The issue with AlgorithmChecker and SystemChecker is that methods/attributes are not forwarded to the wrapped class which reduces the flexibility of DynaDojo.
For example, you cannot access an underlying Tensorflow or Pytorch neural net model without first calling ._alg which means you can't move the model to GPU natively.
The issue with AlgorithmChecker and SystemChecker is that methods/attributes are not forwarded to the wrapped class which reduces the flexibility of DynaDojo.
For example, you cannot access an underlying Tensorflow or Pytorch neural net model without first calling
._alg
which means you can't move the model to GPU natively.I think we should migrate to using decorators.
Relevant stackoverflow post: https://stackoverflow.com/a/63644860