open-mmlab/mmdetection · error · TypeError

pretrained must be a str or None

Error message

pretrained must be a str or None

What it means

RegNet (inherits ResNet-style init) raises TypeError when the deprecated `pretrained` argument is not a str or None. Only a checkpoint path string or None is accepted; use init_cfg for new code.

Source

Thrown at mmdet/models/backbones/regnet.py:193

            'init_cfg and pretrained cannot be specified at the same time'
        if 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:
            if init_cfg is None:
                self.init_cfg = [
                    dict(type='Kaiming', layer='Conv2d'),
                    dict(
                        type='Constant',
                        val=1,
                        layer=['_BatchNorm', 'GroupNorm'])
                ]
                if self.zero_init_residual:
                    block_init_cfg = dict(
                        type='Constant', val=0, override=dict(name='norm3'))
        else:
            raise TypeError('pretrained must be a str or None')

        self.inplanes = stem_channels
        self.res_layers = []
        for i, num_blocks in enumerate(self.stage_blocks):
            stride = self.strides[i]
            dilation = self.dilations[i]
            group_width = self.group_widths[i]
            width = int(round(self.stage_widths[i] * self.bottleneck_ratio[i]))
            stage_groups = width // group_width

            dcn = self.dcn if self.stage_with_dcn[i] else None
            if self.plugins is not None:
                stage_plugins = self.make_stage_plugins(self.plugins, i)
            else:
                stage_plugins = None

            res_layer = self.make_res_layer(
                block=self.block,

View on GitHub (pinned to cfd5d3a985)

Solutions

1. Pass pretrained as a str checkpoint path or None
2. Better: use init_cfg=dict(type='Pretrained', checkpoint='regnetx_400mf.pth') and drop pretrained

Example fix

// before
backbone=dict(type='RegNet', arch='regnetx_400mf', pretrained='open-mmlab://regnetx_400mf')
// after
backbone=dict(type='RegNet', arch='regnetx_400mf', init_cfg=dict(type='Pretrained', checkpoint='regnetx_400mf_8xb128-...pth'))

assert pretrained is None or isinstance(pretrained, str)

When it happens

Trigger: RegNet(pretrained=dict(...)) or pretrained=Path(...); mixing pretrained and init_cfg (which trips a separate assert).

Common situations: Porting old mmdet configs using pretrained='open-mmlab://detectron2/resnet50_caffe' style entries to RegNet; passing the whole init_cfg dict into the pretrained slot by mistake.

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/634f788fc52d94f0. Report an issue: GitHub.