PCA data dimension problem

Thanks for your excellent work. While reading the code, I noticed that the first dimension of `components_to_data` in `pca.npz` is `512`, but the first dimension of `explained_variance_ratio` is `1024`. Why are these two sizes different? ```python import numpy as np import torch import torch.nn.functional as F pca = np.load("./unitalker_data_release_V1/D1_vocaset/pca.npz") print(pca.files) # explained_variance_ratio shape print("Explained Variance Ratio Shape:", pca['explained_variance_ratio'].shape) # components_to_data shape print("Components to Data Shape:", pca['components_to_data'].shape) # original_data_mean shape print("Original Data Mean Shape:", pca['original_data_mean'].shape) # S shape print("S Shape:", pca['S'].shape) # data_components_mean shape print("Data Components Mean Shape:", pca['data_components_mean'].shape) # data_components_std shape print("Data Components Std Shape:", pca['data_components_std'].shape) ``` ``` ['explained_variance_ratio', 'components_to_data', 'original_data_mean', 'S', 'data_components_mean', 'data_components_std'] Explained Variance Ratio Shape: (1024,) Components to Data Shape: (512, 15069) Original Data Mean Shape: (15069,) S Shape: (1024,) Data Components Mean Shape: (1024,) Data Components Std Shape: (1024,) ```