Why is np.float64 the only return dtype used for RandomForestRegressor predict?

[ethanglaser]

I am facing an issue when using hardware that is not float64/double compatible and have traced it to sklearn's RandomForestRegressor predict function, specifically: https://github.com/scikit-learn/scikit-learn/blob/d99b728b3/sklearn/ensemble/_forest.py#L989-L993. I am wondering why np.float64 is used here - wouldn't it make more sense to use X.dtype?

[glemaitre]

X.dtype is not really meaningful because this is the type of y. The trees are implemented in Cython so and the implementation of the tree expect to have a float64 target as input.