Fit for MMPretrain EfficientFormer (#2108)

* add ef in rgtest metafile

* fix for EfficientFormerClsHead to LinearClsHead

* fix lint

* update benchmark

* fix lint

* fix docs

* fix docs

* add openvino test

docs/en/03-benchmark/benchmark.md

@@ -625,6 +625,28 @@ Users can directly test the performance through [how_to_evaluate_a_model.md](../
     <td align="center">-</td>
     <td align="center">-</td>
     <td align="center">-</td>
+  <tr>
+    <td align="center" rowspan="2"><a href="https://github.com/open-mmlab/mmpretrain/blob/main/configs/efficientformer/efficientformer-l1_8xb128_in1k.py">EfficientFormer</a></td>
+    <td align="center">top-1</td>
+    <td align="center">80.46</td>
+    <td align="center">80.45</td>
+    <td align="center">80.46</td>
+    <td align="center">80.46</td>
+    <td align="center">-</td>
+    <td align="center">-</td>
+    <td align="center">-</td>
+    <td align="center">-</td>
+  </tr>
+  <tr>
+    <td align="center">top-5</td>
+    <td align="center">94.99</td>
+    <td align="center">94.98</td>
+    <td align="center">94.99</td>
+    <td align="center">94.99</td>
+    <td align="center">-</td>
+    <td align="center">-</td>
+    <td align="center">-</td>
+    <td align="center">-</td>
   </tr>
 </tbody>
 </table>

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

@@ -185,3 +185,4 @@ Besides python API, mmdeploy SDK also provides other FFI (Foreign Function Inter
 | [MobileOne](https://github.com/open-mmlab/mmpretrain/tree/main/configs/mobileone)                  |      Y      |      Y       |    N     |  N   |   ?   |    N     |
 | [EfficientNet](https://github.com/open-mmlab/mmpretrain/tree/main/configs/efficientnet)            |      Y      |      Y       |    N     |  N   |   ?   |    N     |
 | [Conformer](https://github.com/open-mmlab/mmpretrain/tree/main/configs/conformer)                  |      Y      |      Y       |    N     |  N   |   ?   |    N     |
+| [EfficientFormer](https://github.com/open-mmlab/mmpretrain/tree/main/configs/efficientformer)      |      Y      |      Y       |    Y     |  N   |   ?   |    Y     |

docs/zh_cn/03-benchmark/benchmark.md

@@ -608,6 +608,7 @@ GPU: ncnn, TensorRT, PPLNN
     <td align="center">81.18</td>
     <td align="center">-</td>
     <td align="center">-</td>
+    <td align="center">-</td>
   </tr>
   <tr>
     <td align="center">top-5</td>
@@ -618,6 +619,30 @@ GPU: ncnn, TensorRT, PPLNN
     <td align="center">95.61</td>
     <td align="center">-</td>
     <td align="center">-</td>
+    <td align="center">-</td>
+  <tr>
+    <td align="center" rowspan="2"><a href="https://github.com/open-mmlab/mmpretrain/blob/main/configs/efficientformer/efficientformer-l1_8xb128_in1k.py">EfficientFormer</a></td>
+    <td align="center">top-1</td>
+    <td align="center">80.46</td>
+    <td align="center">80.45</td>
+    <td align="center">80.46</td>
+    <td align="center">80.46</td>
+    <td align="center">-</td>
+    <td align="center">-</td>
+    <td align="center">-</td>
+    <td align="center">-</td>
+  </tr>
+  <tr>
+    <td align="center">top-5</td>
+    <td align="center">94.99</td>
+    <td align="center">94.98</td>
+    <td align="center">94.99</td>
+    <td align="center">94.99</td>
+    <td align="center">-</td>
+    <td align="center">-</td>
+    <td align="center">-</td>
+    <td align="center">-</td>
+  </tr>
   </tr>
 </tbody>
 </table>

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

@@ -190,3 +190,4 @@ for label_id, score in result:
 | [MobileOne](https://github.com/open-mmlab/mmpretrain/tree/main/configs/mobileone)                  |      Y      |      Y       |    N     |  N   |   ?   |    N     |
 | [EfficientNet](https://github.com/open-mmlab/mmpretrain/tree/main/configs/efficientnet)            |      Y      |      Y       |    N     |  N   |   ?   |    N     |
 | [Conformer](https://github.com/open-mmlab/mmpretrain/tree/main/configs/conformer)                  |      Y      |      Y       |    N     |  N   |   ?   |    N     |
+| [EfficientFormer](https://github.com/open-mmlab/mmpretrain/tree/main/configs/efficientformer)      |      Y      |      Y       |    Y     |  N   |   ?   |    Y     |

mmdeploy/codebase/mmpretrain/deploy/classification.py

@@ -332,6 +332,9 @@ class Classification(BaseTask):
             dict: Composed of the postprocess information.
         """
         postprocess = self.model_cfg.model.head
+        if postprocess['type'] == 'EfficientFormerClsHead':
+            postprocess['type'] = 'LinearClsHead'
+
         if 'topk' not in postprocess:
             topk = (1, )
             logger = get_root_logger()

tests/regression/mmpretrain.yml

@@ -237,3 +237,13 @@ models:
     pipelines:
       - *pipeline_ort_dynamic_fp32
       - *pipeline_trt_dynamic_fp32
+
+  - name: EfficientFormer
+    metafile: configs/efficientformer/metafile.yml
+    model_configs:
+      - configs/efficientformer/efficientformer-l1_8xb128_in1k.py
+    pipelines:
+      - *pipeline_ts_fp32
+      - *pipeline_ort_dynamic_fp32
+      - *pipeline_trt_dynamic_fp32
+      - *pipeline_openvino_dynamic_fp32