AttributeError: 'Tensor' object has no attribute 'astype'

[Kaveh8]

I have got an error complaining about type of img which is in the transforms.py:11, and it uses .astype. Apparently it considers img as a numpy array, while I have using a .map function for a tfdataset.

I think it should consider to convert input to an expected array (numpy for example).
Or, even better, it should only use tensorflow functions in order to make it useful for graph mode execution, such as when using .map function.

    <ipython-input-10-149e589e978b>:16 get_augmentation  *
        aug_data = aug(**data)
    /opt/conda/lib/python3.7/site-packages/volumentations/core/composition.py:24 __call__  *
        data = tr(force_apply, self.targets, **data)
    /opt/conda/lib/python3.7/site-packages/volumentations/core/transforms_interface.py:50 __call__  *
        data[k] = self.apply(v, **params)
    /opt/conda/lib/python3.7/site-packages/volumentations/augmentations/transforms.py:11 apply  *
        return img.astype(np.float32)

    AttributeError: 'Tensor' object has no attribute 'astype'