open-mmlab/mmdetection · error · RuntimeError

The `file_client_args` is deprecated, please use `backend_ar

Error message

The `file_client_args` is deprecated, please use `backend_args` instead, please refer tohttps://github.com/open-mmlab/mmdetection/blob/main/configs/_base_/datasets/coco_detection.py

What it means

CrowdHumanMetric.__init__ raises RuntimeError if the legacy `file_client_args` parameter is passed — the same 3.x deprecation enforcement as CocoMetric. File I/O must be configured via `backend_args`. The message points at the canonical config showing the new backend_args style.

Source

Thrown at mmdet/evaluation/metrics/crowdhuman_metric.py:100

        self.ann_file = ann_file
        # crowdhuman evaluation metrics
        self.metrics = metric if isinstance(metric, list) else [metric]
        allowed_metrics = ['MR', 'AP', 'JI']
        for metric in self.metrics:
            if metric not in allowed_metrics:
                raise KeyError(f"metric should be one of 'MR', 'AP', 'JI',"
                               f'but got {metric}.')

        self.format_only = format_only
        if self.format_only:
            assert outfile_prefix is not None, 'outfile_prefix must be not'
            'None when format_only is True, otherwise the result files will'
            'be saved to a temp directory which will be cleaned up at the end.'
        self.outfile_prefix = outfile_prefix
        self.backend_args = backend_args
        if file_client_args is not None:
            raise RuntimeError(
                'The `file_client_args` is deprecated, '
                'please use `backend_args` instead, please refer to'
                'https://github.com/open-mmlab/mmdetection/blob/main/configs/_base_/datasets/coco_detection.py'  # noqa: E501
            )

        assert eval_mode in [0, 1, 2], \
            "Unknown eval mode. mr_ref should be one of '0', '1', '2'."
        assert compare_matching_method is None or \
               compare_matching_method == 'VOC', \
               'The alternative compare_matching_method is VOC.' \
               'This parameter defaults to CALTECH(None)'
        assert mr_ref == 'CALTECH_-2' or mr_ref == 'CALTECH_-4', \
            "mr_ref should be one of 'CALTECH_-2', 'CALTECH_-4'."
        self.eval_mode = eval_mode
        self.iou_thres = iou_thres
        self.compare_matching_method = compare_matching_method
        self.mr_ref = mr_ref
        self.num_ji_process = num_ji_process

View on GitHub (pinned to cfd5d3a985)

Solutions

Replace file_client_args with backend_args=dict(backend='local') (or 'petrel' with path_mapping for remote)
Strip file_client_args keys from all configs in the repo (grep for it)
Use the config referenced in the message as the migration template

Example fix

# before
val_evaluator = dict(type='CrowdHumanMetric', metric='AP', file_client_args=dict(backend='disk'))
# after
val_evaluator = dict(type='CrowdHumanMetric', metric='AP', backend_args=dict(backend='local'))

Defensive patterns

Strategy: validation

Validate before calling

ev = cfg.get('val_evaluator', cfg.get('test_evaluator', {}))
assert 'file_client_args' not in ev, 'CrowdHumanMetric: use backend_args, not file_client_args'

Type guard

def config_is_mmdet3_compatible(eval_cfg: dict) -> bool:
    return 'file_client_args' not in eval_cfg

Prevention

Migrate all project configs in one sweep (grep file_client_args)
Wrap evaluator construction in a helper that injects backend_args
Track mmdetection release notes for API migrations

When it happens

Trigger: CrowdHumanMetric(..., file_client_args={'backend':'disk'}) or a CrowdHuman val/test config that still carries file_client_args after upgrading mmdetection to 3.x.

Common situations: Porting old CrowdHuman project configs (projects/CrowdHuman) to 3.x; keeping legacy kwargs in a custom subclass or a wrapper script.

Related errors

AI-assisted analysis of open-mmlab/mmdetection@cfd5d3a985 (2026-08-27). Data as JSON: /api/errors/43290ac5aae1f714. Report an issue: GitHub.