ecjtujemmy
commented
2 years ago 谢谢您的解答，可是我在加载预模型时又出现问题，报错“No pretrained weights exist for this model. Using random initialization.”您知道是怎么回事吗 def cmt_b(pretrained=False, kwargs): """ CMT-Base """ model_kwargs = dict( qkv_bias=True, embed_dims=[76, 152, 304, 608], stem_channel=38, num_heads=[1, 2, 4, 8], depths=[4, 4, 20, 4], mlp_ratios=[4, 4, 4, 4], qk_ratio=1, sr_ratios=[8, 4, 2, 1], dp=0.3, kwargs) model = _create_cmt_model(pretrained=pretrained, **model_kwargs) if pretrained: checkpoint = torch.load('D:\preweight\cmt_base.pth') # todo pass path as argument model.load_state_dict(checkpoint, strict=False) print("load CMT pretrained") return model
https://github.com/huawei-noah/Efficient-AI-Backbones/tree/master/cmt_pytorch#cmt-on-imagenet-1k-classification