mdouze / sample_the_earth Goto Github PK

Good morning,
Mr.mdouze , i have saw you replied many questions about faiss at GitHub . I want to ask you some questions sincerely.
If i want to transfer the index from cpu to gpu ,and transfer the index from gpu to cpu , what interface or method should i use , i ran the following code but reported the following error:
#code
index_name = "BIVF32,SQfp16"
index_temp = faiss.index_binary_factory(dim, index_name.split(",")[-2] + ",Flat")
res = faiss.StandardGpuResources()
index_gpu = faiss.index_cpu_to_gpu(res,0,index_temp)
#code end
#error
Traceback (most recent call last):
File "my_train_ivf.py", line 130, in
train_idx_and_validate()
File "my_train_ivf.py", line 57, in train_idx_and_validate
index_gpu = faiss.index_cpu_to_gpu(res,0,index_temp)
File "/opt/conda/lib/python3.7/site-packages/faiss/init.py", line 448, in replacement_function
result = original_function(*args)
NotImplementedError: Wrong number or type of arguments for overloaded function 'index_cpu_to_gpu'.
Possible C/C++ prototypes are:
faiss::gpu::index_cpu_to_gpu(faiss::gpu::GpuResources *,int,faiss::Index const *,faiss::gpu::GpuClonerOptions const *)
faiss::gpu::index_cpu_to_gpu(faiss::gpu::GpuResources *,int,faiss::Index const *)
#error end
In fact, i use method faiss.index_cpu_to_gpu(index_temp) like this early , but it occurs to reportd error also, i viewed the faiss code , and use faiss.index_cpu_to_gpu(res,0,index_temp).
Can you help me to find out the matter. I guess my server computer's faiss version is lower than the version i used in code ,but i can't sure it ,and i don't know how to see the faiss version of server computer.
If you have time to view my question and have any idea about it ,you can reply me email me any time.
Looking forward to you reply.

Best Regards！