open-mmlab/mmdetection · warning

DeprecationWarning: pretrained is deprecated, please use "in

Error message

DeprecationWarning: pretrained is deprecated, please use "init_cfg" instead

What it means

Darknet backbone __init__ warns that the 'pretrained' string argument is deprecated in favor of 'init_cfg'. The model still builds — the checkpoint is converted into init_cfg=dict(type='Pretrained', checkpoint=...) — but the old API may be removed later.

Source

Thrown at mmdet/models/backbones/darknet.py:138

        cfg = dict(conv_cfg=conv_cfg, norm_cfg=norm_cfg, act_cfg=act_cfg)

        self.conv1 = ConvModule(3, 32, 3, padding=1, **cfg)

        self.cr_blocks = ['conv1']
        for i, n_layers in enumerate(self.layers):
            layer_name = f'conv_res_block{i + 1}'
            in_c, out_c = self.channels[i]
            self.add_module(
                layer_name,
                self.make_conv_res_block(in_c, out_c, n_layers, **cfg))
            self.cr_blocks.append(layer_name)

        self.norm_eval = norm_eval

        assert not (init_cfg and pretrained), \
            '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'])
                ]
        else:
            raise TypeError('pretrained must be a str or None')

    def forward(self, x):
        outs = []
        for i, layer_name in enumerate(self.cr_blocks):
            cr_block = getattr(self, layer_name)

View on GitHub (pinned to cfd5d3a985)

Solutions

1. Replace pretrained='path_or_url' with init_cfg=dict(type='Pretrained', checkpoint='path_or_url') in the backbone config
2. Note both cannot be set simultaneously (assertion forbids init_cfg + pretrained together)
3. Update any saved config pipelines/scripts that generate model dicts

Example fix

# before
model = dict(backbone=dict(type='Darknet', pretrained='open-mmlab://darknet53'))
# after
model = dict(backbone=dict(type='Darknet', init_cfg=dict(type='Pretrained', checkpoint='open-mmlab://darknet53')))

cfg = dict(type='Darknet')
assert 'pretrained' not in cfg, 'use init_cfg instead of pretrained'

When it happens

Trigger: Instantiating Darknet(pretrained='open-mmlab://darknet53') or a config using pretrained=... in model backbone init args.

Common situations: Old MMDetection 2.x configs or tutorials passed backbone=dict(type='Darknet', pretrained='...'). New configs should use init_cfg.

Related errors

• invalid depth {depth} for darknet
• pretrained must be a str or None
• frozen_stages must be in range(-1, len(arch_setting) + 1). B
• frozen_stages must be in range(-1, len(arch_setting) + 1). B
• `init_cfg` must contain the key "type"

AI-assisted analysis of open-mmlab/mmdetection@cfd5d3a985 (2026-08-27). Data as JSON: /api/errors/3451dcee7d2106c0. Report an issue: GitHub.