| """\ | |
| comparing gt_mel and predcted_mel | |
| """ | |
| import torch | |
| import numpy as np | |
| import matplotlib.pyplot as plt | |
| # load a GT mel | |
| gt_mel = np.load("../config/dataset/LJSpeech/Dataset/LJSpeech-1.1/mels/LJ001-0001.npy") | |
| print(f"GT mel shape: {gt_mel.shape}") | |
| print(f"GT mel min: {gt_mel.min():.4f}") | |
| print(f"GT mel max: {gt_mel.max():.4f}") | |
| print(f"GT mel mean: {gt_mel.mean():.4f}") | |
| print(f"GT mel std: {gt_mel.std():.4f}") | |
| # load predicted mel from inference | |
| # save predicted mel before passing to vocoder in inference.py | |
| # add this line before vocoder call: | |
| # np.save("predicted_mel.npy", final_mel.cpu().numpy()) | |
| predicted_mel = np.load("predicted_mel.npy") | |
| print(f"\nPredicted mel shape: {predicted_mel.shape}") | |
| print(f"Predicted mel min: {predicted_mel.min():.4f}") | |
| print(f"Predicted mel max: {predicted_mel.max():.4f}") | |
| print(f"Predicted mel mean: {predicted_mel.mean():.4f}") | |
| print(f"Predicted mel std: {predicted_mel.std():.4f}") | |