WZMIAOMIAO/deep-learning-for-image-processing · error · AttributeError

'{}' object has no attribute '{}'

Error message

'{}' object has no attribute '{}'

What it means

SmoothedValue/MetricLogger's __getattr__ looks up requested attributes in its meters dict and then its instance __dict__; if the attribute is in neither, it raises AttributeError. This typically surfaces as a typo of a loss/metric name logged via meter.meters keys.

Source

Thrown at pytorch_object_detection/mask_rcnn/train_utils/distributed_utils.py:137

class MetricLogger(object):
    def __init__(self, delimiter="\t"):
        self.meters = defaultdict(SmoothedValue)
        self.delimiter = delimiter

    def update(self, **kwargs):
        for k, v in kwargs.items():
            if isinstance(v, torch.Tensor):
                v = v.item()
            assert isinstance(v, (float, int))
            self.meters[k].update(v)

    def __getattr__(self, attr):
        if attr in self.meters:
            return self.meters[attr]
        if attr in self.__dict__:
            return self.__dict__[attr]
        raise AttributeError("'{}' object has no attribute '{}'".format(
            type(self).__name__, attr))

    def __str__(self):
        loss_str = []
        for name, meter in self.meters.items():
            loss_str.append(
                "{}: {}".format(name, str(meter))
            )
        return self.delimiter.join(loss_str)

    def synchronize_between_processes(self):
        for meter in self.meters.values():
            meter.synchronize_between_processes()

    def add_meter(self, name, meter):
        self.meters[name] = meter

    def log_every(self, iterable, print_freq, header=None):

View on GitHub (pinned to 1ec3fe6f37)

Solutions

  1. Check metric_logger.meters keys (print them) and use an existing name
  2. Ensure the model's loss_dict keys are registered via metric_logger.update(loss_dict) before access
  3. Use hasattr(metric_logger, name) or metric_logger.meters.get(name) for optional metrics

Example fix

// before
avg_mask_loss = metric_logger.loss_mask.global_avg  # may not exist
// after
if "loss_mask" in metric_logger.meters:
    avg_mask_loss = metric_logger.loss_mask.global_avg
Defensive patterns

Strategy: type-guard

Validate before calling

name = 'loss_mask'
if name in metric_logger.meters:
    value = metric_logger.meters[name].global_avg
else:
    logger.warning(f'meter {name} not registered')

Type guard

def has_meter(logger_obj, name):
    return name in getattr(logger_obj, 'meters', {})

Try / catch

try:
    value = metric_logger.loss_mask.global_avg
except AttributeError as e:
    logger.warning(f'missing meter: {e}; available: {list(metric_logger.meters)}')
    value = None

Prevention

When it happens

Trigger: Accessing e.g. metric_logger.loss_mask when no 'loss_mask' meter was registered (update_losses never saw that key), or any attribute not added via meters.update/track.

Common situations: Accessing a loss key not produced by the model (name mismatch between model output dict and logger); reading stats before the first iteration; typos in attribute names.

Related errors

AI-assisted analysis of WZMIAOMIAO/deep-learning-for-image-processing@1ec3fe6f37 (2026-08-30). Data as JSON: /api/errors/a66c98728679056d. Report an issue: GitHub.