Debug / Fix an Extremely Randomized Trees (ExtraTrees) Implementation in NumPy

Task: Debug / Fix an Extremely Randomized Trees (ExtraTrees) Implementation in NumPy You are given a NumPy-only implementation of Extremely Randomized Trees (ExtraTrees) for either classification or regression. The code contains bugs that cause one or more of the following: training crashes (shape/indexing/numerical issues), or training runs but predictions do not match expected outputs, or results are not reproducible / randomness does not match the intended behavior. Without introducing any third-party libraries beyond NumPy, identify and fix the issues so that the implementation passes the provided unit tests / validation script. Required functionality (follow the provided code/tests) fit(X, y): train an ensemble of extremely randomized trees. predict(X): generate predictions; for classification, return class labels (or the probability/score format required by the tests); for regression, return continuous values. Randomness: when a seed is provided (e.g., random_state), results must be reproducible. Constraints Use Python + NumPy only. You may refactor internal logic as needed, but must keep the public API compatible with the tests. All provided tests must pass. I/O (test-driven) Training input: X: np.ndarray of shape (n_samples, n_features) y: np.ndarray of shape (n_samples,) or (n_samples, 1) Prediction output: Regression: shape (n_samples,) Classification: shape (n_samples,) (class ids) or an equivalent format required by the tests Example (format only; follow the real tests) Input: X is 100x10, y has length 100 Output: predict(X_test) returns an array of length len(X_test) Note: This is a take-home/debug style task. The core is reading existing code, locating bugs, and fixing behavior to match the ExtraTrees definition and the provided tests. Example Input (Provided by take-home: code + failing unit tests) Output All unit tests pass; deterministic outputs under fixed seed