为什么多线程使用pose+stgcn的过程中会遇到注册表变化

[lmc2000]

我对每检测的一个人使用多线程,同时预测动作,先加载pose模型然后加载stgcn模型,然后再pose模型提取骨骼在使用stgcn模型预测动作,但是我发现加载模型的时候,会提示我注册表为mmpose,没有gcnhead
Loads checkpoint by local backend from path: weights/rtmpose-l_simcc-coco-wholebody_pt-aic-coco_270e-384x288-eaeb96c8_20230125.pth

Loads checkpoint by local backend from path: weights/stgcn_acc81_class2_5.pth
04/22 18:00:25 - mmengine - WARNING - The current default scope "mmaction" is not "mmpose", init_default_scope will force set the currentdefault scope to "mmpose".
04/22 18:00:25 - mmengine - WARNING - The current default scope "mmdet" is not "mmpose", init_default_scope will force set the currentdefault scope to "mmpose".
Loads checkpoint by local backend from path: weights/rtmpose-l_simcc-coco-wholebody_pt-aic-coco_270e-384x288-eaeb96c8_20230125.pth
04/22 18:00:26 - mmengine - WARNING - The current default scope "mmpose" is not "mmaction", init_default_scope will force set the currentdefault scope to "mmaction".

Process Process-1:
Traceback (most recent call last):
File "/data/miniconda3/envs/openmmlab/lib/python3.8/multiprocessing/process.py", line 315, in _bootstrap
self.run()
File "/data/miniconda3/envs/openmmlab/lib/python3.8/multiprocessing/process.py", line 108, in run
self._target(*self._args, **self._kwargs)
File "/home/www/models/mmaction/inference/显存计算.py", line 152, in Inspection
p,label=future.result()
File "/data/miniconda3/envs/openmmlab/lib/python3.8/concurrent/futures/_base.py", line 437, in result
return self.__get_result()
File "/data/miniconda3/envs/openmmlab/lib/python3.8/concurrent/futures/_base.py", line 389, in __get_result
raise self._exception
File "/data/miniconda3/envs/openmmlab/lib/python3.8/concurrent/futures/thread.py", line 57, in run
result = self.fn(*self.args, **self.kwargs)
File "/home/www/models/mmaction/inference/显存计算.py", line 90, in action_recognition
stgcn_model = init_recognizer(args.action_config, args.action_checkpoint, args.device)
File "/home/www/models/mmaction/mmaction/apis/inference.py", line 50, in init_recognizer
model = MODELS.build(config.model)
File "/data/miniconda3/envs/openmmlab/lib/python3.8/site-packages/mmengine/registry/registry.py", line 570, in build
return self.build_func(cfg, *args, **kwargs, registry=self)
File "/data/miniconda3/envs/openmmlab/lib/python3.8/site-packages/mmengine/registry/build_functions.py", line 232, in build_model_from_cfg
return build_from_cfg(cfg, registry, default_args)
File "/data/miniconda3/envs/openmmlab/lib/python3.8/site-packages/mmengine/registry/build_functions.py", line 121, in build_from_cfg
obj = obj_cls(**args) # type: ignore
File "/home/www/models/mmaction/mmaction/models/recognizers/base.py", line 122, in init
self.cls_head = MODELS.build(cls_head)
File "/data/miniconda3/envs/openmmlab/lib/python3.8/site-packages/mmengine/registry/registry.py", line 570, in build
return self.build_func(cfg, *args, **kwargs, registry=self)
File "/data/miniconda3/envs/openmmlab/lib/python3.8/site-packages/mmengine/registry/build_functions.py", line 232, in build_model_from_cfg
return build_from_cfg(cfg, registry, default_args)
File "/data/miniconda3/envs/openmmlab/lib/python3.8/site-packages/mmengine/registry/build_functions.py", line 100, in build_from_cfg
raise KeyError(
KeyError: 'GCNHead is not in the mmpose::model registry. Please check whether the value of GCNHead is correct or it was registered as expected. More details can be found at https://mmengine.readthedocs.io/en/latest/advanced_tutorials/config.html#import-the-custom-module'

[lmc2000]

if len(person_d) >= len(det_results) / 4:
        KP=KeypointProcessor()
        with LockWrapper(model_lock):
            PE=PoseExtraction()
            stgcn_model = init_recognizer(args.action_config, args.action_checkpoint, args.device)
        # 插值插帧
        person_det=DetectionProcessor().interpolation_frame(person_det)
        anno = PE.ntu_pose_extraction(frames, person_det)
        anno = KP.del_keypoints(anno, nobody=True)
        anno = KP.hide_keypoints(anno,thr=0.5)
        stgcn_result = inference_skeleton(stgcn_model, [anno], anno['img_shape'])
        stgcn_pred_score = stgcn_result.pred_score
        stgcn_pred=stgcn_result.pred_label
        label=stgcn_result.pred_label.to('cpu')
        #线程结束,清除显存
        del stgcn_model
        del PE
        torch.cuda.empty_cache()