Lightning-AI/pytorch-lightning · error · TypeError

Expected a list as "images", found {type(images)}

Error message

Expected a list as "images", found {type(images)}

What it means

WandbLogger.log_image requires the images argument to be a Python list (each element is then wrapped into wandb.Image with per-item kwargs). Passing a tensor, numpy array, tuple, or generator triggers this TypeError before anything is logged.

Source

Thrown at src/lightning/pytorch/loggers/wandb.py:492

        step: Optional[int] = None,
    ) -> None:
        """Log text as a Table.

        Can be defined either with `columns` and `data` or with `dataframe`.

        """

        self.log_table(key, columns, data, dataframe, step)

    @rank_zero_only
    def log_image(self, key: str, images: list[Any], step: Optional[int] = None, **kwargs: Any) -> None:
        """Log images (tensors, numpy arrays, PIL Images or file paths).

        Optional kwargs are lists passed to each image (ex: caption, masks, boxes).

        """
        if not isinstance(images, list):
            raise TypeError(f'Expected a list as "images", found {type(images)}')
        n = len(images)
        for k, v in kwargs.items():
            if len(v) != n:
                raise ValueError(f"Expected {n} items but only found {len(v)} for {k}")
        kwarg_list = [{k: kwargs[k][i] for k in kwargs} for i in range(n)]

        import wandb

        metrics = {key: [wandb.Image(img, **kwarg) for img, kwarg in zip(images, kwarg_list)]}
        self.log_metrics(metrics, step)  # type: ignore[arg-type]

    @rank_zero_only
    def log_audio(self, key: str, audios: list[Any], step: Optional[int] = None, **kwargs: Any) -> None:
        r"""Log audios (numpy arrays, or file paths).

        Args:
            key: The key to be used for logging the audio files
            audios: The list of audio file paths, or numpy arrays to be logged

View on GitHub (pinned to 9fed5c27d2)

Solutions

1. Wrap the batch into a list: list(tensor) or images.tolist()/images.cpu() split per sample
2. Convert numpy arrays: [np_img for np_img in arr]
3. Unpack tuples/generators: list(images)

Example fix

# before
logger.log_image(images=batch_tensor, caption=["a", "b"])
# after
logger.log_image(images=list(batch_tensor), caption=["a", "b"])

assert isinstance(images, list), "images must be a list"

def is_image_list(x) -> bool:
    return isinstance(x, list) and len(x) > 0 and all(isinstance(i, (str,)) or hasattr(i, "__array__") or torch.is_tensor(i) for i in x)

When it happens

Trigger: logger.log_image(tensor_batch, ...) or logger.log_image(np.array([...]), ...) instead of logger.log_image([t1, t2], ...); also passing a tuple.

Common situations: Passing a raw (C,H,W)/(B,C,H,W) torch tensor directly from a training step instead of converting to a list of per-sample tensors/images.

Related errors

• Expected {n} items but only found {len(v)} for {k}
• Expected a list as "audios", found {type(audios)}
• Expected a list as "videos", found {type(videos)}
• `name` must be a str, found {name}
• Expected `torch.nn.Module` or `torch.optim.Optimizer`, got:

AI-assisted analysis of Lightning-AI/pytorch-lightning@9fed5c27d2 (2026-08-28). Data as JSON: /api/errors/45d1523372e7f799. Report an issue: GitHub.