| from evaluate.visualization import radar_plot | |
| import matplotlib.pyplot as plt | |
| def compute_sample_scores(results, prompt): | |
| # 示例人工评分（实际可扩展为平均多个 prompt 的客观指标） | |
| return { | |
| "sd_v1_5": {"G": 4, "R": 4, "A": 4, "E": 3}, | |
| "openjourney_v4": {"G": 3, "R": 4, "A": 5, "E": 3}, | |
| "ldm_256": {"G": 2, "R": 3, "A": 3, "E": 5} | |
| } | |
| def plot_radar(scores_dict, out_path="radar.png"): | |
| radar_plot(scores_dict, model_names=list(scores_dict.keys()), title="GRACE Radar Comparison") | |
| plt.savefig(out_path) | |
| plt.close() | |