Fix retinanet wrong visualization with torch1.13+openvino (#1977)

* fix

* fix ci

* fix dead links

Author: RunningLeon

.circleci/test.yml

@@ -85,7 +85,7 @@ commands:
             cd mmdeploy
             python -m mim install -r requirements/codebases.txt
             python -m pip install -r requirements.txt
-            python -m pip install -U numpy clip numba
+            python -m pip install -U numpy clip numba transformers
             cd ..
   perform_model_converter_ut:
     steps:
@@ -145,7 +145,7 @@ jobs:
             docker exec mmdeploy pip install onnxruntime==1.8.1
             docker exec mmdeploy mim install -r mmdeploy/requirements/codebases.txt
             docker exec mmdeploy pip install -r mmdeploy/requirements.txt
-            docker exec mmdeploy pip install -U numpy clip numba
+            docker exec mmdeploy pip install -U numpy clip numba transformers
       - run:
           name: Perform Model Converter unittests
           command: |

.github/workflows/backend-ort.yml

@@ -17,7 +17,7 @@ concurrency:
   cancel-in-progress: true
 
 jobs:
-  buils_ort:
+  build_ort:
     runs-on: ubuntu-20.04
     strategy:
       matrix:

.github/workflows/build.yml

@@ -49,7 +49,7 @@ jobs:
           python -m pip install -r requirements/backends.txt
           python -m mim install "mmcv>=2.0.0rc1"
           python -m mim install -r requirements/codebases.txt
-          python -m pip install -U numpy clip numba
+          python -m pip install -U numpy clip numba transformers
           python -m pip list
       - name: Install mmyolo
         run: |
@@ -149,7 +149,7 @@ jobs:
           python -m pip install -r requirements/backends.txt
           python -m mim install "mmcv>=2.0.0rc1"
           CFLAGS=$CFLAGS python -m mim install -r requirements/codebases.txt
-          python -m pip install -U pycuda numpy clip numba
+          python -m pip install -U pycuda numpy clip numba transformers
           python -m pip list
       - name: Build and install
         run: |
@@ -194,7 +194,7 @@ jobs:
           python -m pip install -r requirements/backends.txt
           python -m mim install "mmcv>=2.0.0rc1"
           python -m mim install -r requirements/codebases.txt
-          python -m pip install -U pycuda numpy clip numba
+          python -m pip install -U pycuda numpy clip numba transformers
           python -m pip list
       - name: Build and install
         run: |
@@ -232,7 +232,7 @@ jobs:
           apt update && apt install unzip
           python3 -V
           python3 -m pip install opencv-python==4.5.4.60 opencv-python-headless==4.5.4.60 opencv-contrib-python==4.5.4.60
-          python3 -m pip install openmim numpy pycuda clip
+          python3 -m pip install openmim numpy pycuda clip transformers
           python3 -m pip install -r requirements.txt
           python3 -m mim install $(cat requirements/codebases.txt | grep mmcls)
           python3 -m pip list

docs/en/04-supported-codebases/mmaction2.md

@@ -21,7 +21,7 @@ ______________________________________________________________________
 
 ### Install mmaction2
 
-Please follow the [installation guide](https://github.com/open-mmlab/mmaction2/tree/1.x#installation) to install mmaction2.
+Please follow the [installation guide](https://github.com/open-mmlab/mmaction2/tree/main#installation) to install mmaction2.
 
 ### Install mmdeploy
 
@@ -184,7 +184,7 @@ Besides python API, mmdeploy SDK also provides other FFI (Foreign Function Inter
 
 ## Supported models
 
-| Model                                                                                     | TorchScript | ONNX Runtime | TensorRT | ncnn | PPLNN | OpenVINO |
-| :---------------------------------------------------------------------------------------- | :---------: | :----------: | :------: | :--: | :---: | :------: |
-| [TSN](https://github.com/open-mmlab/mmaction2/tree/1.x/configs/recognition/tsn)           |      N      |      Y       |    Y     |  N   |   N   |    N     |
-| [SlowFast](https://github.com/open-mmlab/mmaction2/tree/1.x/configs/recognition/slowfast) |      N      |      Y       |    Y     |  N   |   N   |    N     |
+| Model                                                                                      | TorchScript | ONNX Runtime | TensorRT | ncnn | PPLNN | OpenVINO |
+| :----------------------------------------------------------------------------------------- | :---------: | :----------: | :------: | :--: | :---: | :------: |
+| [TSN](https://github.com/open-mmlab/mmaction2/tree/main/configs/recognition/tsn)           |      N      |      Y       |    Y     |  N   |   N   |    N     |
+| [SlowFast](https://github.com/open-mmlab/mmaction2/tree/main/configs/recognition/slowfast) |      N      |      Y       |    Y     |  N   |   N   |    N     |

docs/en/04-supported-codebases/mmocr.md

@@ -24,7 +24,7 @@ ______________________________________________________________________
 
 ### Install mmocr
 
-Please follow the [installation guide](https://mmocr.readthedocs.io/en/latest/install.html) to install mmocr.
+Please follow the [installation guide](https://mmocr.readthedocs.io/en/latest/get_started/install.html) to install mmocr.
 
 ### Install mmdeploy
 

docs/en/07-developer-guide/architecture.md

@@ -93,7 +93,7 @@ Operators customized for ncnn are in the `csrc/mmdeploy/backend_ops/ncnn/` direc
 
 - topk
 - tensorslice
-- shpe
+- shape
 - gather
 - expand
 - constantofshape

docs/zh_cn/04-supported-codebases/mmaction2.md

@@ -21,7 +21,7 @@ ______________________________________________________________________
 
 ### 安装 mmaction2
 
-请参考[官网安装指南](https://github.com/open-mmlab/mmaction2/tree/1.x#installation).
+请参考[官网安装指南](https://github.com/open-mmlab/mmaction2/tree/main#installation).
 
 ### 安装 mmdeploy
 
@@ -187,7 +187,7 @@ for label_id, score in result:
 
 ## 模型支持列表
 
-| Model                                                                                     | TorchScript | ONNX Runtime | TensorRT | ncnn | PPLNN | OpenVINO |
-| :---------------------------------------------------------------------------------------- | :---------: | :----------: | :------: | :--: | :---: | :------: |
-| [TSN](https://github.com/open-mmlab/mmaction2/tree/1.x/configs/recognition/tsn)           |      N      |      Y       |    Y     |  N   |   N   |    N     |
-| [SlowFast](https://github.com/open-mmlab/mmaction2/tree/1.x/configs/recognition/slowfast) |      N      |      Y       |    Y     |  N   |   N   |    N     |
+| Model                                                                                      | TorchScript | ONNX Runtime | TensorRT | ncnn | PPLNN | OpenVINO |
+| :----------------------------------------------------------------------------------------- | :---------: | :----------: | :------: | :--: | :---: | :------: |
+| [TSN](https://github.com/open-mmlab/mmaction2/tree/main/configs/recognition/tsn)           |      N      |      Y       |    Y     |  N   |   N   |    N     |
+| [SlowFast](https://github.com/open-mmlab/mmaction2/tree/main/configs/recognition/slowfast) |      N      |      Y       |    Y     |  N   |   N   |    N     |

docs/zh_cn/04-supported-codebases/mmocr.md

@@ -24,7 +24,7 @@ ______________________________________________________________________
 
 ### 安装 mmocr
 
-请参考[官网安装指南](https://mmocr.readthedocs.io/en/latest/install.html).
+请参考[官网安装指南](https://mmocr.readthedocs.io/en/latest/get_started/install.html).
 
 ### 安装 mmdeploy
 

docs/zh_cn/07-developer-guide/architecture.md

@@ -93,7 +93,7 @@ $ tree -L 1
 
 - topk
 - tensorslice
-- shpe
+- shape
 - gather
 - expand
 - constantofshape

mmdeploy/codebase/mmdet/models/task_modules/coders/delta_xywh_bbox_coder.py

@@ -101,8 +101,9 @@ def delta2bbox(rois,
     dxy = denorm_deltas[..., :2]
     dwh = denorm_deltas[..., 2:]
 
-    xy1 = rois[..., None, :2]
-    xy2 = rois[..., None, 2:]
+    # fix openvino on torch1.13
+    xy1 = rois[..., :2].unsqueeze(2)
+    xy2 = rois[..., 2:].unsqueeze(2)
 
     pxy = (xy1 + xy2) * 0.5
     pwh = xy2 - xy1