update setup and docs in v0.2.7

.github/CONTRIBUTING.md

@@ -1 +1 @@
-We appreciate all contributions to improve OpenMixup. We follow the developing standard of MMLab. Please refer to [CONTRIBUTING.md](https://github.com/open-mmlab/mmcv/blob/master/CONTRIBUTING.md) in MMCV for more details about the contributing guideline.
+We appreciate all contributions to improve OpenMixup. Currently, we follow the developing standard of MMLab. Please refer to [CONTRIBUTING.md](https://github.com/open-mmlab/mmcv/blob/master/CONTRIBUTING.md) in MMCV for more details about the contributing guideline.

.github/ISSUE_TEMPLATE/----.md

@@ -19,7 +19,7 @@ assignees: ''
 
 ### 相关信息
 
-1. `pip list | grep "mmcv\|mmcls\|^torch"` 命令的输出
+1. `pip list | grep "openmixup\|^torch"` 命令的输出
    \[填写这里\]
 2. 如果你修改了，或者使用了新的配置文件，请在这里写明
 
@@ -29,5 +29,5 @@ assignees: ''
 
 3. 如果你是在训练过程中遇到的问题，请填写完整的训练日志和报错信息
    \[填写这里\]
-4. 如果你对 `mmcls` 文件夹下的代码做了其他相关的修改，请在这里写明
+4. 如果你对 `openmixup` 文件夹下的代码做了其他相关的修改，请在这里写明
    \[填写这里\]

.github/ISSUE_TEMPLATE/---bug.md

@@ -24,7 +24,7 @@ assignees: ''
 
 ### 相关信息
 
-1. `pip list | grep "mmcv\|mmcls\|^torch"` 命令的输出
+1. `pip list | grep "openmixup\|^torch"` 命令的输出
    \[填写这里\]
 2. 如果你修改了，或者使用了新的配置文件，请在这里写明
 
@@ -34,7 +34,7 @@ assignees: ''
 
 3. 如果你是在训练过程中遇到的问题，请填写完整的训练日志和报错信息
    \[填写这里\]
-4. 如果你对 `mmcls` 文件夹下的代码做了其他相关的修改，请在这里写明
+4. 如果你对 `openmixup` 文件夹下的代码做了其他相关的修改，请在这里写明
    \[填写这里\]
 
 ### 附加内容

.github/ISSUE_TEMPLATE/bug_report.md

@@ -22,7 +22,7 @@ The command you executed.
 
 ### Post related information
 
-1. The output of `pip list | grep "mmcv\|mmcls\|^torch"`
+1. The output of `pip list | grep "openmixup\|^torch"`
    \[here\]
 2. Your config file if you modified it or created a new one.
 
@@ -32,7 +32,7 @@ The command you executed.
 
 3. Your train log file if you meet the problem during training.
    \[here\]
-4. Other code you modified in the `mmcls` folder.
+4. Other code you modified in the `openmixup` folder.
    \[here\]
 
 ### Additional context

.github/ISSUE_TEMPLATE/general-questions.md

@@ -17,7 +17,7 @@ assignees: ''
 
 ### Post related information
 
-1. The output of `pip list | grep "mmcv\|mmcls\|^torch"`
+1. The output of `pip list | grep "openmixup\|^torch"`
    \[here\]
 2. Your config file if you modified it or created a new one.
 
@@ -27,5 +27,5 @@ assignees: ''
 
 3. Your train log file if you meet the problem during training.
    \[here\]
-4. Other code you modified in the `mmcls` folder.
+4. Other code you modified in the `openmixup` folder.
    \[here\]

.github/pull_request_template.md

@@ -1,4 +1,4 @@
-Thanks for your contribution to OpenMixup and we appreciate it a lot. The following instructions would make your pull request more healthy and more easily get feedback. We follow the standard in MMLab, e.g., [MMClassification](https://github.com/open-mmlab/mmclassification). If you do not understand some items, don't worry, just make the pull request and seek help from maintainers.
+Thanks for your contribution to [OpenMixup](https://github.com/Westlake-AI/openmixup) and we appreciate it a lot. The following instructions would make your pull request more healthy and more easily get feedback. Currently, we follow the standard in MMLab, e.g., [MMClassification](https://github.com/open-mmlab/mmclassification). If you do not understand some items, don't worry, just make the pull request and seek help from maintainers.
 
 ## Motivation
 

.gitignore

@@ -104,17 +104,15 @@ venv.bak/
 # mypy
 .mypy_cache/
 
-openmixup/version.py
-version.py
+# custom
 data
 .vscode
 .idea
-
-# custom
 *.pkl
 *.pkl.json
 *.log.json
 work_dirs/
+/openmixup/.mim
 tools/exp_bash/
 pretrains
 
@@ -137,4 +135,5 @@ configs/selfsup_IP89
 # configs/semisup
 # configs/selfsup
 *.json
+*.toml
 openmixup/models/classifiers/backup

LICENSE

@@ -186,7 +186,7 @@
       same "printed page" as the copyright notice for easier
       identification within third-party archives.
 
-   Copyright 2020-2021 Open-MMLab.
+   Copyright 2021-2022 CAIRI AI Lab.
 
    Licensed under the Apache License, Version 2.0 (the "License");
    you may not use this file except in compliance with the License.

MANIFEST.in

@@ -0,0 +1,3 @@
+include requirements/*.txt
+recursive-include openmixup/.mim/configs *.py *.yml
+recursive-include openmixup/.mim/tools *.sh *.py

README.md

@@ -1,6 +1,7 @@
 # OpenMixup
-[![Release](https://img.shields.io/badge/release-V0.2.7-%09%2360004F)](https://github.com/Westlake-AI/openmixup/releases)
-[![Docs](https://img.shields.io/badge/docs-latest-%23002FA7)](https://openmixup.readthedocs.io/en/latest/)
+[![release](https://img.shields.io/badge/release-V0.2.7-%09%2360004F)](https://github.com/Westlake-AI/openmixup/releases)
+[![PyPI](https://img.shields.io/pypi/v/openmixup)](https://pypi.org/project/openmixup)
+[![docs](https://img.shields.io/badge/docs-latest-%23002FA7)](https://openmixup.readthedocs.io/en/latest/)
 [![license](https://img.shields.io/badge/license-Apache--2.0-%23B7A800)](https://github.com/Westlake-AI/openmixup/blob/main/LICENSE)
 ![open issues](https://img.shields.io/github/issues-raw/Westlake-AI/openmixup?color=%23FF9600)
 [![issue resolution](https://img.shields.io/badge/issue%20resolution-1%20d-%23009763)](https://github.com/Westlake-AI/openmixup/issues)
@@ -12,14 +13,6 @@
 [🔍Awesome MIM](https://openmixup.readthedocs.io/en/latest/awesome_selfsup/MIM.html) |
 [🆕News](https://openmixup.readthedocs.io/en/latest/changelog.html)
 
-## News and Updates
-
-[2022-12-16] `OpenMixup` v0.2.7 is released (issue [#35](https://github.com/Westlake-AI/openmixup/issues/35)).
-
-[2022-12-02] Update new features and documents of `OpenMixup` v0.2.6 (issue [#24](https://github.com/Westlake-AI/openmixup/issues/24), issue [#25](https://github.com/Westlake-AI/openmixup/issues/25), issue [#31](https://github.com/Westlake-AI/openmixup/issues/31), and issue [#33](https://github.com/Westlake-AI/openmixup/issues/33)). Update the official implementation of [MogaNet](https://arxiv.org/abs/2211.03295).
-
-[2022-09-14] `OpenMixup` v0.2.6 is released (issue [#20](https://github.com/Westlake-AI/openmixup/issues/20)).
-
 ## Introduction
 
 The main branch works with **PyTorch 1.8** (required by some self-supervised methods) or higher (we recommend **PyTorch 1.12**). You can still use **PyTorch 1.6** for supervised classification methods.
@@ -63,13 +56,18 @@ The main branch works with **PyTorch 1.8** (required by some self-supervised met
   </ol>
 </details>
 
-<p align="right">(<a href="#top">back to top</a>)</p>
+## News and Updates
+
+[2022-12-16] `OpenMixup` v0.2.7 is released (issue [#35](https://github.com/Westlake-AI/openmixup/issues/35)).
+
+[2022-12-02] Update new features and documents of `OpenMixup` v0.2.6 (issue [#24](https://github.com/Westlake-AI/openmixup/issues/24), issue [#25](https://github.com/Westlake-AI/openmixup/issues/25), issue [#31](https://github.com/Westlake-AI/openmixup/issues/31), and issue [#33](https://github.com/Westlake-AI/openmixup/issues/33)). Update the official implementation of [MogaNet](https://arxiv.org/abs/2211.03295).
+
+[2022-09-14] `OpenMixup` v0.2.6 is released (issue [#20](https://github.com/Westlake-AI/openmixup/issues/20)).
 
 ## Installation
 
-OpenMixup is compatible with **Python 3.7/3.8/3.9** and **PyTorch >= 1.8**. Here are installation steps for development:
+OpenMixup is compatible with **Python 3.6/3.7/3.8/3.9** and **PyTorch >= 1.6**. Here are quick installation steps for development:
 
-### From Source
 ```shell
 conda create -n openmixup python=3.8 pytorch=1.12 cudatoolkit=11.3 torchvision -c pytorch -y
 conda activate openmixup
@@ -79,63 +77,30 @@ git clone https://github.com/Westlake-AI/openmixup.git
 cd openmixup
 python setup.py develop
 ```
-### From PyPI
-```shell
-conda create -n openmixup python=3.8 pytorch=1.12 cudatoolkit=11.3 torchvision -c pytorch -y
-conda activate openmixup
-pip install openmim
-mim install mmcv-full
-pip install openmixup
-cd openmixup
-python setup.py develop
-```
 
-Please refer to [install.md](docs/en/install.md) for more detailed installation and dataset preparation instructions.
+Please refer to [install.md](docs/en/install.md) for more detailed installation and dataset preparation.
 
 ## Getting Started
 
-OpenMixup supports Linux, macOS and Windows. It enables easy implementation and extensions of mixup data augmentation methods in existing supervised, self-, and semi-supervised visual recognition models. Please see [get_started.md](docs/en/get_started.md) for the basic usage of OpenMixup.
+OpenMixup supports Linux and macOS. It enables easy implementation and extensions of mixup data augmentation methods in existing supervised, self-, and semi-supervised visual recognition models. Please see [get_started.md](docs/en/get_started.md) for the basic usage of OpenMixup.
 
-### Quick Start
-This is an example of how to quickly set up the OpenMixup on your device. You can get a local copy up by running the belowing example steps.
-#### Step0: Create your environment
-```shell
-conda create -n openmixup python=3.8 pytorch=1.12 cudatoolkit=11.3 torchvision -c pytorch -y
-conda activate openmixup
-```
-#### Step1: Install the required packages
-```shell
-pip install -U openmim
-mim install mmcv-full
-```
+### Training and Evaluation Scripts
 
-#### Step2: Clone and develop the project
-```shell
-git clone https://github.com/Westlake-AI/openmixup.git
-cd openmixup
-python setup.py develop
-```
-Now you can use the image you just built for your own project. 
-
-### Training Script
-
-Here, we provide example scripts for you to quickly start the accelerated end-to-end multiple GPUs training with specified `CONFIG_FILE`. 
+Here, we provide scripts for starting a quick end-to-end training with multiple `GPUs` and the specified `CONFIG_FILE`. 
 ```shell
 bash tools/dist_train.sh ${CONFIG_FILE} ${GPUS} [optional arguments]
 ```
-To be more specific, you can run the script below to train a designated mixup CIFAR100 classification algorithm with 4 GPUs:
+For example, you can run the script below to train a ResNet-50 classifier on ImageNet with 4 GPUs:
 ```shell
-CUDA_VISIBLE_DEVICES=0,1,2,3 bash tools/dist_train.sh openmixup\configs\classification\cifar100\mixups\basic\r18_mixups_CE_none.py 4
+CUDA_VISIBLE_DEVICES=0,1,2,3 PORT=29500 bash tools/dist_train.sh configs/classification/imagenet/resnet/resnet50_4xb64_cos_ep100.py 4
 ```
-
-### Evaluation Script
-After trianing, you can test the trained models with the corresponding evaluation script. An example with 4 GPUs evaluation is as follows:
-
+After trianing, you can test the trained models with the corresponding evaluation script:
 ```shell
-CUDA_VISIBLE_DEVICES=0,1,2,3 bash tools/dist_test.sh ${CONFIG_FILE} ${GPUS} ${PATH_TO_MODEL} [optional arguments]
+bash tools/dist_test.sh ${CONFIG_FILE} ${GPUS} ${PATH_TO_MODEL} [optional arguments]
 ```
 
-### Develop
+### Development
+
 Please see [Tutorials](docs/en/tutorials) for more developing examples and tech details:
 
 - [config files](docs/en/tutorials/0_config.md)
@@ -276,7 +241,7 @@ This project is released under the [Apache 2.0 license](LICENSE). See `LICENSE`
 
 ## Citation
 
-If you find this project useful in your research, please consider cite our GitHub repo or [tech report](https://arxiv.org/abs/2209.04851):
+If you find this project useful in your research, please consider star our GitHub repo and cite [tech report](https://arxiv.org/abs/2209.04851):
 
 ```BibTeX
 @misc{2022openmixup,
@@ -301,14 +266,12 @@ If you find this project useful in your research, please consider cite our GitHu
 
 ## Contributors and Contact
 
-For now, the direct contributors include: Siyuan Li ([@Lupin1998](https://github.com/Lupin1998)), Zedong Wang ([@Jacky1128](https://github.com/Jacky1128)), Zicheng Liu ([@pone7](https://github.com/pone7)), Di Wu ([@wudi-bu](https://github.com/wudi-bu)), Tengfei Wang ([@wang-tf](https://github.com/wang-tf)), and Minglong Liu ([@minhlong94](https://github.com/minhlong94)). We thank contributors from MMSelfSup and MMClassification and all public contributors!
+For help, new features, or reporting bugs associated with OpenMixup, please open a [GitHub issue](https://github.com/Westlake-AI/openmixup/issues) and [pull request](https://github.com/Westlake-AI/openmixup/pulls) with the tag "help wanted" or "enhancement". For now, the direct contributors include: Siyuan Li ([@Lupin1998](https://github.com/Lupin1998)), Zedong Wang ([@Jacky1128](https://github.com/Jacky1128)), and Zicheng Liu ([@pone7](https://github.com/pone7)). We thank all public contributors and contributors from MMSelfSup and MMClassification!
 
 This repo is currently maintained by:
 
 - Siyuan Li ([email protected]), Westlake University
 - Zedong Wang ([email protected]), Westlake University
 - Zicheng Liu ([email protected]), Westlake University
 
-If you have suggestions that would make OpenMixup better, please fork the repo and create a pull request. It is also encouraged to open an issue with the tag "help wanted" or "enhancement". Don't forget to give our OpenMixup a star! Thanks again!
-
 <p align="right">(<a href="#top">back to top</a>)</p>

configs/classification/imagenet/deit/deit_base_8xb128_ep300.py

@@ -6,6 +6,7 @@
 
 # data
 data = dict(imgs_per_gpu=128, workers_per_gpu=10)
+sampler = "RepeatAugSampler"  # the official repo uses `repeated_aug` for more stable training
 
 # additional hooks
 update_interval = 1  # 128 x 8gpus x 1 accumulates = bs1024

configs/classification/imagenet/deit/deit_base_adan_8xb256_fp16_ep150.py

@@ -0,0 +1,43 @@
+_base_ = [
+    '../../_base_/models/deit/deit_base_p16_sz224.py',
+    '../../_base_/datasets/imagenet/swin_sz224_8xbs128.py',
+    '../../_base_/default_runtime.py',
+]
+
+# data
+data = dict(imgs_per_gpu=256, workers_per_gpu=12)
+sampler = "RepeatAugSampler"  # the official repo uses `repeated_aug` for more stable training
+
+# additional hooks
+update_interval = 1  # 256 x 8gpus x 1 accumulates = bs2048
+
+# optimizer
+optimizer = dict(
+    type='Adan',
+    lr=1.5e-2,  # lr = 1.5e-2 / bs2048
+    weight_decay=0.02, eps=1e-8, betas=(0.98, 0.92, 0.99),
+    max_grad_norm=5.0,
+    paramwise_options={
+        '(bn|ln|gn)(\d+)?.(weight|bias)': dict(weight_decay=0.),
+        'norm': dict(weight_decay=0.),
+        'bias': dict(weight_decay=0.),
+        'cls_token': dict(weight_decay=0.),
+        'pos_embed': dict(weight_decay=0.),
+    })
+
+# fp16
+use_fp16 = True
+fp16 = dict(type='mmcv', loss_scale='dynamic')
+optimizer_config = dict(update_interval=update_interval)
+
+# lr scheduler
+lr_config = dict(
+    policy='CosineAnnealing',
+    by_epoch=False, min_lr=1e-8,
+    warmup='linear',
+    warmup_iters=60, warmup_by_epoch=True,  # warmup 60 epochs.
+    warmup_ratio=1e-8,
+)
+
+# runtime settings
+runner = dict(type='EpochBasedRunner', max_epochs=150)

configs/classification/imagenet/deit/deit_base_adan_8xb256_fp16_ep300.py

@@ -0,0 +1,4 @@
+_base_ = './deit_base_adan_8xb256_fp16_ep150.py'
+
+# runtime settings
+runner = dict(type='EpochBasedRunner', max_epochs=300)

configs/classification/imagenet/deit/deit_small_8xb128_ep300.py

@@ -6,6 +6,7 @@
 
 # data
 data = dict(imgs_per_gpu=128, workers_per_gpu=10)
+# sampler = "RepeatAugSampler"  # this repo reproduce the performance without `repeated_aug`
 
 # additional hooks
 update_interval = 1  # 128 x 8gpus x 1 accumulates = bs1024

configs/classification/imagenet/deit/deit_small_adan_8xb256_fp16_ep150.py

@@ -1,18 +1,9 @@
 _base_ = [
     '../../_base_/models/deit/deit_small_p16_sz224.py',
-    '../../_base_/datasets/imagenet/deit_adan_sz224_8xbs256.py',
+    '../../_base_/datasets/imagenet/swin_sz224_8xbs128.py',
     '../../_base_/default_runtime.py',
 ]
 
-# model settings
-model = dict(
-    head=dict(
-        type='VisionTransformerClsHead',  # mixup BCE + label smooth
-        loss=dict(type='LabelSmoothLoss',
-            label_smooth_val=0.1, num_classes=1000, mode='multi_label', loss_weight=1.0),
-        in_channels=384, num_classes=1000)
-)
-
 # data
 data = dict(imgs_per_gpu=256, workers_per_gpu=12)
 
@@ -22,17 +13,17 @@
 # optimizer
 optimizer = dict(
     type='Adan',
-    lr=1.5e-3,  # lr = 1.5e-3 / bs2048
+    lr=1.5e-2,  # lr = 1.5e-2 / bs2048
     weight_decay=0.02, eps=1e-8, betas=(0.98, 0.92, 0.99),
-    max_grad_norm=0.0,
+    max_grad_norm=5.0,
     paramwise_options={
         '(bn|ln|gn)(\d+)?.(weight|bias)': dict(weight_decay=0.),
         'norm': dict(weight_decay=0.),
         'bias': dict(weight_decay=0.),
         'cls_token': dict(weight_decay=0.),
         'pos_embed': dict(weight_decay=0.),
     })
-                     
+
 # fp16
 use_fp16 = True
 fp16 = dict(type='mmcv', loss_scale='dynamic')