Conflict with yggdrasil decision forests #45

Due to Numpy version compatibility, ydf will not work.
here's the finding:

Your environment uses NumPy 2.0.2, YDF 0.13.0, and Pandas 2.2.2. The zillionare/ZigZag library may not be fully compatible with NumPy 2.0 due to changes in NumPy’s ABI (Application Binary Interface) or array handling, which can cause binary incompatibilities. This is evident from errors like numpy.ndarray size changed, may indicate binary incompatibility reported in jbn/ZigZag issues. If ZigZag was installed or imported in your Colab session, it could introduce a stale or incompatible NumPy state, potentially causing the YDF TypeError: PopulateColumnNumericalNPFloat32 error you encountered, where the dataset initializes with 0 records.
YDF and NumPy Interaction: YDF 0.13.0 expects NumPy arrays in specific formats (e.g., float32 for numerical features), and your previous YDF failures suggest a data-passing issue in the C++ backend, possibly exacerbated by NumPy’s internal state being altered by ZigZag’s Cython extensions. The zillionare/ZigZag build process involves Cython compilation, which can be sensitive to NumPy versions, and any mismatch could corrupt the session state, requiring a restart to clear it.