where is the get_similar_vectors function code?

[mengchengyao]

Hi
Thank you for your great work in Gene embedding with GenePT.

When I run aorta_data_analysis.ipynb, it did not work with error of could not find get_similar_vectors function code ? I tried to sovle this problem with scGPT code (get_similar_vectors), it worked, but the accuracy_score was low (sklearn.metrics.accuracy_score(gt_list, pred_list): 0.83), do is right?

when I use follow code, the accuracy_score is 0.87, almost the same to your result.

class Similarity(nn.Module):
    '''
    Dot product or cosine similarity
    '''
    def __init__(self, temp):
        super().__init__()
        self.temp = temp
        self.cos = nn.CosineSimilarity(dim=-1)
    def forward(self, x, y):
        # 将 NumPy 数组转换为 PyTorch 张量
        x_tensor = torch.from_numpy(x)
        y_tensor = torch.from_numpy(y)
        # 计算余弦相似度
        cos_sim = self.cos(x_tensor, y_tensor) / self.temp
        # 如果需要返回 NumPy 数组，可以在这里进行转换
        return cos_sim.numpy()
def get_similar_vectors(vector, ref, top_k=10):
    sim = Similarity(temp=0.5)
    # sims = cos_sim(vector, ref)
    sims=sim(vector, ref)
    # cos_sim=torch.mm(vector, ref)
    # sims = l2_sim(vector, ref) ###scGPT用的是此函数，但是结果较差
    top_k_idx = np.argsort(sims)[::-1][:top_k]
    return top_k_idx, sims[top_k_idx]

Yourse