HarryBarry123
HarryBarry123
when I tried to train File "/home/pe/projects/czy/StreamYOLO-main/exps/train_utils/double_trainer.py", line 314, in resume_train ckpt = torch.load(ckpt_file, map_location=self.device)["model"] │ │ │ │ └ 'cuda:0' │ │ │ └ │ │ └ '/home/pe/projects/czy/StreamYOLO-main/tools/yolox_s.pth' │...
当我加载grounding_dino_swin-t_pretrain_obj365_goldg_grit9m_v3det_20231204_095047-b448804b.pth权重进行训练时，碰到了 File "/media/ubuntu/11632161-4946-47f0-8f2c-3fda15a5d5d7/projects/czy/mmdetection-main/mmdet/models/backbones/swin.py", line 728, in init_weights table_current = self.state_dict()[table_key] KeyError: 'backbone.stages.0.blocks.0.attn.w_msa.relative_position_bias_table' 的问题，请问该如何解决
In the example of grounding_dino_swin-t_finetune_8xb4_20e_cat.py, with the optim_wrapper = dict( optimizer=dict(lr=0.0001), paramwise_cfg=dict( custom_keys={ 'absolute_pos_embed': dict(decay_mult=0.), 'backbone': dict(lr_mult=0.0), 'language_model': dict(lr_mult=0.0) })), which parameters are updated during training, and how can other...
