diff --git a/README.md b/README.md index 3f1f7a7b7c..7e9c02bc46 100644 --- a/README.md +++ b/README.md @@ -31,6 +31,16 @@ English | [简体中文](README_zh-CN.md) +## Highlights + +The OpenMMLab team released a new generation of training engine [MMEngine](https://github.com/open-mmlab/mmengine) at the World Artificial Intelligence Conference on September 1, 2022. It is a foundational library for training deep learning models. Compared with MMCV, it provides a universal and powerful runner, an open architecture with a more unified interface, and a more customizable training process. + +At the same time, MMCV released [2.x](https://github.com/open-mmlab/mmcv/tree/2.x) release candidate version and will release 2.x official version on January 1, 2023. + +In version 2.x, it removed components related to the training process and added a data transformation module. Also, starting from 2.x, it renamed the package names **mmcv** to **mmcv-lite** and **mmcv-full** to **mmcv**. For details, see [Compatibility Documentation](docs/en/compatibility.md). + +MMCV will maintain both `1.x` and `2.x` versions. For details, see [Branch Maintenance Plan](README.md#branch-maintenance-plan). + ## Introduction MMCV is a foundational library for computer vision research and it provides the following functionalities: @@ -110,6 +120,16 @@ pip install -U openmim mim install mmcv ``` +## Branch Maintenance Plan + +MMCV currently has two branches, the master and 2.x branches, which go through the following three phases. + +| Phase | Time | Branch | description | +| -------------------- | --------------------- | --------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------- | +| RC Period | 2022/9/1 - 2022.12.31 | Release candidate code (2.x version) will be released on 2.x branch. Default master branch is still 1.x version | Master and 2.x branches iterate normally | +| Compatibility Period | 2023/1/1 - 2023.12.31 | **Default master branch will be switched to 2.x branch**, and 1.x branch will correspond to 1.x version | We still maintain the old version 1.x, respond to user needs, but try not to introduce changes that break compatibility; master branch iterates normally | +| Maintenance Period | From 2024/1/1 | Default master branch corresponds to 2.x version and 1.x branch is 1.x version | 1.x branch is in maintenance phase, no more new feature support; master branch is iterating normally | + ## Supported projects - [MIM](https://github.com/open-mmlab/mim): MIM installs OpenMMLab packages. diff --git a/README_zh-CN.md b/README_zh-CN.md index cd024becc6..d63940655c 100644 --- a/README_zh-CN.md +++ b/README_zh-CN.md @@ -31,6 +31,14 @@ [English](README.md) | 简体中文 +## Highlights + +OpenMMLab 团队于 2022 年 9 月 1 日在世界人工智能大会发布了新一代训练引擎 [MMEngine](https://github.com/open-mmlab/mmengine)，它是一个用于训练深度学习模型的基础库。相比于 MMCV，它提供了更高级且通用的训练器、接口更加统一的开放架构以及可定制化程度更高的训练流程。 + +与此同时，MMCV 发布了 [2.x](https://github.com/open-mmlab/mmcv/tree/2.x) 预发布版本，并将于 2023 年 1 月 1 日发布 2.x 正式版本。在 2.x 版本中，它删除了和训练流程相关的组件，并新增了数据变换模块。另外，从 2.x 版本开始，重命名包名 **mmcv** 为 **mmcv-lite** 以及 **mmcv-full** 为 **mmcv**。详情见[兼容性文档](docs/zh_cn/compatibility.md)。 + +MMCV 会同时维护 1.x 和 2.x 版本，详情见[分支维护计划](README_zh-CN.md#分支维护计划)。 + ## 简介 MMCV 是一个面向计算机视觉的基础库，它提供了以下功能： @@ -110,6 +118,16 @@ pip install -U openmim mim install mmcv ``` +## 分支维护计划 + +MMCV 目前有两个分支，分别是 master 和 2.x 分支，它们会经历以下三个阶段： + +| 阶段 | 时间 | 分支 | 说明 | +| ------ | --------------------- | ------------------------------------------------------------ | -------------------------------------------------------------------------------------------------------- | +| 公测期 | 2022/9/1 - 2022.12.31 | 公测版代码发布在 2.x 分支；默认主分支 master 仍对应 1.x 版本 | master 和 2.x 分支正常进行迭代 | +| 兼容期 | 2023/1/1 - 2023.12.31 | **切换默认主分支 master 为 2.x 版本**；1.x 分支对应 1.x 版本 | 保持对旧版本 1.x 的维护和开发，响应用户需求，但尽量不引进破坏旧版本兼容性的改动；master 分支正常进行迭代 | +| 维护期 | 2024/1/1 - 待定 | 默认主分支 master 为 2.x 版本；1.x 分支对应 1.x 版本 | 1.x 分支进入维护阶段，不再进行新功能支持；master 分支正常进行迭代 | + ## 支持的开源项目 - [MIM](https://github.com/open-mmlab/mim): MIM 是 OpenMMlab 项目、算法、模型的统一入口 diff --git a/docs/en/compatibility.md b/docs/en/compatibility.md index c8618388f4..fc8516c492 100644 --- a/docs/en/compatibility.md +++ b/docs/en/compatibility.md @@ -1,3 +1,53 @@ +### v2.0.0rc1 + +The OpenMMLab team released a new generation of training engine [MMEngine](https://github.com/open-mmlab/mmengine) at the World Artificial Intelligence Conference on September 1, 2022. It is a foundational library for training deep learning models. Compared with MMCV, it provides a universal and powerful runner, an open architecture with a more unified interface, and a more customizable training process. + +At the same time, MMCV released [2.x](https://github.com/open-mmlab/mmcv/tree/2.x) release candidate version and will release 2.x official version on January 1, 2023. In version 2.x, it has the following changes: + +(1) It removed the following components: + +- `mmcv.fileio` module, removed in PR [#2179](https://github.com/open-mmlab/mmcv/pull/2179). FileIO module from mmengine will be used wherever required. +- `mmcv.runner`, `mmcv.parallel`, `mmcv. engine` and `mmcv.device`, removed in PR [#2216](https://github.com/open-mmlab/mmcv/pull/2216). +- All classes in `mmcv.utils` (eg `Config` and `Registry`) and many functions, removed in PR [#2217](https://github.com/open-mmlab/mmcv/pull/2217). Only a few functions related to mmcv are reserved. +- `mmcv.onnex`, `mmcv.tensorrt` modules and related functions, removed in PR [#2225](https://github.com/open-mmlab/mmcv/pull/2225). + +(2) It added the [`mmcv.transforms`](https://github.com/open-mmlab/mmcv/tree/2.x/mmcv/transforms) data transformation module. + +(3) It renamed the package name **mmcv** to **mmcv-lite** and **mmcv-full** to **mmcv** in PR [#2235](https://github.com/open-mmlab/mmcv/pull/2235). Also, change the default value of the environment variable `MMCV_WITH_OPS` from 0 to 1. + + + + + + + + + + + + +
MMCV < 2.0MMCV >= 2.0
+ +```bash +# Contains ops, because the highest version of mmcv-full is less than 2.0.0, so there is no need to add version restrictions +pip install mmcv-full -f xxxx + +# do not contain ops +pip install "mmcv < 2.0.0" +``` + + + +```bash +# Contains ops +pip install "mmcv>=2.0.0rc1" -f xxxx + +# Ops are not included, because the starting version of mmcv-lite is 2.0.0rc1, so there is no need to add version restrictions +pip install mmcv-lite +``` + +
+ ### v1.3.18 Some ops have different implementations on different devices. Lots of macros and type checks are scattered in several files, which makes the code hard to maintain. For example: diff --git a/docs/zh_cn/compatibility.md b/docs/zh_cn/compatibility.md index 6bda560927..9f8bbfa931 100644 --- a/docs/zh_cn/compatibility.md +++ b/docs/zh_cn/compatibility.md @@ -1,3 +1,53 @@ +### v2.0.0rc1 + +OpenMMLab 团队于 2022 年 9 月 1 日在世界人工智能大会发布了新一代训练引擎 [MMEngine](https://github.com/open-mmlab/mmengine)，它是一个用于训练深度学习模型的基础库。相比于 MMCV，它提供了更高级且通用的训练器、接口更加统一的开放架构以及可定制化程度更高的训练流程。 + +与此同时，MMCV 发布了 [2.x](https://github.com/open-mmlab/mmcv/tree/2.x) 预发布版本并将于 2023 年 1 月 1 日发布 2.x 正式版本。在 2.x 版本中，它有以下变化： + +（1）删除了以下组件： + +- `mmcv.fileio` 模块，删除于 PR [#2179](https://github.com/open-mmlab/mmcv/pull/2179)。在需要使用 FileIO 的地方使用 mmengine 中的 FileIO 模块 +- `mmcv.runner`、`mmcv.parallel`、`mmcv.engine` 和 `mmcv.device`，删除于 PR [#2216](https://github.com/open-mmlab/mmcv/pull/2216) +- `mmcv.utils` 的所有类（例如 `Config` 和 `Registry`）和大部分函数，删除于 PR [#2217](https://github.com/open-mmlab/mmcv/pull/2217)，只保留少数和 mmcv 相关的函数 +- `mmcv.onnex`、`mmcv.tensorrt` 模块以及相关的函数，删除于 PR [#2225](https://github.com/open-mmlab/mmcv/pull/2225) + +（2）新增了 [`mmcv.transforms`](https://github.com/open-mmlab/mmcv/tree/2.x/mmcv/transforms) 数据变换模块 + +（3）在 PR [#2235](https://github.com/open-mmlab/mmcv/pull/2235) 中将包名 **mmcv** 重命名为 **mmcv-lite**、 **mmcv-full** 重命名为 **mmcv**。此外，将环境变量 `MMCV_WITH_OPS` 的默认值从 0 改为 1 + + + + + + + + + + + + +
MMCV < 2.0MMCV >= 2.0
+ +```bash +# 包含算子，因为 mmcv-full 的最高版本小于 2.0.0，所以无需加版本限制 +pip install mmcv-full -f xxxx + +# 不包含算子 +pip install "mmcv < 2.0.0" +``` + + + +```bash +# 包含算子 +pip install "mmcv>=2.0.0rc1" -f xxxx + +# 不包含算子，因为 mmcv-lite 的起始版本为 2.0.0rc1，所以无需加版本限制 +pip install mmcv-lite +``` + +
+ ### v1.3.18 部分自定义算子对于不同的设备有不同实现，为此添加的大量宏命令与类型检查使得代码变得难以维护。例如： diff --git a/mmcv/__init__.py b/mmcv/__init__.py index 14c556acdf..e87858c594 100644 --- a/mmcv/__init__.py +++ b/mmcv/__init__.py @@ -1,5 +1,7 @@ # Copyright (c) OpenMMLab. All rights reserved. # flake8: noqa +import warnings + from .arraymisc import * from .fileio import * from .image import * @@ -14,3 +16,11 @@ # - parallel # - op # - device + +warnings.warn( + 'On January 1, 2023, MMCV will release v2.0.0, in which it will remove ' + 'components related to the training process and add a data transformation ' + 'module. In addition, it will rename the package names mmcv to mmcv-lite ' + 'and mmcv-full to mmcv. ' + 'See https://github.com/open-mmlab/mmcv/blob/master/docs/en/compatibility.md ' + 'for more details.')