open-mmlab/mmdetection · error · TypeError

pretrained must be a str or None

Error message

pretrained must be a str or None

What it means

ResNet.__init__ raises TypeError when the deprecated `pretrained` is neither a str nor None. A str becomes init_cfg=dict(type='Pretrained', ...); None means use init_cfg; other types are rejected, and passing both init_cfg and pretrained trips an assert.

Source

Thrown at mmdet/models/backbones/resnet.py:426

                    dict(
                        type='Constant',
                        val=1,
                        layer=['_BatchNorm', 'GroupNorm'])
                ]
                block = self.arch_settings[depth][0]
                if self.zero_init_residual:
                    if block is BasicBlock:
                        block_init_cfg = dict(
                            type='Constant',
                            val=0,
                            override=dict(name='norm2'))
                    elif block is Bottleneck:
                        block_init_cfg = dict(
                            type='Constant',
                            val=0,
                            override=dict(name='norm3'))
        else:
            raise TypeError('pretrained must be a str or None')

        self.depth = depth
        if stem_channels is None:
            stem_channels = base_channels
        self.stem_channels = stem_channels
        self.base_channels = base_channels
        self.num_stages = num_stages
        assert num_stages >= 1 and num_stages <= 4
        self.strides = strides
        self.dilations = dilations
        assert len(strides) == len(dilations) == num_stages
        self.out_indices = out_indices
        assert max(out_indices) < num_stages
        self.style = style
        self.deep_stem = deep_stem
        self.avg_down = avg_down
        self.frozen_stages = frozen_stages
        self.conv_cfg = conv_cfg

View on GitHub (pinned to cfd5d3a985)

Solutions

1. Use init_cfg=dict(type='Pretrained', checkpoint='...') and remove pretrained
2. If keeping pretrained, pass only a str path and remove init_cfg
3. Pass a Path to a checkpoint file, not a loaded state_dict

Example fix

// before
backbone=dict(type='ResNet', depth=50, pretrained='torchvision://resnet50', init_cfg=None)
// after
backbone=dict(type='ResNet', depth=50, init_cfg=dict(type='Pretrained', checkpoint='torchvision://resnet50'))

assert not (init_cfg and pretrained), 'cannot set both'\nassert pretrained is None or isinstance(pretrained, str)

When it happens

Trigger: ResNet(pretrained=dict(type='Pretrained', checkpoint=...)); ResNet(pretrained=True); combining pretrained and init_cfg.

Common situations: Legacy configs from mmdet v1/v2 using pretrained='torchvision://resnet50'; converting configs and leaving both keys; passing a loaded state_dict object instead of a path.

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/24c9c631f2e58fb5. Report an issue: GitHub.