Dear VecFlow authors,
I tried to build the WikiANN datasets from the HF, but the read out data are all beyond 1e8.
Note I tried some fix, since the naive script are not able to be runned.
final_A[500_000*(i-1):500_000*i] = np.fromfile(fname, dtype=np.float32).reshape(-1, 768)
Could you kindly tell me how could I build this dataset, very apperciate!
Best
Dear VecFlow authors,
I tried to build the WikiANN datasets from the HF, but the read out data are all beyond 1e8.
Note I tried some fix, since the naive script are not able to be runned.
Could you kindly tell me how could I build this dataset, very apperciate!
Best