about compute_feats.py and train_tcga.py Number of output classes question

[Lliu666liuliu]

您好，请问使用Camelyon16数据集，使用simclr模型权重文件提取实例特征compute_feats.py该文件 --num_classes是1还是2，提取完实例特征进行mil聚合时，train_tcga.py --num_classes是1还是2，一时没有搞明白，希望您能解答一下

[binli123]

It may not matter because the aggregator (including the last classification layer) is discarded during features extracting.