lllyasviel/ControlNet · error · ValueError

inverted_residual_setting should be non-empty or a 4-element

Error message

inverted_residual_setting should be non-empty or a 4-element list, got {}

What it means

MobileNetV2-based MLSd large model constructor validates inverted_residual_setting: it must be a non-empty list whose (first) entries are 4-element tuples (t, c, n, s) describing each inverted residual block. Empty lists or wrong-length tuples raise this ValueError.

Source

Thrown at annotator/mlsd/models/mbv2_mlsd_large.py:186

        input_channel = 32
        last_channel = 1280
        width_mult = 1.0
        round_nearest = 8

        inverted_residual_setting = [
            # t, c, n, s
            [1, 16, 1, 1],
            [6, 24, 2, 2],
            [6, 32, 3, 2],
            [6, 64, 4, 2],
            [6, 96, 3, 1],
            #[6, 160, 3, 2],
            #[6, 320, 1, 1],
        ]

        # only check the first element, assuming user knows t,c,n,s are required
        if len(inverted_residual_setting) == 0 or len(inverted_residual_setting[0]) != 4:
            raise ValueError("inverted_residual_setting should be non-empty "
                             "or a 4-element list, got {}".format(inverted_residual_setting))

        # building first layer
        input_channel = _make_divisible(input_channel * width_mult, round_nearest)
        self.last_channel = _make_divisible(last_channel * max(1.0, width_mult), round_nearest)
        features = [ConvBNReLU(4, input_channel, stride=2)]
        # building inverted residual blocks
        for t, c, n, s in inverted_residual_setting:
            output_channel = _make_divisible(c * width_mult, round_nearest)
            for i in range(n):
                stride = s if i == 0 else 1
                features.append(block(input_channel, output_channel, stride, expand_ratio=t))
                input_channel = output_channel

        self.features = nn.Sequential(*features)
        self.fpn_selected = [1, 3, 6, 10, 13]
        # weight initialization
        for m in self.modules():

View on GitHub (pinned to ed85cd1e25)

Solutions

1. Keep at least one full [t, c, n, s] entry in inverted_residual_setting
2. Revert to the file's default inverted_residual_setting if you modified it
3. Validate the setting length programmatically before constructing the model

Example fix

# before
setting = []  # all rows commented out
# after
setting = [[1, 16, 1, 1], [6, 24, 2, 2]]  # proper [t,c,n,s] rows

assert len(inverted_residual_setting) > 0 and all(len(r) == 4 for r in inverted_residual_setting), 'need non-empty list of [t,c,n,s] rows'

def is_valid_setting(s) -> bool:
    return isinstance(s, (list, tuple)) and len(s) > 0 and all(len(r) == 4 for r in s)

When it happens

Trigger: Constructing the MLSd large network with an overridden inverted_residual_setting that is [] or whose first element has more/fewer than 4 values (e.g. commented-out rows left in a modified copy of the default settings list).

Common situations: Editing the default layer table to trim layers and accidentally removing all entries; passing a config dict value that deserializes to an empty list.

Related errors

• inverted_residual_setting should be non-empty or a 4-element
• resize_method {self.__resize_method} not implemented
• provide num_res_blocks either as an int (globally constant)
• 'order' must be '1' or '2' or '3'.
• 'solver_type' must be either 'dpm_solver' or 'taylor', got {

AI-assisted analysis of lllyasviel/ControlNet@ed85cd1e25 (2026-08-27). Data as JSON: /api/errors/caaeacc8629e1d9b. Report an issue: GitHub.