huggingface/pytorch-image-models · error · RuntimeError

features_only not implemented for Vision Transformer models.

Error message

features_only not implemented for Vision Transformer models.

What it means

The visformer factories explicitly reject features_only=True because Visformer's feature extraction isn't wired through build_model_with_cfg's feature-only pathway; a RuntimeError (not ValueError) is raised at model creation.

Source

Thrown at timm/models/visformer.py:485

            x = self.stage3(x)

        x = self.norm(x)
        return x

    def forward_head(self, x, pre_logits: bool = False):
        x = self.global_pool(x)
        x = self.head_drop(x)
        return x if pre_logits else self.head(x)

    def forward(self, x):
        x = self.forward_features(x)
        x = self.forward_head(x)
        return x


def _create_visformer(variant, pretrained=False, default_cfg=None, **kwargs):
    if kwargs.get('features_only', None):
        raise RuntimeError('features_only not implemented for Vision Transformer models.')
    model = build_model_with_cfg(Visformer, variant, pretrained, **kwargs)
    return model


def _cfg(url='', **kwargs):
    return {
        'url': url,
        'num_classes': 1000, 'input_size': (3, 224, 224), 'pool_size': (7, 7),
        'crop_pct': .9, 'interpolation': 'bicubic', 'fixed_input_size': True,
        'mean': IMAGENET_DEFAULT_MEAN, 'std': IMAGENET_DEFAULT_STD,
        'first_conv': 'stem.0', 'classifier': 'head',
        'license': 'apache-2.0',
        **kwargs
    }


default_cfgs = generate_default_cfgs({
    'visformer_tiny.in1k': _cfg(hf_hub_id='timm/'),

View on GitHub (pinned to 9a5261e31b)

Solutions

1. Create the model without features_only and call forward_intermediates() to get intermediate features
2. Choose a backbone that supports features_only if a feature pyramid API is required

Example fix

# before
model = timm.create_model('visformer_tiny', features_only=True)
# after
model = timm.create_model('visformer_tiny', features_only=False)
feats = model.forward_intermediates(x, indices=[1, 3, 5, 7])

kwargs.pop('features_only', False) if name.startswith('visformer') else None
model = timm.create_model(name, **kwargs)

def supports_features_only(name: str) -> bool:
    return not name.startswith('visformer')

try:
    model = timm.create_model(name, features_only=True)
except RuntimeError:
    model = timm.create_model(name)  # use forward_intermediates instead

When it happens

Trigger: timm.create_model('visformer_tiny', features_only=True) or any visformer_* with features_only in kwargs.

Common situations: Generic FPN/segmentation code that blindly passes features_only=True to every backbone in a model zoo.

Understand the failure class

Background: UnsupportedOperationException and "is not supported" errors: when a library deliberately refuses a call — this error's family across 30 libraries.

Related errors

• features_only not implemented for Vision Transformer models.
• features_only not implemented for Vision Transformer models.
• features_only not implemented for ConvMixer models.
• features_only not implemented for Vision Transformer models.
• Input image must have positive dimensions, got H={height}, W

AI-assisted analysis of huggingface/pytorch-image-models@9a5261e31b (2026-08-27). Data as JSON: /api/errors/c90274562e2c7b41. Report an issue: GitHub.