open-mmlab/mmdetection · error · TypeError

pretrained must be a str or None

Error message

pretrained must be a str or None

What it means

SSDVGG.__init__ raises TypeError when the deprecated `pretrained` is neither a str nor None. Like other mmdet backbones, str is converted to init_cfg and new code should use init_cfg directly.

Source

Thrown at mmdet/models/backbones/ssd_vgg.py:97

        self.out_feature_indices = out_feature_indices

        assert not (init_cfg and pretrained), \
            'init_cfg and pretrained cannot be specified at the same time'

        if init_cfg is not None:
            self.init_cfg = init_cfg
        elif isinstance(pretrained, str):
            warnings.warn('DeprecationWarning: pretrained is deprecated, '
                          'please use "init_cfg" instead')
            self.init_cfg = dict(type='Pretrained', checkpoint=pretrained)
        elif pretrained is None:
            self.init_cfg = [
                dict(type='Kaiming', layer='Conv2d'),
                dict(type='Constant', val=1, layer='BatchNorm2d'),
                dict(type='Normal', std=0.01, layer='Linear'),
            ]
        else:
            raise TypeError('pretrained must be a str or None')

        if input_size is not None:
            warnings.warn('DeprecationWarning: input_size is deprecated')
        if l2_norm_scale is not None:
            warnings.warn('DeprecationWarning: l2_norm_scale in VGG is '
                          'deprecated, it has been moved to SSDNeck.')

    def init_weights(self, pretrained=None):
        super(VGG, self).init_weights()

    def forward(self, x):
        """Forward function."""
        outs = []
        for i, layer in enumerate(self.features):
            x = layer(x)
            if i in self.out_feature_indices:
                outs.append(x)

View on GitHub (pinned to cfd5d3a985)

Solutions

1. Use init_cfg=dict(type='Pretrained', checkpoint='vgg16_caffe-...pth')
2. Or pass pretrained as a checkpoint path string only

Example fix

// before
backbone=dict(type='SSDVGG', depth=16, pretrained='open-mmlab://vgg16_caffe')
// after
backbone=dict(type='SSDVGG', depth=16, with_last_pool=False, init_cfg=dict(type='Pretrained', checkpoint='vgg16_caffe-...pth'))

assert pretrained is None or isinstance(pretrained, str)

When it happens

Trigger: SSDVGG(pretrained=dict(checkpoint=...)) or pretrained=Path; also triggered by passing pretrained alongside init_cfg in older code paths.

Common situations: Old SSD configs (v1-era) using pretrained='open-mmlab://vgg16_caffe'; migrating SSD pipelines to the init_cfg API.

Related errors

• pretrained must be a str or None
• pretrained must be a str or None
• pretrained must be a str or None
• pretrained must be a str or None
• pretrained must be a str or None

AI-assisted analysis of open-mmlab/mmdetection@cfd5d3a985 (2026-08-27). Data as JSON: /api/errors/bfe04aeb08510db2. Report an issue: GitHub.