Mmpretrain

Highlights

- Support Tokens-to-Token ViT backbone and Res2Net backbone. Welcome to use!
- Support ImageNet21k dataset.
- Add a pipeline visualization tool. Try it with the [tutorials](https://mmclassification.readthedocs.io/en/latest/tools/visualization.html#pipeline-visualization)!

New Features

- Add Tokens-to-Token ViT backbone and converted checkpoints. ([467](https://github.com/open-mmlab/mmclassification/pull/467))
- Add Res2Net backbone and converted weights. ([465](https://github.com/open-mmlab/mmclassification/pull/465))
- Support ImageNet21k dataset. ([461](https://github.com/open-mmlab/mmclassification/pull/461))
- Support seesaw loss. ([500](https://github.com/open-mmlab/mmclassification/pull/500))
- Add a pipeline visualization tool. ([406](https://github.com/open-mmlab/mmclassification/pull/406))
- Add a tool to find broken files. ([482](https://github.com/open-mmlab/mmclassification/pull/482))
- Add a tool to test TorchServe. ([468](https://github.com/open-mmlab/mmclassification/pull/468))

Improvements

- Refator Vision Transformer. ([395](https://github.com/open-mmlab/mmclassification/pull/395))
- Use context manager to reuse matplotlib figures. ([432](https://github.com/open-mmlab/mmclassification/pull/432))

Bug Fixes

- Remove `DistSamplerSeedHook` if use `IterBasedRunner`. ([501](https://github.com/open-mmlab/mmclassification/pull/501))
- Set the priority of `EvalHook` to "LOW" to avoid a bug when using `IterBasedRunner`. ([488](https://github.com/open-mmlab/mmclassification/pull/488))
- Fix a wrong parameter of `get_root_logger` in `apis/train.py`. ([486](https://github.com/open-mmlab/mmclassification/pull/486))
- Fix version check in dataset builder. ([474](https://github.com/open-mmlab/mmclassification/pull/474))

Docs Update

- Add English Colab tutorials and update Chinese Colab tutorials. ([483](https://github.com/open-mmlab/mmclassification/pull/483), [#497](https://github.com/open-mmlab/mmclassification/pull/497))
- Add tutuorial for config files. ([487](https://github.com/open-mmlab/mmclassification/pull/487))
- Add model-pages in Model Zoo. ([480](https://github.com/open-mmlab/mmclassification/pull/480))
- Add code-spell pre-commit hook and fix a large mount of typos. ([470](https://github.com/open-mmlab/mmclassification/pull/470))

Contributors
A total of 6 developers contributed to this release.

mzr1996 Ezra-Yu tansor youqingxiaozhua 0x4f5da2 okotaku

Highlights

- We have improved compatibility with downstream repositories like MMDetection and MMSegmentation. We will add some examples about how to use our backbones in MMDetection.
- Add RepVGG backbone and checkpoints. Welcome to use it!
- Add timm backbones wrapper, now you can simply use backbones of pytorch-image-models in MMClassification!

New Features

- Add RepVGG backbone and checkpoints. ([414](https://github.com/open-mmlab/mmclassification/pull/414))
- Add timm backbones wrapper. ([427](https://github.com/open-mmlab/mmclassification/pull/427))

Improvements

- Fix TnT compatibility and verbose warning. ([436](https://github.com/open-mmlab/mmclassification/pull/436))
- Support setting `--out-items` in `tools/test.py`. ([437](https://github.com/open-mmlab/mmclassification/pull/437))
- Add datetime info and saving model using torch<1.6 format. ([439](https://github.com/open-mmlab/mmclassification/pull/439))
- Improve downstream repositories compatibility. ([421](https://github.com/open-mmlab/mmclassification/pull/421))
- Rename the option `--options` to `--cfg-options` in some tools. ([425](https://github.com/open-mmlab/mmclassification/pull/425))
- Add PyTorch 1.9 and Python 3.9 build workflow, and remove some CI. ([422](https://github.com/open-mmlab/mmclassification/pull/422))

Bug Fixes

- Fix format error in `test.py` when metric returns `np.ndarray`. ([441](https://github.com/open-mmlab/mmclassification/pull/441))
- Fix `publish_model` bug if no parent of `out_file`. ([463](https://github.com/open-mmlab/mmclassification/pull/463))
- Fix num_classes bug in pytorch2onnx.py. ([458](https://github.com/open-mmlab/mmclassification/pull/458))
- Fix missing runtime requirement `packaging`. ([459](https://github.com/open-mmlab/mmclassification/pull/459))
- Fix saving simplified model bug in ONNX export tool. ([438](https://github.com/open-mmlab/mmclassification/pull/438))

Docs Update

- Update `getting_started.md` and `install.md`. And rewrite `finetune.md`. ([466](https://github.com/open-mmlab/mmclassification/pull/466))
- Use PyTorch style docs theme. ([457](https://github.com/open-mmlab/mmclassification/pull/457))
- Update metafile and Readme. ([435](https://github.com/open-mmlab/mmclassification/pull/435))
- Add `CITATION.cff`. ([428](https://github.com/open-mmlab/mmclassification/pull/428))

Contributors
A total of 8 developers contributed to this release.
Charlyo Ezra-Yu mzr1996 amirassov RangiLyu zhaoxin111 uniyushu zhangrui-wolf

Highlights
- Support `hparams` argument in `AutoAugment` and `RandAugment` to provide hyperparameters for sub-policies.
- Support custom squeeze channels in `SELayer`.
- Support classwise weight in losses.

New Features

- Add `hparams` argument in `AutoAugment` and `RandAugment` and some other improvement. ([398](https://github.com/open-mmlab/mmclassification/pull/398))
- Support classwise weight in losses ([388](https://github.com/open-mmlab/mmclassification/pull/388))
- Enhence `SELayer` to support custom squeeze channels. ([417](https://github.com/open-mmlab/mmclassification/pull/417))

Code Refactor

- Better result visualization ([419](https://github.com/open-mmlab/mmclassification/pull/419))
- Use `post_process` function to handle pred result processing. ([390](https://github.com/open-mmlab/mmclassification/pull/390))
- Update `digit_version` function. ([402](https://github.com/open-mmlab/mmclassification/pull/402))
- Avoid albumentations to install both opencv and opencv-headless. ([397](https://github.com/open-mmlab/mmclassification/pull/397))
- Avoid unnecessary listdir when building ImageNet. ([396](https://github.com/open-mmlab/mmclassification/pull/396))
- Use dynamic mmcv download link in TorchServe dockerfile. ([387](https://github.com/open-mmlab/mmclassification/pull/387))

Docs Improvement

- Add readme of some algorithms and update meta yml ([418](https://github.com/open-mmlab/mmclassification/pull/418))
- Add Copyright information. ([413](https://github.com/open-mmlab/mmclassification/pull/413))
- Add PR template and modify issue template ([380](https://github.com/open-mmlab/mmclassification/pull/380))

Contributors
A total of 5 developers contributed to this release.
azad96 Ezra-Yu mzr1996 mmeendez8 sovrasov

Highlights
- Add transformer-in-transformer backbone and pretrain checkpoints, refers to [the paper](https://arxiv.org/abs/2103.00112).
- Add Chinese colab tutorial.
- Provide dockerfile to build mmcls dev docker image.

New Features

- Add transformer in transformer backbone and pretrain checkpoints. ([339](https://github.com/open-mmlab/mmclassification/pull/339))
- Support mim, welcome to use mim to manage your mmcls project. ([376](https://github.com/open-mmlab/mmclassification/pull/376))
- Add Dockerfile. ([365](https://github.com/open-mmlab/mmclassification/pull/365))
- Add ResNeSt configs. ([332](https://github.com/open-mmlab/mmclassification/pull/332))

Improvements

- Use the `presistent_works` option if available, to accelerate training. ([349](https://github.com/open-mmlab/mmclassification/pull/349))
- Add Chinese ipynb tutorial. ([306](https://github.com/open-mmlab/mmclassification/pull/306))
- Refactor unit tests. ([321](https://github.com/open-mmlab/mmclassification/pull/321))
- Support to test mmdet inference with mmcls backbone. ([343](https://github.com/open-mmlab/mmclassification/pull/343))
- Use zero as default value of `thrs` in metrics. ([341](https://github.com/open-mmlab/mmclassification/pull/341))

Bug Fixes

- Fix ImageNet dataset annotation file parse bug. ([370](https://github.com/open-mmlab/mmclassification/pull/370))
- Fix docstring typo and init bug in ShuffleNetV1. ([374](https://github.com/open-mmlab/mmclassification/pull/374))
- Use local ATTENTION registry to avoid conflict with other repositories. ([376](https://github.com/open-mmlab/mmclassification/pull/375))
- Fix swin transformer config bug. ([355](https://github.com/open-mmlab/mmclassification/pull/355))
- Fix `patch_cfg` argument bug in SwinTransformer. ([368](https://github.com/open-mmlab/mmclassification/pull/368))
- Fix duplicate `init_weights` call in ViT init function. ([373](https://github.com/open-mmlab/mmclassification/pull/373))
- Fix broken `_base_` link in a resnet config. ([361](https://github.com/open-mmlab/mmclassification/pull/361))
- Fix vgg-19 model link missing. ([363](https://github.com/open-mmlab/mmclassification/pull/363))

Contributors

A total of 8 developers contributed to this release.

Ezra-Yu, HIT-cwh, Junjun2016, LXXXXR, mzr1996, pvys, wangruohui, ZwwWayne

New Features

- Support Swin-Transformer backbone and add training configs for Swin-Transformer on ImageNet. (271)
- Add pretrained model of RegNetX. (269)
- Support adding custom hooks in the config file. (305)
- Improve and add Chinese translation of `CONTRIBUTING.md` and all tools tutorials. (320)
- Dump config before training. (282)
- Add torchscript and torchserve deployment tools. (279, 284)

Improvements

- Improve test tools and add some new tools. (322)
- Correct MobilenetV3 backbone structure and add pretained models. (291)
- Refactor `PatchEmbed` and `HybridEmbed` as independent components. (330)
- Refactor mixup and cutmix as `Augments` to support more funtions. (278)
- Refactor weights initialization method. (270, 318, 319)
- Refactor `LabelSmoothLoss` to support multiple calculation formulas. (285)

Bug Fixes

- Fix bug for CPU training. (286)
- Fix missing test data when `num_imgs` can not be evenly divided by `num_gpus`. (299)
- Fix build compatible with pytorch v1.3-1.5. (301)
- Fix `magnitude_std` bug in `RandAugment`. (309)
- Fix bug when `samples_per_gpu` is 1. (311)

New Features
- Improve and add Chinese translation of `data_pipeline.md` and `new_modules.md`. (265)
- Build Chinese translation on readthedocs. (267)
- Add an argument efficientnet_style to `RandomResizedCrop` and `CenterCrop`. (268)

Improvements
- Only allow directory operation when rank==0 when testing. (258)
- Fix typo in `base_head`. (274)
- Update ResNeXt checkpoints. (283)

Bug Fixes
- Add attribute `data.test` in MNIST configs. (264)
- Download CIFAR/MNIST dataset only on rank 0. (273)
- Fix MMCV version compatibility. (276)
- Fix CIFAR color channels bug and update checkpoints in model zoo. (280)