AttributeError: module 'faiss' has no attribute 'cast_integer_to_long_ptr'

When i run the train_net.py in --eval-only mode with rerank enable, i got the error: [05/28 16:15:42 fastreid.evaluation.evaluator]: Start inference on 19889 images [05/28 16:15:50 fastreid.evaluation.evaluator]: Inference done 11/156. 0.5326 s / batch. ETA=0:01:21 [05/28 16:16:20 fastreid.evaluation.evaluator]: Inference done 64/156. 0.5372 s / batch. ETA=0:00:52 [05/28 16:16:50 fastreid.evaluation.evaluator]: Inference done 116/156. 0.5423 s / batch. ETA=0:00:22 [05/28 16:17:13 fastreid.evaluation.evaluator]: Total inference time: 0:01:26.502893 (0.572867 s / batch per device, on 1 devices) [05/28 16:17:13 fastreid.evaluation.evaluator]: Total inference pure compute time: 0:01:22 (0.544687 s / batch per device, on 1 devices) [05/28 16:17:17 fastreid.evaluation.reid_evaluation]: Test with rerank setting Traceback (most recent call last): File "tools/train_net.py", line 57, in <module> args=(args,), File "./fastreid/engine/launch.py", line 71, in launch main_func(*args) File "tools/train_net.py", line 39, in main res = DefaultTrainer.test(cfg, model) File "./fastreid/engine/defaults.py", line 454, in test results_i = inference_on_dataset(model, data_loader, evaluator, flip_test=cfg.TEST.FLIP.ENABLED) File "./fastreid/evaluation/evaluator.py", line 156, in inference_on_dataset results = evaluator.evaluate() File "./fastreid/evaluation/reid_evaluation.py", line 102, in evaluate rerank_dist = build_dist(query_features, gallery_features, metric="jaccard", k1=k1, k2=k2) File "/home/ccf/anaconda3/envs/fastreid/lib/python3.7/site-packages/torch/autograd/grad_mode.py", line 15, in decorate_context return func(*args, **kwargs) File "./fastreid/utils/compute_dist.py", line 53, in build_dist dist = compute_jaccard_distance(feat, k1=kwargs["k1"], k2=kwargs["k2"], search_option=0) File "/home/ccf/anaconda3/envs/fastreid/lib/python3.7/site-packages/torch/autograd/grad_mode.py", line 15, in decorate_context return func(*args, **kwargs) File "./fastreid/utils/compute_dist.py", line 78, in compute_jaccard_distance _, initial_rank = search_raw_array_pytorch(res, features, features, k1) File "./fastreid/utils/faiss_utils.py", line 87, in search_raw_array_pytorch I_ptr = swig_ptr_from_LongTensor(I) File "./fastreid/utils/faiss_utils.py", line 19, in swig_ptr_from_LongTensor return faiss.cast_integer_to_long_ptr( AttributeError: module 'faiss' has no attribute 'cast_integer_to_long_ptr'

[AlphaPlusTT]

yes, faiss<=1.6.4 is fine

[gallardorafael]

Just to confirm: faiss<=1.6.4 solves the problem.

The problem is that the ‘cast_integer_to_long_ptr’ disappeared after 1.6.4.