更换backbone求助

[coptersack]

十分感谢您的卓越工作，在更换backbone时，我已经将新的backbone通过register注册过了，新的config文件也添加了，加载模型时，core中的workspace中报错显示 Traceback (most recent call last): File "D:\Medical_segmentation\Kingmed\RT-DETR-main\rtdetrv2_pytorch\tools\train.py", line 71, in main(args) File "D:\Medical_segmentation\Kingmed\RT-DETR-main\rtdetrv2_pytorch\tools\train.py", line 35, in main solver.fit() File "D:\Medical_segmentation\Kingmed\RT-DETR-main\rtdetrv2_pytorch\tools..\src\solver\det_solver.py", line 20, in fit self.train() File "D:\Medical_segmentation\Kingmed\RT-DETR-main\rtdetrv2_pytorch\tools..\src\solver_solver.py", line 68, in train self._setup() File "D:\Medical_segmentation\Kingmed\RT-DETR-main\rtdetrv2_pytorch\tools..\src\solver_solver.py", line 35, in _setup self.model = cfg.model ^^^^^^^^^ File "D:\Medical_segmentation\Kingmed\RT-DETR-main\rtdetrv2_pytorch\tools..\src\core\yaml_config.py", line 36, in model self._model = create(self.yaml_cfg['model'], self.global_cfg) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "D:\Medical_segmentation\Kingmed\RT-DETR-main\rtdetrv2_pytorch\tools..\src\core\workspace.py", line 144, in create raise ValueError(f'Missing inject config of {_k}.') ValueError: Missing inject config of repvit_m2_3. 不知道是哪里缺少了步骤或者哪里出了问题，希望您百忙之余可以帮我解答

[lyuwenyu]

在 __init__.py文件import下你新增的backbone

[coptersack]

在 __init__.py文件import下你新增的backbone

感谢您的解答，但是这一步我也做了的

[lyuwenyu]

我看下你怎么注册的; v2版本和v1的实现有点差异

[coptersack]

class RepViT(nn.Module):
    def __init__(self, cfgs, distillation=False, pretrained=None, init_cfg=None, out_indices=[]):
        super(RepViT, self).__init__()
        # setting of inverted residual blocks
        self.cfgs = cfgs

        # building first layer
        input_channel = self.cfgs[0][2]
        patch_embed = torch.nn.Sequential(Conv2d_BN(3, input_channel // 2, 3, 2, 1), torch.nn.GELU(),
                                          Conv2d_BN(input_channel // 2, input_channel, 3, 2, 1))
        layers = [patch_embed]
        # building inverted residual blocks
        block = RepViTBlock
        for k, t, c, use_se, use_hs, s in self.cfgs:
            output_channel = _make_divisible(c, 8)
            exp_size = _make_divisible(input_channel * t, 8)
            layers.append(block(input_channel, exp_size, output_channel, k, s, use_se, use_hs))
            input_channel = output_channel
        self.features = nn.ModuleList(layers)

        self.init_cfg = init_cfg
        # assert (self.init_cfg is not None)
        self.out_indices = out_indices
        self._init_weights()
        self = torch.nn.SyncBatchNorm.convert_sync_batchnorm(self)
        self.train()

    def train(self, mode=True):
        """Convert the model into training mode while keep layers freezed."""
        super(RepViT, self).train(mode)
        if mode:
            for m in self.modules():
                if isinstance(m, _BatchNorm):
                    m.eval()

    def forward(self, x):
        outs = []
        for i, f in enumerate(self.features):
            x = f(x)
            if i in self.out_indices:
                outs.append(x)
        assert (len(outs) == 4)
        return outs[1:4]

@register
def repvit_m1_1(pretrained=False, distillation=False, init_cfg=None, out_indices=[], **kwargs):
    cfgs = [
        # k, t, c, SE, HS, s
    ]
    return RepViT(cfgs, init_cfg=init_cfg, pretrained=pretrained, distillation=distillation, out_indices=out_indices)

这是backbone/repvit中的注册代码，config就是在resnet50上加了

RTDETR:
  backbone: repvit_m1_1

十分感谢您在百忙之中还帮我解决这些低级问题，哈哈哈哈

[lyuwenyu]

更新下注册那行

# v1
@register

# to

# v2 
@register()

[coptersack]

感谢，已经解决了