[DOC] update npu doc

docs/en/device/npu.md

@@ -2,7 +2,8 @@
 
 ## Usage
 
-Please refer to [link](https://github.com/open-mmlab/mmcv/blob/master/docs/zh_cn/get_started/build.md) installing mmcv on NPU Devices.
+Please refer to [link](https://mmcv.readthedocs.io/en/latest/get_started/build.html#build-mmcv-full-on-ascend-npu-machine) installing mmcv
+on NPU Devices.
 
 Here we use 8 NPUs on your computer to train the model with the following command:
 
@@ -16,20 +17,29 @@ Also, you can use only one NPU to train the model with the following command:
 python tools/train.py configs/ssd/ssd300_coco.py
 ```
 
-## Verified Models
+## Models Results
 
 |        Model         | box AP | mask AP | Config                                                                                                                        | Download                                                                                                     |
 | :------------------: | :----: | :-----: | :---------------------------------------------------------------------------------------------------------------------------- | :----------------------------------------------------------------------------------------------------------- |
 |     [ssd300](<>)     |  25.6  |   ---   | [config](https://github.com/open-mmlab/mmdetection/blob/master/configs/ssd/ssd300_fp16_coco.py)                               | [log](https://download.openmmlab.com/mmdetection/v2.0/npu/ssd300_coco.log.json)                              |
 |     [ssd512](<>)     |  29.4  |   ---   | [config](https://github.com/open-mmlab/mmdetection/blob/master/configs/ssd/ssd512_fp16_coco.py)                               | [log](https://download.openmmlab.com/mmdetection/v2.0/npu/ssd512_coco.log.json)                              |
-| [\*ssdlite-mbv2](<>) |  20.2  |   ---   | [config](https://github.com/open-mmlab/mmdetection/blob/master/configs/ssd/ssdlite_mobilenetv2_scratch_600e_coco.py)          | [log](https://download.openmmlab.com/mmdetection/v2.0/npu/ssdlite_mobilenetv2_scratch_600e_coco.log.json)    |
+| [ssdlite-mbv2\*](<>) |  20.2  |   ---   | [config](https://github.com/open-mmlab/mmdetection/blob/master/configs/ssd/ssdlite_mobilenetv2_scratch_600e_coco.py)          | [log](https://download.openmmlab.com/mmdetection/v2.0/npu/ssdlite_mobilenetv2_scratch_600e_coco.log.json)    |
+| [retinanet-r18](<>)  |  31.8  |   ---   | [config](https://github.com/open-mmlab/mmdetection/blob/master/configs/retinanet/retinanet_r18_fpn_1x8_1x_coco.py)            | [log](https://download.openmmlab.com/mmdetection/v2.0/npu/retinanet_r18_fpn_1x8_1x_coco.log.json)            |
 | [retinanet-r50](<>)  |  36.6  |   ---   | [config](https://github.com/open-mmlab/mmdetection/blob/master/configs/retinanet/retinanet_r50_fpn_fp16_1x_coco.py)           | [log](https://download.openmmlab.com/mmdetection/v2.0/npu/retinanet_r50_fpn_1x_coco.log.json)                |
-|   [\*fcos-r50](<>)   |  36.1  |   ---   | [config](https://github.com/open-mmlab/mmdetection/blob/master/configs/fcos/fcos_r50_caffe_fpn_gn-head_fp16_1x_bs8x8_coco.py) | [log](https://download.openmmlab.com/mmdetection/v2.0/npu/fcos_r50_caffe_fpn_gn-head_1x_coco_bs8x8.log.json) |
+|   [yolov3-608](<>)   |  34.7  |   ---   | [config](https://github.com/open-mmlab/mmdetection/blob/master/configs/yolo/yolov3_d53_fp16_mstrain-608_273e_coco.py)         | [log](https://download.openmmlab.com/mmdetection/v2.0/npu/yolov3_d53_fp16_mstrain-608_273e_coco.log.json)    |
+|  [yolox-s\*\*](<>)   |  39.9  |   ---   | [config](https://github.com/open-mmlab/mmdetection/blob/master/configs/yolox/yolox_s_8x8_300e_coco.py)                        | [log](https://download.openmmlab.com/mmdetection/v2.0/npu/yolox_s_8x8_300e_coco.log.json)                    |
+| [centernet-r18](<>)  |  26.1  |   ---   | [config](https://github.com/open-mmlab/mmdetection/blob/master/configs/centernet/centernet_resnet18_140e_cocoo.py)            | [log](https://download.openmmlab.com/mmdetection/v2.0/npu/centernet_resnet18_140e_coco.log.jsonn)            |
+|   [fcos-r50\*](<>)   |  36.1  |   ---   | [config](https://github.com/open-mmlab/mmdetection/blob/master/configs/fcos/fcos_r50_caffe_fpn_gn-head_fp16_1x_bs8x8_coco.py) | [log](https://download.openmmlab.com/mmdetection/v2.0/npu/fcos_r50_caffe_fpn_gn-head_1x_coco_bs8x8.log.json) |
 |   [solov2-r50](<>)   |  ---   |  34.7   | [config](https://github.com/open-mmlab/mmdetection/blob/master/configs/solov2/solov2_r50_fpn_1x_coco.py)                      | [log](https://download.openmmlab.com/mmdetection/v2.0/npu/solov2_r50_fpn_1x_coco.log.json)                   |
 
 **Notes:**
 
-- If not specially marked, the results are same between results on the NPU and results on the GPU with FP32.
-- (\*) The results on the NPU of these models are aligned with the results of the mixed-precision training on the GPU, but are lower than the results of the FP32. This situation is mainly related to the phase of the model itself in mixed-precision training, users please adjust the hyperparameters to achieve the best result by self.
+- If not specially marked, the results on NPU are the same as those on the GPU with FP32.
+- (\*) The results on the NPU of these models are aligned with the results of the mixed-precision training on the GPU,
+  but are lower than the results of the FP32. This situation is mainly related to the phase of the model itself in
+  mixed-precision training, users may need to adjust the hyperparameters to achieve better results.
+- (\*\*) The accuracy of yolox-s on the GPU in mixed precision is 40.1, with `persister_woker=True` in the data loader config by default.
+  There are currently some bugs on NPUs that prevent the last few epochs from running, but the accuracy is less affected and the difference can be ignored.
+  affected and can be ignored.
 
 **All above models are provided by Huawei Ascend group.**

docs/zh_cn/device/npu.md

@@ -0,0 +1,42 @@
+# NPU (华为 昇腾)
+
+## 使用方法
+
+请参考该链接 [link](https://mmcv.readthedocs.io/en/latest/get_started/build.html#build-mmcv-full-on-ascend-npu-machine) 安装npu版本的mmcv。
+
+以下展示单机八卡场景的运行指令:
+
+```shell
+bash tools/dist_train.sh configs/ssd/ssd300_coco.py 8
+```
+
+以下展示单机单卡下的运行指令:
+
+```shell
+python tools/train.py configs/ssd/ssd300_coco.py
+```
+
+## 模型验证结果
+
+|        Model         | box AP | mask AP | Config                                                                                                                        | Download                                                                                                     |
+| :------------------: | :----: | :-----: | :---------------------------------------------------------------------------------------------------------------------------- | :----------------------------------------------------------------------------------------------------------- |
+|     [ssd300](<>)     |  25.6  |   ---   | [config](https://github.com/open-mmlab/mmdetection/blob/master/configs/ssd/ssd300_fp16_coco.py)                               | [log](https://download.openmmlab.com/mmdetection/v2.0/npu/ssd300_coco.log.json)                              |
+|     [ssd512](<>)     |  29.4  |   ---   | [config](https://github.com/open-mmlab/mmdetection/blob/master/configs/ssd/ssd512_fp16_coco.py)                               | [log](https://download.openmmlab.com/mmdetection/v2.0/npu/ssd512_coco.log.json)                              |
+| [ssdlite-mbv2\*](<>) |  20.2  |   ---   | [config](https://github.com/open-mmlab/mmdetection/blob/master/configs/ssd/ssdlite_mobilenetv2_scratch_600e_coco.py)          | [log](https://download.openmmlab.com/mmdetection/v2.0/npu/ssdlite_mobilenetv2_scratch_600e_coco.log.json)    |
+| [retinanet-r18](<>)  |  31.8  |   ---   | [config](https://github.com/open-mmlab/mmdetection/blob/master/configs/retinanet/retinanet_r18_fpn_1x8_1x_coco.py)            | [log](https://download.openmmlab.com/mmdetection/v2.0/npu/retinanet_r18_fpn_1x8_1x_coco.log.json)            |
+| [retinanet-r50](<>)  |  36.6  |   ---   | [config](https://github.com/open-mmlab/mmdetection/blob/master/configs/retinanet/retinanet_r50_fpn_fp16_1x_coco.py)           | [log](https://download.openmmlab.com/mmdetection/v2.0/npu/retinanet_r50_fpn_1x_coco.log.json)                |
+|   [yolov3-608](<>)   |  34.7  |   ---   | [config](https://github.com/open-mmlab/mmdetection/blob/master/configs/yolo/yolov3_d53_fp16_mstrain-608_273e_coco.py)         | [log](https://download.openmmlab.com/mmdetection/v2.0/npu/yolov3_d53_fp16_mstrain-608_273e_coco.log.json)    |
+|  [yolox-s\*\*](<>)   |  39.9  |   ---   | [config](https://github.com/open-mmlab/mmdetection/blob/master/configs/yolox/yolox_s_8x8_300e_coco.py)                        | [log](https://download.openmmlab.com/mmdetection/v2.0/npu/yolox_s_8x8_300e_coco.log.json)                    |
+| [centernet-r18](<>)  |  26.1  |   ---   | [config](https://github.com/open-mmlab/mmdetection/blob/master/configs/centernet/centernet_resnet18_140e_cocoo.py)            | [log](https://download.openmmlab.com/mmdetection/v2.0/npu/centernet_resnet18_140e_coco.log.jsonn)            |
+|   [fcos-r50\*](<>)   |  36.1  |   ---   | [config](https://github.com/open-mmlab/mmdetection/blob/master/configs/fcos/fcos_r50_caffe_fpn_gn-head_fp16_1x_bs8x8_coco.py) | [log](https://download.openmmlab.com/mmdetection/v2.0/npu/fcos_r50_caffe_fpn_gn-head_1x_coco_bs8x8.log.json) |
+|   [solov2-r50](<>)   |  ---   |  34.7   | [config](https://github.com/open-mmlab/mmdetection/blob/master/configs/solov2/solov2_r50_fpn_1x_coco.py)                      | [log](https://download.openmmlab.com/mmdetection/v2.0/npu/solov2_r50_fpn_1x_coco.log.json)                   |
+
+**Notes:**
+
+- 如果没有特别标记，NPU 上的结果与使用 FP32 的 GPU 上的结果结果相同。
+- (\*) 这些模型在 NPU 上的结果与 GPU 上的混合精度训练结果一致，但低于 FP32 的结果。这种情况主要与模型本身在混合精度训练中的特点有关，
+  用户可以自行调整超参数来获得更高精度。
+- (\*\*) GPU 上 yolox-s 在混合精度下的精度为 40.1 低于readme中 40.5 的水平;默认情况下，yolox-s 启用 persister_woker，但这个参数
+  目前在NPU上存在一些bug，会导致在最后几个epoch由于资源耗尽报错退出，对整体精度影响有限可以忽略。
+
+**All above models are provided by Huawei Ascend group.**

docs/zh_cn/index.rst

@@ -48,6 +48,12 @@ Welcome to MMDetection's documentation!
 
    api.rst
 
+.. toctree::
+   :maxdepth: 1
+   :caption: 设备支持
+
+   device/npu.md
+
 
 Indices and tables
 ==================