AF3 OpenFoldDataset类get_stochastic_train_filter_prob方法解读

AlphaFold3 data_modules 模块的 OpenFoldDataset 类的 get_stochastic_train_filter_prob 方法是类的静态方法（@staticmethod），该方法接收一个 cache_entry（字典或类似的结构），用于存储蛋白质的相关信息。返回一个 浮点数，表示当前样本的采样概率。

源代码：

@staticmethoddef get_stochastic_train_filter_prob(cache_entry: Any,*args, **kwargs) -> float:# Stochastic filtersprobabilities = []cluster_size = cache_entry.get("cluster_size", None)if cluster_size is not None and cluster_size > 0:probabilities.append(1 / cluster_size)chain_length = len(cache_entry["seq"])probabilities.append((1 / 512) * (max(min(chain_length, 512), 256)))# Risk of underflow here?out = 1for p in probabilities: