如何从状态字典创建模型？
[英]How do I create a model from a state dict?
问题描述
I am trying to load a checkpoint pth file from the faster_rcnn_resnet101 model which is not currently in the PyTorch model zoo.
我正在尝试从当前不在 PyTorch 模型动物园中的 faster_rcnn_resnet101 模型加载检查点 pth 文件。 This causes PyTorch to throw a KeyError saying that I the layers in the state dict does not match the model architecture of faster_rcnn_fpn_resnet50 that I've loaded from the model zoo.这会导致 PyTorch 抛出一个 KeyError 说我状态字典中的层与我从模型动物园加载的 faster_rcnn_fpn_resnet50 的模型架构不匹配。
Note: I tried posting the architecture of the faster_rcnn_fpn_resnet50 model, but I went over the character limit of 30.000注意：我尝试发布 fast_rcnn_fpn_resnet50 模型的架构，但我超过了 30.000 的字符限制
print(state_dict[model].keys()):打印（状态字典[模型].keys（））：
backbone.stem.conv1.weight
backbone.stem.conv1.norm.weight
backbone.stem.conv1.norm.bias
backbone.stem.conv1.norm.running_mean
backbone.stem.conv1.norm.running_var
backbone.res2.0.shortcut.weight
backbone.res2.0.shortcut.norm.weight
backbone.res2.0.shortcut.norm.bias
backbone.res2.0.shortcut.norm.running_mean
backbone.res2.0.shortcut.norm.running_var
backbone.res2.0.conv1.weight
backbone.res2.0.conv1.norm.weight
backbone.res2.0.conv1.norm.bias
backbone.res2.0.conv1.norm.running_mean
backbone.res2.0.conv1.norm.running_var
backbone.res2.0.conv2.weight
backbone.res2.0.conv2.norm.weight
backbone.res2.0.conv2.norm.bias
backbone.res2.0.conv2.norm.running_mean
backbone.res2.0.conv2.norm.running_var
backbone.res2.0.conv3.weight
backbone.res2.0.conv3.norm.weight
backbone.res2.0.conv3.norm.bias
backbone.res2.0.conv3.norm.running_mean
backbone.res2.0.conv3.norm.running_var
backbone.res2.1.conv1.weight
backbone.res2.1.conv1.norm.weight
backbone.res2.1.conv1.norm.bias
backbone.res2.1.conv1.norm.running_mean
backbone.res2.1.conv1.norm.running_var
backbone.res2.1.conv2.weight
backbone.res2.1.conv2.norm.weight
backbone.res2.1.conv2.norm.bias
backbone.res2.1.conv2.norm.running_mean
backbone.res2.1.conv2.norm.running_var
backbone.res2.1.conv3.weight
backbone.res2.1.conv3.norm.weight
backbone.res2.1.conv3.norm.bias
backbone.res2.1.conv3.norm.running_mean
backbone.res2.1.conv3.norm.running_var
backbone.res2.2.conv1.weight
backbone.res2.2.conv1.norm.weight
backbone.res2.2.conv1.norm.bias
backbone.res2.2.conv1.norm.running_mean
backbone.res2.2.conv1.norm.running_var
backbone.res2.2.conv2.weight
backbone.res2.2.conv2.norm.weight
backbone.res2.2.conv2.norm.bias
backbone.res2.2.conv2.norm.running_mean
backbone.res2.2.conv2.norm.running_var
backbone.res2.2.conv3.weight
backbone.res2.2.conv3.norm.weight
backbone.res2.2.conv3.norm.bias
backbone.res2.2.conv3.norm.running_mean
backbone.res2.2.conv3.norm.running_var
backbone.res3.0.shortcut.weight
backbone.res3.0.shortcut.norm.weight
backbone.res3.0.shortcut.norm.bias
backbone.res3.0.shortcut.norm.running_mean
backbone.res3.0.shortcut.norm.running_var
backbone.res3.0.conv1.weight
backbone.res3.0.conv1.norm.weight
backbone.res3.0.conv1.norm.bias
backbone.res3.0.conv1.norm.running_mean
backbone.res3.0.conv1.norm.running_var
backbone.res3.0.conv2.weight
backbone.res3.0.conv2.norm.weight
backbone.res3.0.conv2.norm.bias
backbone.res3.0.conv2.norm.running_mean
backbone.res3.0.conv2.norm.running_var
backbone.res3.0.conv3.weight
backbone.res3.0.conv3.norm.weight
backbone.res3.0.conv3.norm.bias
backbone.res3.0.conv3.norm.running_mean
backbone.res3.0.conv3.norm.running_var
backbone.res3.1.conv1.weight
backbone.res3.1.conv1.norm.weight
backbone.res3.1.conv1.norm.bias
backbone.res3.1.conv1.norm.running_mean
backbone.res3.1.conv1.norm.running_var
backbone.res3.1.conv2.weight
backbone.res3.1.conv2.norm.weight
backbone.res3.1.conv2.norm.bias
backbone.res3.1.conv2.norm.running_mean
backbone.res3.1.conv2.norm.running_var
backbone.res3.1.conv3.weight
backbone.res3.1.conv3.norm.weight
backbone.res3.1.conv3.norm.bias
backbone.res3.1.conv3.norm.running_mean
backbone.res3.1.conv3.norm.running_var
backbone.res3.2.conv1.weight
backbone.res3.2.conv1.norm.weight
backbone.res3.2.conv1.norm.bias
backbone.res3.2.conv1.norm.running_mean
backbone.res3.2.conv1.norm.running_var
backbone.res3.2.conv2.weight
backbone.res3.2.conv2.norm.weight
backbone.res3.2.conv2.norm.bias
backbone.res3.2.conv2.norm.running_mean
backbone.res3.2.conv2.norm.running_var
backbone.res3.2.conv3.weight
backbone.res3.2.conv3.norm.weight
backbone.res3.2.conv3.norm.bias
backbone.res3.2.conv3.norm.running_mean
backbone.res3.2.conv3.norm.running_var
backbone.res3.3.conv1.weight
backbone.res3.3.conv1.norm.weight
backbone.res3.3.conv1.norm.bias
backbone.res3.3.conv1.norm.running_mean
backbone.res3.3.conv1.norm.running_var
backbone.res3.3.conv2.weight
backbone.res3.3.conv2.norm.weight
backbone.res3.3.conv2.norm.bias
backbone.res3.3.conv2.norm.running_mean
backbone.res3.3.conv2.norm.running_var
backbone.res3.3.conv3.weight
backbone.res3.3.conv3.norm.weight
backbone.res3.3.conv3.norm.bias
backbone.res3.3.conv3.norm.running_mean
backbone.res3.3.conv3.norm.running_var
backbone.res4.0.shortcut.weight
backbone.res4.0.shortcut.norm.weight
backbone.res4.0.shortcut.norm.bias
backbone.res4.0.shortcut.norm.running_mean
backbone.res4.0.shortcut.norm.running_var
backbone.res4.0.conv1.weight
backbone.res4.0.conv1.norm.weight
backbone.res4.0.conv1.norm.bias
backbone.res4.0.conv1.norm.running_mean
backbone.res4.0.conv1.norm.running_var
backbone.res4.0.conv2.weight
backbone.res4.0.conv2.norm.weight
backbone.res4.0.conv2.norm.bias
backbone.res4.0.conv2.norm.running_mean
backbone.res4.0.conv2.norm.running_var
backbone.res4.0.conv3.weight
backbone.res4.0.conv3.norm.weight
backbone.res4.0.conv3.norm.bias
backbone.res4.0.conv3.norm.running_mean
backbone.res4.0.conv3.norm.running_var
backbone.res4.1.conv1.weight
backbone.res4.1.conv1.norm.weight
backbone.res4.1.conv1.norm.bias
backbone.res4.1.conv1.norm.running_mean
backbone.res4.1.conv1.norm.running_var
backbone.res4.1.conv2.weight
backbone.res4.1.conv2.norm.weight
backbone.res4.1.conv2.norm.bias
backbone.res4.1.conv2.norm.running_mean
backbone.res4.1.conv2.norm.running_var
backbone.res4.1.conv3.weight
backbone.res4.1.conv3.norm.weight
backbone.res4.1.conv3.norm.bias
backbone.res4.1.conv3.norm.running_mean
backbone.res4.1.conv3.norm.running_var
backbone.res4.2.conv1.weight
backbone.res4.2.conv1.norm.weight
backbone.res4.2.conv1.norm.bias
backbone.res4.2.conv1.norm.running_mean
backbone.res4.2.conv1.norm.running_var
backbone.res4.2.conv2.weight
backbone.res4.2.conv2.norm.weight
backbone.res4.2.conv2.norm.bias
backbone.res4.2.conv2.norm.running_mean
backbone.res4.2.conv2.norm.running_var
backbone.res4.2.conv3.weight
backbone.res4.2.conv3.norm.weight
backbone.res4.2.conv3.norm.bias
backbone.res4.2.conv3.norm.running_mean
backbone.res4.2.conv3.norm.running_var
backbone.res4.3.conv1.weight
backbone.res4.3.conv1.norm.weight
backbone.res4.3.conv1.norm.bias
backbone.res4.3.conv1.norm.running_mean
backbone.res4.3.conv1.norm.running_var
backbone.res4.3.conv2.weight
backbone.res4.3.conv2.norm.weight
backbone.res4.3.conv2.norm.bias
backbone.res4.3.conv2.norm.running_mean
backbone.res4.3.conv2.norm.running_var
backbone.res4.3.conv3.weight
backbone.res4.3.conv3.norm.weight
backbone.res4.3.conv3.norm.bias
backbone.res4.3.conv3.norm.running_mean
backbone.res4.3.conv3.norm.running_var
backbone.res4.4.conv1.weight
backbone.res4.4.conv1.norm.weight
backbone.res4.4.conv1.norm.bias
backbone.res4.4.conv1.norm.running_mean
backbone.res4.4.conv1.norm.running_var
backbone.res4.4.conv2.weight
backbone.res4.4.conv2.norm.weight
backbone.res4.4.conv2.norm.bias
backbone.res4.4.conv2.norm.running_mean
backbone.res4.4.conv2.norm.running_var
backbone.res4.4.conv3.weight
backbone.res4.4.conv3.norm.weight
backbone.res4.4.conv3.norm.bias
backbone.res4.4.conv3.norm.running_mean
backbone.res4.4.conv3.norm.running_var
backbone.res4.5.conv1.weight
backbone.res4.5.conv1.norm.weight
backbone.res4.5.conv1.norm.bias
backbone.res4.5.conv1.norm.running_mean
backbone.res4.5.conv1.norm.running_var
backbone.res4.5.conv2.weight
backbone.res4.5.conv2.norm.weight
backbone.res4.5.conv2.norm.bias
backbone.res4.5.conv2.norm.running_mean
backbone.res4.5.conv2.norm.running_var
backbone.res4.5.conv3.weight
backbone.res4.5.conv3.norm.weight
backbone.res4.5.conv3.norm.bias
backbone.res4.5.conv3.norm.running_mean
backbone.res4.5.conv3.norm.running_var
backbone.res4.6.conv1.weight
backbone.res4.6.conv1.norm.weight
backbone.res4.6.conv1.norm.bias
backbone.res4.6.conv1.norm.running_mean
backbone.res4.6.conv1.norm.running_var
backbone.res4.6.conv2.weight
backbone.res4.6.conv2.norm.weight
backbone.res4.6.conv2.norm.bias
backbone.res4.6.conv2.norm.running_mean
backbone.res4.6.conv2.norm.running_var
backbone.res4.6.conv3.weight
backbone.res4.6.conv3.norm.weight
backbone.res4.6.conv3.norm.bias
backbone.res4.6.conv3.norm.running_mean
backbone.res4.6.conv3.norm.running_var
backbone.res4.7.conv1.weight
backbone.res4.7.conv1.norm.weight
backbone.res4.7.conv1.norm.bias
backbone.res4.7.conv1.norm.running_mean
backbone.res4.7.conv1.norm.running_var
backbone.res4.7.conv2.weight
backbone.res4.7.conv2.norm.weight
backbone.res4.7.conv2.norm.bias
backbone.res4.7.conv2.norm.running_mean
backbone.res4.7.conv2.norm.running_var
backbone.res4.7.conv3.weight
backbone.res4.7.conv3.norm.weight
backbone.res4.7.conv3.norm.bias
backbone.res4.7.conv3.norm.running_mean
backbone.res4.7.conv3.norm.running_var
backbone.res4.8.conv1.weight
backbone.res4.8.conv1.norm.weight
backbone.res4.8.conv1.norm.bias
backbone.res4.8.conv1.norm.running_mean
backbone.res4.8.conv1.norm.running_var
backbone.res4.8.conv2.weight
backbone.res4.8.conv2.norm.weight
backbone.res4.8.conv2.norm.bias
backbone.res4.8.conv2.norm.running_mean
backbone.res4.8.conv2.norm.running_var
backbone.res4.8.conv3.weight
backbone.res4.8.conv3.norm.weight
backbone.res4.8.conv3.norm.bias
backbone.res4.8.conv3.norm.running_mean
backbone.res4.8.conv3.norm.running_var
backbone.res4.9.conv1.weight
backbone.res4.9.conv1.norm.weight
backbone.res4.9.conv1.norm.bias
backbone.res4.9.conv1.norm.running_mean
backbone.res4.9.conv1.norm.running_var
backbone.res4.9.conv2.weight
backbone.res4.9.conv2.norm.weight
backbone.res4.9.conv2.norm.bias
backbone.res4.9.conv2.norm.running_mean
backbone.res4.9.conv2.norm.running_var
backbone.res4.9.conv3.weight
backbone.res4.9.conv3.norm.weight
backbone.res4.9.conv3.norm.bias
backbone.res4.9.conv3.norm.running_mean
backbone.res4.9.conv3.norm.running_var
backbone.res4.10.conv1.weight
backbone.res4.10.conv1.norm.weight
backbone.res4.10.conv1.norm.bias
backbone.res4.10.conv1.norm.running_mean
backbone.res4.10.conv1.norm.running_var
backbone.res4.10.conv2.weight
backbone.res4.10.conv2.norm.weight
backbone.res4.10.conv2.norm.bias
backbone.res4.10.conv2.norm.running_mean
backbone.res4.10.conv2.norm.running_var
backbone.res4.10.conv3.weight
backbone.res4.10.conv3.norm.weight
backbone.res4.10.conv3.norm.bias
backbone.res4.10.conv3.norm.running_mean
backbone.res4.10.conv3.norm.running_var
backbone.res4.11.conv1.weight
backbone.res4.11.conv1.norm.weight
backbone.res4.11.conv1.norm.bias
backbone.res4.11.conv1.norm.running_mean
backbone.res4.11.conv1.norm.running_var
backbone.res4.11.conv2.weight
backbone.res4.11.conv2.norm.weight
backbone.res4.11.conv2.norm.bias
backbone.res4.11.conv2.norm.running_mean
backbone.res4.11.conv2.norm.running_var
backbone.res4.11.conv3.weight
backbone.res4.11.conv3.norm.weight
backbone.res4.11.conv3.norm.bias
backbone.res4.11.conv3.norm.running_mean
backbone.res4.11.conv3.norm.running_var
backbone.res4.12.conv1.weight
backbone.res4.12.conv1.norm.weight
backbone.res4.12.conv1.norm.bias
backbone.res4.12.conv1.norm.running_mean
backbone.res4.12.conv1.norm.running_var
backbone.res4.12.conv2.weight
backbone.res4.12.conv2.norm.weight
backbone.res4.12.conv2.norm.bias
backbone.res4.12.conv2.norm.running_mean
backbone.res4.12.conv2.norm.running_var
backbone.res4.12.conv3.weight
backbone.res4.12.conv3.norm.weight
backbone.res4.12.conv3.norm.bias
backbone.res4.12.conv3.norm.running_mean
backbone.res4.12.conv3.norm.running_var
backbone.res4.13.conv1.weight
backbone.res4.13.conv1.norm.weight
backbone.res4.13.conv1.norm.bias
backbone.res4.13.conv1.norm.running_mean
backbone.res4.13.conv1.norm.running_var
backbone.res4.13.conv2.weight
backbone.res4.13.conv2.norm.weight
backbone.res4.13.conv2.norm.bias
backbone.res4.13.conv2.norm.running_mean
backbone.res4.13.conv2.norm.running_var
backbone.res4.13.conv3.weight
backbone.res4.13.conv3.norm.weight
backbone.res4.13.conv3.norm.bias
backbone.res4.13.conv3.norm.running_mean
backbone.res4.13.conv3.norm.running_var
backbone.res4.14.conv1.weight
backbone.res4.14.conv1.norm.weight
backbone.res4.14.conv1.norm.bias
backbone.res4.14.conv1.norm.running_mean
backbone.res4.14.conv1.norm.running_var
backbone.res4.14.conv2.weight
backbone.res4.14.conv2.norm.weight
backbone.res4.14.conv2.norm.bias
backbone.res4.14.conv2.norm.running_mean
backbone.res4.14.conv2.norm.running_var
backbone.res4.14.conv3.weight
backbone.res4.14.conv3.norm.weight
backbone.res4.14.conv3.norm.bias
backbone.res4.14.conv3.norm.running_mean
backbone.res4.14.conv3.norm.running_var
backbone.res4.15.conv1.weight
backbone.res4.15.conv1.norm.weight
backbone.res4.15.conv1.norm.bias
backbone.res4.15.conv1.norm.running_mean
backbone.res4.15.conv1.norm.running_var
backbone.res4.15.conv2.weight
backbone.res4.15.conv2.norm.weight
backbone.res4.15.conv2.norm.bias
backbone.res4.15.conv2.norm.running_mean
backbone.res4.15.conv2.norm.running_var
backbone.res4.15.conv3.weight
backbone.res4.15.conv3.norm.weight
backbone.res4.15.conv3.norm.bias
backbone.res4.15.conv3.norm.running_mean
backbone.res4.15.conv3.norm.running_var
backbone.res4.16.conv1.weight
backbone.res4.16.conv1.norm.weight
backbone.res4.16.conv1.norm.bias
backbone.res4.16.conv1.norm.running_mean
backbone.res4.16.conv1.norm.running_var
backbone.res4.16.conv2.weight
backbone.res4.16.conv2.norm.weight
backbone.res4.16.conv2.norm.bias
backbone.res4.16.conv2.norm.running_mean
backbone.res4.16.conv2.norm.running_var
backbone.res4.16.conv3.weight
backbone.res4.16.conv3.norm.weight
backbone.res4.16.conv3.norm.bias
backbone.res4.16.conv3.norm.running_mean
backbone.res4.16.conv3.norm.running_var
backbone.res4.17.conv1.weight
backbone.res4.17.conv1.norm.weight
backbone.res4.17.conv1.norm.bias
backbone.res4.17.conv1.norm.running_mean
backbone.res4.17.conv1.norm.running_var
backbone.res4.17.conv2.weight
backbone.res4.17.conv2.norm.weight
backbone.res4.17.conv2.norm.bias
backbone.res4.17.conv2.norm.running_mean
backbone.res4.17.conv2.norm.running_var
backbone.res4.17.conv3.weight
backbone.res4.17.conv3.norm.weight
backbone.res4.17.conv3.norm.bias
backbone.res4.17.conv3.norm.running_mean
backbone.res4.17.conv3.norm.running_var
backbone.res4.18.conv1.weight
backbone.res4.18.conv1.norm.weight
backbone.res4.18.conv1.norm.bias
backbone.res4.18.conv1.norm.running_mean
backbone.res4.18.conv1.norm.running_var
backbone.res4.18.conv2.weight
backbone.res4.18.conv2.norm.weight
backbone.res4.18.conv2.norm.bias
backbone.res4.18.conv2.norm.running_mean
backbone.res4.18.conv2.norm.running_var
backbone.res4.18.conv3.weight
backbone.res4.18.conv3.norm.weight
backbone.res4.18.conv3.norm.bias
backbone.res4.18.conv3.norm.running_mean
backbone.res4.18.conv3.norm.running_var
backbone.res4.19.conv1.weight
backbone.res4.19.conv1.norm.weight
backbone.res4.19.conv1.norm.bias
backbone.res4.19.conv1.norm.running_mean
backbone.res4.19.conv1.norm.running_var
backbone.res4.19.conv2.weight
backbone.res4.19.conv2.norm.weight
backbone.res4.19.conv2.norm.bias
backbone.res4.19.conv2.norm.running_mean
backbone.res4.19.conv2.norm.running_var
backbone.res4.19.conv3.weight
backbone.res4.19.conv3.norm.weight
backbone.res4.19.conv3.norm.bias
backbone.res4.19.conv3.norm.running_mean
backbone.res4.19.conv3.norm.running_var
backbone.res4.20.conv1.weight
backbone.res4.20.conv1.norm.weight
backbone.res4.20.conv1.norm.bias
backbone.res4.20.conv1.norm.running_mean
backbone.res4.20.conv1.norm.running_var
backbone.res4.20.conv2.weight
backbone.res4.20.conv2.norm.weight
backbone.res4.20.conv2.norm.bias
backbone.res4.20.conv2.norm.running_mean
backbone.res4.20.conv2.norm.running_var
backbone.res4.20.conv3.weight
backbone.res4.20.conv3.norm.weight
backbone.res4.20.conv3.norm.bias
backbone.res4.20.conv3.norm.running_mean
backbone.res4.20.conv3.norm.running_var
backbone.res4.21.conv1.weight
backbone.res4.21.conv1.norm.weight
backbone.res4.21.conv1.norm.bias
backbone.res4.21.conv1.norm.running_mean
backbone.res4.21.conv1.norm.running_var
backbone.res4.21.conv2.weight
backbone.res4.21.conv2.norm.weight
backbone.res4.21.conv2.norm.bias
backbone.res4.21.conv2.norm.running_mean
backbone.res4.21.conv2.norm.running_var
backbone.res4.21.conv3.weight
backbone.res4.21.conv3.norm.weight
backbone.res4.21.conv3.norm.bias
backbone.res4.21.conv3.norm.running_mean
backbone.res4.21.conv3.norm.running_var
backbone.res4.22.conv1.weight
backbone.res4.22.conv1.norm.weight
backbone.res4.22.conv1.norm.bias
backbone.res4.22.conv1.norm.running_mean
backbone.res4.22.conv1.norm.running_var
backbone.res4.22.conv2.weight
backbone.res4.22.conv2.norm.weight
backbone.res4.22.conv2.norm.bias
backbone.res4.22.conv2.norm.running_mean
backbone.res4.22.conv2.norm.running_var
backbone.res4.22.conv3.weight
backbone.res4.22.conv3.norm.weight
backbone.res4.22.conv3.norm.bias
backbone.res4.22.conv3.norm.running_mean
backbone.res4.22.conv3.norm.running_var
proposal_generator.rpn_head.conv.weight
proposal_generator.rpn_head.conv.bias
proposal_generator.rpn_head.objectness_logits.weight
proposal_generator.rpn_head.objectness_logits.bias
proposal_generator.rpn_head.anchor_deltas.weight
proposal_generator.rpn_head.anchor_deltas.bias
roi_heads.res5.0.shortcut.weight
roi_heads.res5.0.shortcut.norm.weight
roi_heads.res5.0.shortcut.norm.bias
roi_heads.res5.0.shortcut.norm.running_mean
roi_heads.res5.0.shortcut.norm.running_var
roi_heads.res5.0.conv1.weight
roi_heads.res5.0.conv1.norm.weight
roi_heads.res5.0.conv1.norm.bias
roi_heads.res5.0.conv1.norm.running_mean
roi_heads.res5.0.conv1.norm.running_var
roi_heads.res5.0.conv2.weight
roi_heads.res5.0.conv2.norm.weight
roi_heads.res5.0.conv2.norm.bias
roi_heads.res5.0.conv2.norm.running_mean
roi_heads.res5.0.conv2.norm.running_var
roi_heads.res5.0.conv3.weight
roi_heads.res5.0.conv3.norm.weight
roi_heads.res5.0.conv3.norm.bias
roi_heads.res5.0.conv3.norm.running_mean
roi_heads.res5.0.conv3.norm.running_var
roi_heads.res5.1.conv1.weight
roi_heads.res5.1.conv1.norm.weight
roi_heads.res5.1.conv1.norm.bias
roi_heads.res5.1.conv1.norm.running_mean
roi_heads.res5.1.conv1.norm.running_var
roi_heads.res5.1.conv2.weight
roi_heads.res5.1.conv2.norm.weight
roi_heads.res5.1.conv2.norm.bias
roi_heads.res5.1.conv2.norm.running_mean
roi_heads.res5.1.conv2.norm.running_var
roi_heads.res5.1.conv3.weight
roi_heads.res5.1.conv3.norm.weight
roi_heads.res5.1.conv3.norm.bias
roi_heads.res5.1.conv3.norm.running_mean
roi_heads.res5.1.conv3.norm.running_var
roi_heads.res5.2.conv1.weight
roi_heads.res5.2.conv1.norm.weight
roi_heads.res5.2.conv1.norm.bias
roi_heads.res5.2.conv1.norm.running_mean
roi_heads.res5.2.conv1.norm.running_var
roi_heads.res5.2.conv2.weight
roi_heads.res5.2.conv2.norm.weight
roi_heads.res5.2.conv2.norm.bias
roi_heads.res5.2.conv2.norm.running_mean
roi_heads.res5.2.conv2.norm.running_var
roi_heads.res5.2.conv3.weight
roi_heads.res5.2.conv3.norm.weight
roi_heads.res5.2.conv3.norm.bias
roi_heads.res5.2.conv3.norm.running_mean
roi_heads.res5.2.conv3.norm.running_var
roi_heads.box_predictor.cls_score.weight
roi_heads.box_predictor.cls_score.bias
roi_heads.box_predictor.bbox_pred.weight
roi_heads.box_predictor.bbox_pred.bias
Is there a way to dynamically create the model architecture of the faster_rcnn_resnet101 model in torch by iterating over the model dict or is there another way?有没有办法通过迭代模型字典在torch中动态创建faster_rcnn_resnet101模型的模型架构，还是有其他方法？
I have tried searching for a GitHub repository with a hubconf.py for torch.hub, but to no avail.我曾尝试使用 hubconf.py 为 torch.hub 搜索 GitHub 存储库，但无济于事。
1 个解决方案
解决方案1
1 已采纳 2022-07-05 09:56:02
TLDR; TLDR； the short answer is you can't.简短的回答是你不能。
The state_dict of a nn.Module contains the module's state, but not its function. state_dict的nn.Module包含模块的状态，但不包含其功能。 It is not possible to instantiate a model with its state dictionary alone.仅使用状态字典来实例化模型是不可能的。
You can manage to initialize some of the sub-modules based on the state dict keys and the weight shapes, but this is not guaranteed.您可以根据状态字典键和权重形状设法初始化一些子模块，但这不能保证。 In any case you won't be able to know the model's behaviour because this information is simply not contained in its state dictionary.在任何情况下，您都无法知道模型的行为，因为此信息根本不包含在其状态字典中。 Therefore you are required to have access to the forward definition of the model in order to know the forward logic, ie which functions are applied on the input and intermediate outputs and in which order those submodules are used.因此，您需要访问模型的forward定义以了解前向逻辑，即哪些函数应用于输入和中间输出以及这些子模块的使用顺序。
Have a look at this minimal example showing two models with identical state dictionaries.看看这个最小的例子，展示了两个具有相同状态字典的模型。 However, their forward is different from one another:然而，他们的forward却各不相同：
class A(nn.Linear):
def forward(self, x):
return super().forward(x)
class B(nn.Linear):
def forward(self, x):
return super().forward(x)**2
Here, a and b are initialized and the state dict is copied from one to the other:在这里， a和b被初始化，状态字典从一个复制到另一个：
>>> a = A(2,1)
>>> b = B(2,1)
>>> a.load_state_dict(b.state_dict())
Both a and b have the exact same state... but actually implement two different functions! a和b都具有完全相同的状态......但实际上实现了两个不同的功能！
声明:本站的技术帖子网页，遵循CC BY-SA 4.0协议，如果您需要转载，请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.
相关问题
如何从列表和csv文件创建[dict] - How do I create [dict] from a list and csv file
如何从 SQL 查询的结果创建值的字典？ - How do I create a dict of values from the results of an SQL query?
如何通过在 python 中的分隔符上拆分键来从 dict 创建嵌套的 dict？ - How do I create a nested dict from a dict by splitting the keys on a delimiter in python?
如何从Python中带有嵌套字典的字典中创建新的字典 - How do I create a new dict of dicts from a dict with nested dicts in Python
我如何从 csv 文件中获取 dict 作为 dict - how do i get a dict as dict back from csv fille
我如何从_ast.Dict变成实际的dict？ - how do i go from an _ast.Dict to an actual dict?
如何从 JSON 字典中按月或年过滤 Django Model？ - How do I filter a Django Model by month or year from a JSON dict?
如何创建模型，使其要么是空字符串，要么是 pydantic 中的 Dict - How can I create model such that either it is empty string or Dict in pydantic
如何从字典填充StructuredProperty？ - How do I populate a StructuredProperty from a dict?
如何从函数返回字典？ - How do I return a dict from a function?
相关标签