训练出来的模型，为什么不能使用torch.load？

[Yzs007]

RuntimeError: Error(s) in loading state_dict for SAFMN: Missing key(s) in state_dict: "to_feat.weight", "to_feat.bias", "feats.0.norm1.weight", "feats.0.norm1.bias", "feats.0.norm2.weight", "feats.0.norm2.bias", "feats.0.safm.mfr.0.weight", "feats.0.safm.mfr.0.bias", "feats.0.safm.mfr.1.weight", "feats.0.safm.mfr.1.bias", "feats.0.safm.mfr.2.weight", "feats.0.safm.mfr.2.bias", "feats.0.safm.mfr.3.weight", "feats.0.safm.mfr.3.bias", "feats.0.safm.aggr.weight", "feats.0.safm.aggr.bias", "feats.0.ccm.ccm.0.weight", "feats.0.ccm.ccm.0.bias", "feats.0.ccm.ccm.2.weight", "feats.0.ccm.ccm.2.bias", Unexpected key(s) in state_dict: "params".

[wangxinchao-bit]

字典保存不一样需要 checkpoint = torch.load(path)["params"]