Lightning-AI/pytorch-lightning · error · AttributeError

'{type(self).__name__}' object has no attribute '{key}'

Error message

'{type(self).__name__}' object has no attribute '{key}'

What it means

This class is a dict subclass whose `__getattr__` maps attribute access to key lookup (`self[key]`). When code accesses an attribute that is neither a real attribute nor a key in the dict, the KeyError is converted into a standard AttributeError with the class and key name, matching Python's normal attribute-error semantics.

Source

Thrown at src/lightning/fabric/utilities/data.py:494

        >>> import torch
        >>> model = torch.nn.Linear(2, 2)
        >>> state = AttributeDict(model=model, iter_num=0)
        >>> state.model
        Linear(in_features=2, out_features=2, bias=True)
        >>> state.iter_num += 1
        >>> state.iter_num
        1
        >>> state
        "iter_num": 1
        "model":    Linear(in_features=2, out_features=2, bias=True)

    """

    def __getattr__(self, key: str) -> Any:
        try:
            return self[key]
        except KeyError as e:
            raise AttributeError(f"'{type(self).__name__}' object has no attribute '{key}'") from e

    def __setattr__(self, key: str, val: Any) -> None:
        self[key] = val

    def __delattr__(self, item: str) -> None:
        if item not in self:
            raise KeyError(item)
        del self[item]

    def __repr__(self) -> str:
        if not len(self):
            return ""
        max_key_length = max(len(str(k)) for k in self)
        tmp_name = "{:" + str(max_key_length + 3) + "s} {}"
        rows = [tmp_name.format(f'"{n}":', self[n]) for n in sorted(self.keys())]
        return "\n".join(rows)

View on GitHub (pinned to 9fed5c27d2)

Solutions

1. Check membership before access: `if 'key' in obj:` or `obj.get('key', default)`.
2. Use explicit key access `obj['key']` with KeyError handling, or verify available keys via `list(obj.keys())`.
3. Regenerate/re-save the dict so expected keys exist; align key names across versions.

Example fix

# before
value = cfg.learning_rate  # AttributeError if key absent

# after
value = cfg.get('learning_rate', default_lr)

if 'learning_rate' in cfg:
    lr = cfg['learning_rate']
else:
    lr = default_lr

def has_key(cfg, key: str) -> bool:
    return isinstance(cfg, dict) and key in cfg

try:
    lr = cfg.learning_rate
except AttributeError:
    lr = default_lr

When it happens

Trigger: Accessing a missing attribute on Lightning's dict-like wrapper (e.g. `_AttributeDict`): `cfg.some_missing_key`, where 'some_missing_key' was never inserted into the dict.

Common situations: Reading configuration/hyperparameter objects (e.g. `hparams` or saved checkpoints restored into an attribute-dict) where a key was renamed between versions or never saved; typos in config key access.

Related errors

• You cannot set both `activation_checkpointing` and `activati
• Device should be CPU, got {device} instead.
• `devices` selected with `CPUAccelerator` should be an int >
• Device should be CUDA, got {device} instead.
• You requested to find {num_devices} devices but there are no

AI-assisted analysis of Lightning-AI/pytorch-lightning@9fed5c27d2 (2026-08-28). Data as JSON: /api/errors/82f31ed409209d43. Report an issue: GitHub.