• Add support for Ascend devices with nms_rotated by @ckirchhoff2021 in #2550
• Add Ascend support for RoIPoolGrad op by @xinlianglalala in #2569
• Add bbox_overlaps npu adapter by @dflhw in #2580
• ms_deform_attn performance optimization by @Wickyzheng in #2616
• Add the support of voxelization op for ascend device by @dflhw in #2613
• Make voxelization operator support the mlu290 platform. by @mahxn0 in #2652
• Add Ascend support for RoIAlign op by @momo609 in #2638
• Add MLU support for Sparse Convolution op by @duzekunKTH in #2589
• Add the support of arf op for ascend device by @dflhw in #2789
  *Sync replace by @defei-coder in #2750
• Add the implementation of diff_iou_rotated with mlu-ops by @Danielmic in #2852
• Add the implementation of dynamic_scatter with mlu-ops (#2847) by @Danielmic in #2853
• Add support of points_in_polyogns for Ascend device by @long11111111 in #2864
• Add the support of BoxIouRotated op for ascend device by @MrShadowY in #2854
• Replace the implementation of deform_roi_pool with mlu-ops by @defei-coder in #2598
• Add torch_npu optimizer by @luomaoling in #2753
• Add missing contiguous calls for nms_rotated by @allanzelener in #2547
• Support NmsRotated with cambricon MLU backend by @liuyuan1-v in #2643
• Add the implementation of box_iou_rotated with mlu-ops. by @ZhangLearning in #2703
• Modify adaptation of bbox_overlaps op for ascend device by @dflhw in #2697
• Support torch_npu v2.1 by @momo609 in #2943
• DvcliveLoggerHook: use by_epoch for determining when to log. by @daavoo in #2149
• Support MultiScaleDeformableAttention with AMP by @nijkah in **#2541

Enhancements

• Replace the implementation of three_nn with mlu-ops by @DanieeelLiu in #2719
• Repalce the implementation of roiaware_pool3d with mlu-ops. by @ZhangLearning in #2699
• Refactor the issue templates by @zhouzaida in #2585
• Remove unused files by @zhouzaida in #2586
• The argument scale of fused_bias_leakyrelu should be greater than 0 by @jayggh in #2623
• Replace carafe op of MLU backend with mlu-ops by @qipengh in #2818
• Replace the implementation of psa_mask with mlu-ops. by @DanieeelLiu in #2756
• Repalce the implementation of rotated_feature_align with ml… by @tudejiang79 in #2659
• Repalce the implementation of roi_align_rotated with mlu-ops by @tudejiang79 in #2657
• Simplify the logic of sparse_conv by @duzekunKTH in #2683
• Replace bbox_overlaps op of MLU backend with mlu-ops by @qirun-uiuc in #2880
• Replace focal_loss_sigmoid op of MLU backend with mlu-ops by @qirun-uiuc in #2877
• Replace roipoint_pool3d op of MLU backend with mlu-ops by @ClowDragon in #2876
• Replace tin_shift op of MLU backend with mlu-ops by @ClowDragon in #2911
• Add torch mluops check before call mluOpsxxx interface by @ClowDragon in #2874
• Refine the implementation of NMS op in Ascend device by @momo609 in #2658
• Optimize the performace of ms_deform_attn for MLU device by @Wickyzheng in #2510
• Update parrots image by @CokeDong in #2522
• Add AMP support for MLU_DCNv2 by @mengpenghui in #2548
• Use official isort by @triple-Mu in #2574
• Increase the dtype limit to maintain consistency with the gpu. by @momo609 in #2724
• Add a comment for the use of masked_conv on Ascend device by @zcc9707 in #2689
• Ignore mlu-ops files by @defei-coder in #2691
• Delete yapf verify by @zhouzaida in #2944
• Add a default value for MMCV_MLU_ARGS by @defei-coder in #2688

Bug Fixes

• Fix a typo in runner/hooks/evaluation.py by @jamiechoi1995 in #2523
• Fix c++/cuda lint failed by @HAOCHENYE in #2528
• Fix CI failed caused by 'Python.h' not found by @ice-tong in #2530
• Fix a dilation bug of MLU-DCNv2 and add limitation of torchvision by @mengpenghui in #2519
• Check the version of torchvision in init of DCN by @mengpenghui in #2556
• Fix typo in docstring by @HAOCHENYE in #2549
• Fix the parameter error of DeformableRoiPool in Ascend device by @dflhw in #2546
• Upgrade the version of isort to fix lint error by @zhouzaida in #2572
• Fix tensor descriptor setting in MLU ball_query. by @DanieeelLiu in #2579
• Fix a typo in the docstring of MSDeformAttn by @nijkah in #2599
• Fix the paper link error in the docstring by @BainOuO in #2618
• Prevent divide-by-zero error on Ascend device for bbox_overlaps by @dflhw in #2644
• Fix boxes.scalar_type in nms_npu by @momo609 in #2731
• Fix torch 1.8 adpater npu bug by @momo609 in #2970
• Fix npu optim by @luomaoling in #2766
• Keep the shape value of ascend iou op's parameter 2 smaller than parameter 1 by @dflhw in #2820
• Fix arf op's write conflict when num_orientations is not 1 by @dflhw in #2822
• Revert "[Fix] Fix arf op's write conflict when num_orientations is not 1" by @dflhw in #2826
• Fix deform_conv ops on Ascend NPU by @Ginray in #2805
• fix format error in roi_align_rotated by @ClowDragon in #2886
• Update mlu-ops version by @ClowDragon in #2922
• nms_rotated_npu accept angle parameter by @18251968697 in #2670
• Move voxel_size to the device where points are by @dflhw in #2681
• Fix parrots bbox_overlaps by @CokeDong in #2684

Others

• Refine the docstring of EvalHook by @Yan-Daojiang in #2576
• Steps to compile mmcv-full on MLU machine by @defei-coder in #2571
• Remove .owners.yml by @zhouzaida in #2712
• Build macos wheel for torch2.0.0 by @zhouzaida in #3008
• Bump version to v1.7.2 by @zhouzaida in #3007
• Update version info by @zhouzaida in #2981
• Replace the ubuntu-18.04 with 22.04 in CI by @zhouzaida in #2793

New Contributors

• @MiniTIckW made their first contribution in #2480
• @Wickyzheng made their first contribution in #2510
• @DaGaiBa made their first contribution in #2487
• @ice-tong ...

😀 We are glad to announce that we are providing the mmcv pre-built packages for PyTorch 2.1.0 starting from MMCV v2.1.0.

pip install -U openmim
mim install mmcv

Features

• Add the implementation of dynamic_scatter with mlu-ops by @Danielmic in #2847
• Add the implementation of diff_iou_rotated with mlu-ops by @Danielmic in #2840
• Add the support of points_in_polyogns for Ascend device by @long11111111 in #2848
• Add the support of BoxIouRotated op for ascend device by @MrShadowY in #2842
• Support modulated_deform_conv and deform_conv with cambricon MLU backend by @qipengh in #2823
• Add the support for group_points ops with Ascend by @673958639 in #2935
• Support torch_npu 2.1 by @momo609 in #2909
• Adapt new dipu device by @CokeDong in #2905
• Use PyTorch WARP_SHFL_DOWN macro for half support by @zstreet87 in #2843

Enhancements

• Support dipu_mock_cuda=False in dipu for mmcv ext ops with cpu fallback by @CokeDong in #2839
• Add torch mluops check before calling mluOpsxxx interface by @ClowDragon in #2871
• Change the order of condition to make fx wok by @youkaichao in #2883
• Rename fast_conv_bn_eval to efficient_conv_bn_eval by @youkaichao in #2884
• Replace focal_loss_sigmoid op of MLU backend with mlu-ops by @qirun-uiuc in #2855
• Replace bbox_overlaps op of MLU backend with mlu-ops by @qirun-uiuc in #2879
• Replace roipoint_pool3d op of MLU backend with mlu-ops by @ClowDragon in #2875
• Replace tin_shift op of MLU backend with mlu-ops by @ClowDragon in #2910
• Use unified interface in Ascend kernel by @momo609 in #2945

Bug Fixes

• Fix roi_align npu bug by @momo609 in #2862
• Fix format error in roi_align_rotated by @ClowDragon in #2887
• Fix data race risk of cache_randomness by @HarborYuan in #2927
• Update mlu-ops version by @ClowDragon in #2921
• Fix setup.py for torch_npu v2.1 by @momo609 in #2928
• Fix the support for nms_rotated in Ascend by @momo609 in #2931
• Fix modulated_deform_conv.py for torch_npu v2.1 by @6Vvv in #2941
• Temporarily disable mps ops for torch2.1.0 by @grimoire in #2958
• Update c++17 in setup.py by @grimoire in #2865
• Adapt to rocm by @xiabo123 in #2827

Others

• Install wheel package in CI to support bdist_wheel command by @zhouzaida in #2851
• Add torch2.1.0 checking in CI by @zhouzaida in #2955
• Bump version to 2.1.0 by @zhouzaida in #2959

New Contributors

• @673958639 made their first contribution in #2935
• @6Vvv made their first contribution in #2941

Full Changelog: v2.0.1...v2.1.0

Features

• Add fast_conv_bn_eval option in ConvModule for fast validation and training in Eval mode by @youkaichao in #2807
• Support mmcv ext with DIOPI impl by @CokeDong in #2790
• Add the support of arf op for ascend device by @dflhw in #2792
• Add the support for rotated_feature_align with MLU by @tudejiang79 in #2809
• Replace the implementation of roi_align_rotated with mlu-ops by @tudejiang79 in #2808

Enhancements

• Refactor the pytorch and cuda combination for testing in CI by @zhouzaida in #2693
• Remove two outdated files by @zhouzaida in #2740
• Update the supported torch and cuda versions in README by @zhouzaida in #2687
• Remove .owners.yml by @zhouzaida in #2743
• Enhance issue templates by @zhouzaida in #2776
• Raise an error if fails to read image by @Ezra-Yu in #2791
• Make build_xxx_layer allow accepting a class type by @HAOCHENYE in #2782
• Replace carafe op of MLU backend with mlu-ops by @qipengh in #2817
• Replace the implementation of psa_mask with mlu-ops. by @DanieeelLiu in #2810
• Simplify the logic of sparse_conv by @duzekunKTH in #2802
• Drop the outdated torch versions in CI by @zhouzaida in #2830
• Add ConvModule.turn_on_fast_conv_bn_eval to reduce repetitive code and dynamically bind conv during forward by @youkaichao in #2835

Bug Fixes

• ops can not be built in github action by @zhouzaida in #2783
• Build CUDA ops in CI by @zhouzaida in #2794
• Fix type hint in _scale_size and rescale_size by @MambaWong in #2799
• Fix the support for ROCm by @xiabo123 in #2811
• Keep the shape value of ascend iou op's parameter 2 smaller than parameter 1 by @dflhw in #2821
• Fix arf op's write conflict when num_orientations is not 1 by @dflhw in #2824
• Correct mmengine version requirement by @wateryzephyr in #2816
• Fix deform_conv ops on Ascend NPU by @Ginray in #2832

Documentations

• Describe branch changes by @zhouzaida in #2741
• Add npu build.md and update roi_pool_npu adpater by @momo609 in #2812
• small fix for RandomResize doc by @i-aki-y in #2829

New Contributors

• @MambaWong made their first contribution in #2799
• @i-aki-y made their first contribution in #2829
• @wateryzephyr made their first contribution in #2816

Full Changelog: v2.0.0...v2.0.1

Highlights

Introducing mmcv v2.0.0, we have made significant updates to the package. Training process-related modules, such as Runner, Hook, and Parallel, have been removed and are now provided by MMEngine. With this release, we have added numerous data transformations, including CenterCrop. Additionally, starting from mmcv v2.0.0, we are providing pre-built packages for PyTorch v2.0.0.

Features

• Introduced a variety of data transformations
• Added StyleGAN3 Ops (#2290)
• Implemented BezierAlign CUDA op (#2393)
• Added box_iou_quadri & nms_quadri (#2277)
• Introduced stack ball query and stack group points ops (#2292)
• Supported LayerScale in FFN (#2451)
• Added C++ implementation for bbox_overlaps (#2477)

Enhancements

• Adopted the new fileio interface from mmengine (#2468)
• Enabled MultiScaleDeformableAttention with AMP (#2541)
• Registered Compose to TRANSFORMS (#2600)
• Supported value_proj_ratio in MultiScaleDeformableAttention (#2452)

Backwards Incompatible Changes

• Removed mmcv.fileio (#2179). The new fileio interfaces can be found at mmengine.fileio
• Removed mmcv.runner, mmcv.parallel, mmcv.engine and mmcv.device (#2216)
• Removed all classes in mmcv.utils (eg Config and Registry) and many functions. Only a few functions related to mmcv are reserved (#2217). The new Config and Registry can be found at mmengine.registry and mmengine.config
• Removed mmcv.onnex, mmcv.tensorrt modules, and related functions (#2225)
• Dropped support for Python versions < 3.7 (#2525)
• Dropped support for PyTorch versions < 1.8 and CUDA versions < 10.1
• Renamed the package name mmcv to mmcv-lite and mmcv-full to mmcv. Also, changed the default value of the environment variable MMCV_WITH_OPS from 0 to 1 (#2235)

pip install openmim

# Contains ops, because the highest version of mmcv-full
# is less than 2.0.0, so there is no need to add version
# restrictions
mim install mmcv-full

# do not contain ops
mim install "mmcv < 2.0.0"

pip install openmim

# Contains ops
mim install mmcv

# Ops are not included, because the starting version of
# mmcv-lite is 2.0.0, so there is no need to add version
# restrictions
mim install mmcv-lite

Bug Fixes

• Improved compatibility of training stylegan2 (#2694)
• Replace torch/extension.h with torch/types.h to fix compilation error on Windows (#2698)
• Fixed torch2.0 dcn/mdcn symbolic (#2695)
• Skipped fused_lrelu op when gcc is less than 6.0 or CUDA is less than 10.2 (#2671)
• Skipped filtered_lrelu ut when CUDA is less than 10.2 (#2677)
• Called operator __half explicitly (#2608)
• Ensured fileio related interfaces maintain backward compatibility (#2539)
• Fixed CI failure due to protobuf version (#2529)
• Corrected _flip_keypoint result of RandomFlip (#2527)
• Fixed Windows OS version to 2019 (#2503)
• Resolved optimizer 'ValueError' when using PReLU activation (#2444)
• Fixed improper assertion in RandomChoiceResize (#2450)
• Fixed Correlation op (#2274)
• Fixed roi align symbolic for torch>=1.13 (#2443)
• Added swap_label_pairs in RandomFlip (#2332)
• Fixed MPS ops for torch1.12+ (#2302)
• Corrected calculation error of out_w in MaskedConv2dFunction (#2264)
• Cast mask type to enable training with amp (#2220)

Documentations

• Fixed torch2.0 dcn/mdcn symbolic (#2695)
• Skipped fused_lrelu op when gcc is less than 6.0 or CUDA is less than 10.2 (#2671)
• Skipped filtered_lrelu ut when CUDA is less than 10.2 (#2677)
• Called operator __half explicitly (#2608)
• Used official isort (#2575)
• Ensured fileio related interfaces maintain backward compatibility (#2539)
• Fixed CI failure due to protobuf version (#2529)
• Corrected _flip_keypoint result of RandomFlip (#2527)
• Fixed Windows OS version to 2019 (#2503)
• Resolved optimizer 'ValueError' when using PReLU activation (#2444)
• Fixed improper assertion in RandomChoiceResize (#2450)
• Fixed Correlation op (#2274)
• Fixed roi align symbolic for torch>=1.13 (#2443)
• Added swap_label_pairs in RandomFlip (#2332)
• Fixed MPS ops for torch1.12+ (#2302)
• Corrected calculation error of out_w in MaskedConv2dFunction (#2264)
• Cast mask type to enable training with amp (#2220)

Contributors

@HAOCHENYE @grimoire @zhouzaida @nijkah @BainOuO @enemy1205 @C1rN09 @triple-Mu @okotaku @vansinhu @Harold-lkk @KeiChiTse @xiexinch @xin-li-67 @VVsssssk @LYMDLUT @MeowZheng @yxzhao2022 @mzr1996 @ZwwWayne @YuanLiuuuuuu @ly015 @plyfager @GT9505 @zytx121 @hhaAndroid @fangyixiao18 @JingweiZhang12 @xcnick @zengyh1900

Features

• Support LayerScale in FFN (#2451)
• Add BezierAlign CUDA op (#2393)
• Pick Ascend ops from the master branch (#2501)

Improvements

• Support value_proj_ratio in MultiScaleDeformableAttention (#2452)
• Use the new interface of fileio from mmengine (#2468, #2539)

Bug fixes

• Fix optimizer 'ValueError' when using PReLU activation (#2444)
• Fix _flip_keypoint of RandomFlip returns a wrong result (#2527)

Documentations

• Update the supported PyTorch and CUDA versions (#2456)
• Add convention for the order of data transforms (#2442)
• Update qq group link (#2454)

Others

• Fix the version of Windows OS to 2019 in CI (#2503)
• Deprecate the support for python3.6 (#2525)
• Fix CI failed due to the version of protobuf (#2529)

Contributors

A total of 9 developers contributed to this release.
@zhouzaida @KeiChiTse @Harold-lkk @vansin @okotaku @grimoire @HAOCHENYE @CokeDong @ckirchhoff2021

New Contributors

• @KeiChiTse made their first contribution in #2444

Full Changelog: v2.0.0rc3...v2.0.0rc4

Features

• Support RoiawarePool3d, MultiScaleDeformableAttn, ModulatedDeformConv2d and NMS with cambricon MLU backend (#2359, #2396, #2411, #2401)
• Add Ascend npu adapter for FusedBiasLeakyRelu, DeformableRoIPool and NMS (#2474, #2481, #2387)
• Support receptive field search of CNN models (#2056)

Bug fixes

• Create Tensor with new_* method to support AMP (#2389)
• Fix DCN TensorRT plugin (#2408)
• Fix the bug when the input mask is not '0-1-tensor' in masked_conv (#2423)
• Fix UT caused by the new version of packaging (#2482)
• Change the block size to 1024 during compilation (#2460)
• Fix skip_layer for RF-Next (#2489)

Documentations

• Update supported pytorch versions (#2382)
• Add code style guide (#2406)
• Add currently-npu-supported-ops in ops.md (#2420)
• Refine contributing.md (#1941)
• Fix typos in contributing docs (#2462, #2485)
• Update qq group link (#2453)
• Replace NPU with Ascend in installation documentatio (#2469)

Others

• Add torch1.13 checking in CI (#2405)
• Update the URL of pyupgrade pre-commit hook (#2448)

Contributors

A total of 24 developers contributed to this release.
@ZhangLearning @grimoire @zhouzaida @cathyzhang222 @DerryHub @ckirchhoff2021 @wangjiangben-hw @zcc-zjut @momo609 @ZShaopeng @Wickyzheng @HAOCHENYE @gasvn @lzyhha @zytx121 @BayMaxBHL @sunyc11 @mengpenghui @RangeKing @vansin @Taited @jayggh @xiabo123 @CokeDong

New Contributors

• @ZhangLearning made their first contribution in #2359
• @DerryHub made their first contribution in #2408
• @gasvn made their first contribution in #2056
• @BayMaxBHL made their first contribution in #2448
• @sunyc11 made their first contribution in #2459
• @mengpenghui made their first contribution in #2411
• @Taited made their first contribution in #2485
• @jayggh made their first contribution in #2474
• @xiabo123 made their first contribution in #2460
• @lzyhha made their first contribution in #2489

Full Changelog: v1.7.0...v1.7.1

Features

• Add stack ball query and stack group points ops (#2292)
• Support RoIAwarePool3d, Iou3d and MultiScaleDeformableAttn with MLU backend (#2359, #2339)
• Support NMS with MLU590 backend (#2401)

Improvements

• Adapt mlu code to cntoolkit3.0.1

Bug fixes

• Fix roi align symbolic for torch>=1.13 (#2443)
• Create Tensor with new_* method to support AMP (#2389)
• Fix three nn op can not accept half tensor (#2348)
• Fix the potential NaN bug in calc_square_dist (#2356)
• Fix Correlation op (#2274)
• Fix a not proper assertion in RandomChoiceResize (#2450)

Documentations

• Make installation commands compatible with windows CMD (#2353)
• Add macOS building notice (#2364)
• Refine contributing.md (#1941)
• Add code style guide (#2406)

Others

• Add torch1.13 checking in CI (#2441)

Contributors

A total of 13 developers contributed to this release.
@LYMDLUT @LYM-fire @zhouzaida @VVsssssk @xin-li-67 @grimoire @ZCMax @ZhangLearning @cathyzhang222 @HAOCHENYE @defei-coder @ZShaopeng @xiexinch

New Contributors

• @LYMDLUT made their first contribution in #2353
• @xiexinch made their first contribution in #2450

Full Changelog: v2.0.0rc2...v2.0.0rc3

Highlights

Starting from v1.7.0, we are providing pre-built packages for PyTorch v1.13.0, as well as for the macOS platform. Note that, for the mac M1 chip, only the pre-built packages on PyTorch v1.13.0 are available.

pip install -U openmim
mim install mmcv-full

Features

• Support training on Ascend device (#2262, #2370)
• Add box_iou_quadri & nms_quadri (#2277)
• Add stack ball query and stack group points ops (#2292)
• Support PrRoI op for Parrots (#2280)
• Support Iou3d with cambricon MLU backend (#2339)
• Add getJobLimitCapability interface and use it in nms (#2337)

Improvements

• Support SiLU with torch < 1.7.0 (#2278)
• Update onnx registry (#2294)
• Adapt mlu code to cntoolkit3.0.1 (#2248)
• Refactor the nms3d op to support MLU (#2296)
• Add ability to pass logger instance to frameworks (#2317)
• Add a new argument define_metric in wandb hook (#2237)

Bug fixes

• Fix warning of CUDA ops (#2324)
• Fix the bug of uninitialized prarmeter (#2345)
• Fix three nn op can not accept half tensor (#2348)
• Fix Correlation op (#2274)
• Fix the potential NaN bug in calc_square_dist() (#2356)
• Fix MPS ops for torch1.12+ (#2302)

Documentations

• Refine README and installation documentations (#1977)
• Update README and fix typo in docs (#2358)
• Add steps to install mmcv-full on Ascend machine (#2363)
• Add maintenance plan and print deprecated message (#2357)

Others

• Upgrade pre commit hooks (#2321)
• Build macOS wheel in Github Action (#2311, #2366)

Contributors

A total of 14 developers contributed to this release.
@okotaku @zhouzaida @grimoire @wangjiangben-hw @ckirchhoff2021 @tpoisonooo @defei-coder @budefei @CokeDong @shcheklein @cathyzhang222 @ZCMax @zytx121 @momo609

New Contributors

• @tpoisonooo made their first contribution in #2324
• @shcheklein made their first contribution in #2317
• @cathyzhang222 made their first contribution in #2339
• @momo609 made their first contribution in #2363

Full Changelog: v1.6.2...v1.7.0

😀 We are glad to announce that we are providing the mmcv pre-built packages for macOS (x86_64 architecture).

pip install -U openmim
mim install 'mmcv>=2.0.0rc2'

Features

• Add TTA transform (#2146)
• Add two CUDA ops, box_iou_quadri and nms_quadri (#2277)
• Add two MLU ops, RoIPointPool3d and Deformable RoIPool (#2247, #2137)
• Register a new activation layer SiLU to ACTIVATION_LAYERS (#2256, #2278)

Bug fixes

• Fix the calculation error of out_w in MaskedConv2dFunction (#2264)
• Fix MPS ops for torch1.12+ (#2302)
• Add swap_labe_pairs in RandomFlip (#2332)

Documentations

• Fix some mistakes in data_process.md (#2283)
• Refactor the API structures of documentation (#2254)
• Add a reminder in docs for macOS M1 users (#2268)
• Translate data_transform.md to English (#2325)
• Refine README and installation documentation (#2340)
• Make installation commands compatible with Windows cmd (#2353)

Others

• Update configs of CircleCI and GitHub Action (#2255)
• Build macOS wheels in Github Action (#2350)
• Upgrade the versions of pre-commit hooks (#2328)
• Remove onnx-related files (#2295)

New Contributors

• @yxzhao2022 made their first contribution in #2283
• @xin-li-67 made their first contribution in #2268
• @LYMDLUT made their first contribution in #2353

Contributors

@yxzhao2022 @LYMDLUT @grimoire @xin-li-67 @zytx121 @MeowZheng @HAOCHENYE @zhouzaida @ZShaopeng @defei-coder @okotaku

Full Changelog: v2.0.0rc1...v2.0.0rc2