Nanne/pytorch-NetVlad

Error with Faiss!

Fleurrr opened this issue · 2 comments

Fleurrr commented

My python version is 3.8, and I installed faiss-gpu=1.6.1.There‘re something wrong when I run main.py
"kmeans = faiss.Kmeans(encoder_dim, opt.num_clusters, niter, verbose=False)
TypeError: init() takes 3 positional arguments but 4 were given"

Could you please tell me the solution!

Nanne commented

Seems like in a newer version they changed niter to be an named argument, rather than a positional one: https://github.com/facebookresearch/faiss/blob/c5975cda72981329af97df36b4986ae6f65550e2/faiss/python/__init__.py#L947-L950

Changing that line to kmeans = faiss.Kmeans(encoder_dim, opt.num_clusters, niter=niter, verbose=False) should fix this. I've committed this change for now without testing, as I assume it won't break anything for older versions. Can you test this and get back to me with if this fixes it?

  • 👍1
Fleurrr commented

Seems like in a newer version they changed niter to be an named argument, rather than a positional one: https://github.com/facebookresearch/faiss/blob/c5975cda72981329af97df36b4986ae6f65550e2/faiss/python/__init__.py#L947-L950

Changing that line to kmeans = faiss.Kmeans(encoder_dim, opt.num_clusters, niter=niter, verbose=False) should fix this. I've committed this change for now without testing, as I assume it won't break anything for older versions. Can you test this and get back to me with if this fixes it?

It works! Thank you very much for your help！