I am a student, studying unsupervised learning. And I need to solve a problem: to cluster 10k 128-dimention data points. I tried k-means++, spectral clustering, HDBSCAN, MeanShift and so on, but they spent too much time and seems not very accurate. To study, I can’t use scikit-learn, I can only import numpy. Could anyone give me some advice, or some sample code?

I tried k-means++, but the shape of data does not fit k-means.
Spectral clustering: takes too much time(I used knn to create weight matrix).
Other methods mentioned takes too much time.