diff --git a/README.md b/README.md
index f0ca668b12..10370f1ec4 100644
--- a/README.md
+++ b/README.md
@@ -71,10 +71,10 @@ Note: MMCV requires Python 3.7+.
There are two versions of MMCV:
-- **mmcv**: comprehensive, with full features and various CUDA ops out of the box. It takes longer time to build.
-- **mmcv-lite**: lite, without CUDA ops but all other features, similar to mmcv\<1.0.0. It is useful when you do not need those CUDA ops.
+- **mmcv**: comprehensive, with full features and various CUDA/hip(ROCm) ops out of the box. It takes longer time to build.
+- **mmcv-lite**: lite, without CUDA/hip(ROCm) ops but all other features, similar to mmcv\<1.0.0. It is useful when you do not need those CUDA/hip(ROCm) ops.
-**Note**: Do not install both versions in the same environment, otherwise you may encounter errors like `ModuleNotFound`. You need to uninstall one before installing the other. `Installing the full version is highly recommended if CUDA is available`.
+**Note**: Do not install both versions in the same environment, otherwise you may encounter errors like `ModuleNotFound`. You need to uninstall one before installing the other. `Installing the full version is highly recommended if CUDA/hip(ROCm) is available`.
### Install mmcv
@@ -93,14 +93,20 @@ If you need to specify the version of mmcv, you can use the following command:
mim install mmcv==2.0.0
```
-If you find that the above installation command does not use a pre-built package ending with `.whl` but a source package ending with `.tar.gz`, you may not have a pre-build package corresponding to the PyTorch or CUDA or mmcv version, in which case you can [build mmcv from source](https://mmcv.readthedocs.io/en/latest/get_started/build.html).
+If you find that the above installation command does not use a pre-built package ending with `.whl` but a source package ending with `.tar.gz`, you may not have a pre-build package corresponding to the PyTorch or CUDA/hip(ROCm) or mmcv version, in which case you can [build mmcv from source](https://mmcv.readthedocs.io/en/latest/get_started/build.html).
Installation log using pre-built packages
-
+
+###### 1. For CUDA pre-built packages
Looking in links: https://download.openmmlab.com/mmcv/dist/cu102/torch1.8.0/index.html
Collecting mmcv
Downloading https://download.openmmlab.com/mmcv/dist/cu102/torch1.8.0/mmcv-2.0.0-cp38-cp38-manylinux1_x86_64.whl
+
+
+###### 2. For ROCm pre-built packages:
+The external [`mmcv-rocm-build` repository](https://github.com/Looong01/mmcv-rocm-build) provides wheels and detailed instructions on how to install MMCV for ROCm.
+If you have any questions about it, please open an issue [here](https://github.com/Looong01/mmcv-rocm-build/issues).
@@ -131,6 +137,8 @@ you may first refer to this [Frequently Asked Questions](https://mmcv.readthedoc
If you face installation problems or runtime issues, you may first refer to this [Frequently Asked Questions](https://mmcv.readthedocs.io/en/latest/faq.html) to see if there is a solution. If the problem is still not solved, feel free to open an [issue](https://github.com/open-mmlab/mmcv/issues).
+If you face some installation issues with ROCm build, you may first refer to [this](https://github.com/Looong01/mmcv-rocm-build/issues) to see if there is a solution or feel free to open an [issue](https://github.com/Looong01/mmcv-rocm-build/issues) if not.
+
## Citation
If you find this project useful in your research, please consider cite:
diff --git a/README_zh-CN.md b/README_zh-CN.md
index e48f33bd7f..27dcdb6896 100644
--- a/README_zh-CN.md
+++ b/README_zh-CN.md
@@ -91,10 +91,10 @@ MMCV 支持多种平台，包括：
MMCV 有两个版本：
-- **mmcv**: 完整版，包含所有的特性以及丰富的开箱即用的 CUDA 算子。注意完整版本可能需要更长时间来编译。
-- **mmcv-lite**: 精简版，不包含 CUDA 算子但包含其余所有特性和功能，类似 MMCV 1.0 之前的版本。如果你不需要使用 CUDA 算子的话，精简版可以作为一个考虑选项。
+- **mmcv**: 完整版，包含所有的特性以及丰富的开箱即用的 CUDA/hip(ROCm) 算子。注意完整版本可能需要更长时间来编译。
+- **mmcv-lite**: 精简版，不包含 CUDA/hip(ROCm) 算子但包含其余所有特性和功能，类似 MMCV 1.0 之前的版本。如果你不需要使用 CUDA/hip(ROCm) 算子的话，精简版可以作为一个考虑选项。
-**注意**: 请不要在同一个环境中安装两个版本，否则可能会遇到类似 `ModuleNotFound` 的错误。在安装一个版本之前，需要先卸载另一个。`如果 CUDA 可用，强烈推荐安装 mmcv`。
+**注意**: 请不要在同一个环境中安装两个版本，否则可能会遇到类似 `ModuleNotFound` 的错误。在安装一个版本之前，需要先卸载另一个。`如果 CUDA/hip(ROCm) 可用，强烈推荐安装 mmcv`。
### 安装 mmcv
@@ -113,14 +113,20 @@ mim install mmcv
mim install mmcv==2.0.0
```
-如果发现上述的安装命令没有使用预编译包（以 `.whl` 结尾）而是使用源码包（以 `.tar.gz` 结尾）安装，则有可能是我们没有提供和当前环境的 PyTorch 版本、CUDA 版本相匹配的 mmcv 预编译包，此时，你可以[源码安装 mmcv](https://mmcv.readthedocs.io/zh_CN/latest/get_started/build.html)。
+如果发现上述的安装命令没有使用预编译包（以 `.whl` 结尾）而是使用源码包（以 `.tar.gz` 结尾）安装，则有可能是我们没有提供和当前环境的 PyTorch 版本、CUDA/hip(ROCm) 版本相匹配的 mmcv 预编译包，此时，你可以[源码安装 mmcv](https://mmcv.readthedocs.io/zh_CN/latest/get_started/build.html)。
使用预编译包的安装日志
-
+
+###### 1. 对于CUDA构建的预编译包:
Looking in links: https://download.openmmlab.com/mmcv/dist/cu102/torch1.8.0/index.html
Collecting mmcv
Downloading https://download.openmmlab.com/mmcv/dist/cu102/torch1.8.0/mmcv-2.0.0-cp38-cp38-manylinux1_x86_64.whl
+
+
+###### 2. 对于 ROCm 构建的预编译包:
+一个外部 [`mmcv-rocm-build` 仓库](https://github.com/Looong01/mmcv-rocm-build) 提供了如何为 ROCm 安装 MMCV 的 wheels 和详细说明。
+如果您对此有任何疑问，请在 [这里](https://github.com/Looong01/mmcv-rocm-build/issues) 打开一个 issue。
@@ -148,6 +154,8 @@ mim install mmcv-lite
如果你遇到了安装问题或者运行时问题，请查看[问题解决页面](https://mmcv.readthedocs.io/zh_CN/latest/faq.html)是否已有解决方案。如果问题仍然没有解决，欢迎提 [issue](https://github.com/open-mmlab/mmcv/issues)。
+如果您在使用 ROCm 构建时遇到一些安装问题，你可以先参考 [这里](https://github.com/Looong01/mmcv-rocm-build/issues)是否已有解决方案。如果问题仍然没有解决，欢迎提 [issue](https://github.com/Looong01/mmcv-rocm-build/issues)。
+
## 贡献指南
我们感谢所有的贡献者为改进和提升 MMCV 所作出的努力。请参考[贡献指南](CONTRIBUTING.md)来了解参与项目贡献的相关指引。