a ^g#@sZddlZddlTdddZdddZddd Zddedd ddfd d Zdedfd dZdS)N)*cCsr|durtt|}t}t}t||D]\}}||||q*t|tr`t||||St ||||SdS)z builds the C++ vectors for the GPU indices and the resources. Handles the case where the resources are assigned to the list of GPUs N) rangelenZGpuResourcesVectorZ Int32VectorzipZ push_back isinstanceZ IndexBinaryZ index_binary_cpu_to_gpu_multipleZindex_cpu_to_gpu_multiple) resourcesindexcogpusZvresZvdeviresr `/var/www/html/cobodadashboardai.evdpl.com/venv/lib/python3.9/site-packages/faiss/gpu_wrappers.pyindex_cpu_to_gpu_multiple_pys    rcCst||d|d}|S)N)r r ngpu)index_cpu_to_gpus_list)rr r index_gpur r rindex_cpu_to_all_gpus&srcCsT|dur|dkrtt}n|dur4|dkr4t|}dd|D}t||||}|S)z Here we can pass list of GPU ids as a parameter or ngpu to use first n GPU's. gpus mut be a list or None. co is a GpuMultipleClonerOptions NrcSsg|] }tqSr )ZStandardGpuResources).0_r r r 4z*index_cpu_to_gpus_list..)rZ get_num_gpusr)rr r rr rr r rr+s rFc CsZ|j\} } |jjrd} n&|jjr,|j}d} ntj|dd}d} t|}|jtj krXt }n|jtj krjt }nt d|j\}}|| ksJ|jjrd}n&|jjr|j}d}ntj|dd}d}t|}|jtj krt }n|jtj krt }nt d|durtj| |ftj d}n&|j| |fks$J|jtj ks6Jt|}|dur^tj| |ftjd}n|j| |fksrJt|}|jtjkrt}n0|j|jkrtjkrnnt}nt dt}||_||_| |_||_||_||_||_||_| |_||_| |_||_||_ ||_!||_"||_#| d ks8| d krHt$||| | n t%||||fS) a Compute the k nearest neighbors of a vector on one GPU without constructing an index Parameters ---------- res : StandardGpuResources GPU resources to use during computation xq : array_like Query vectors, shape (nq, d) where d is appropriate for the index. `dtype` must be float32. xb : array_like Database vectors, shape (nb, d) where d is appropriate for the index. `dtype` must be float32. k : int Number of nearest neighbors. D : array_like, optional Output array for distances of the nearest neighbors, shape (nq, k) I : array_like, optional Output array for the nearest neighbors, shape (nq, k) metric : MetricType, optional Distance measure to use (either METRIC_L2 or METRIC_INNER_PRODUCT) device: int, optional Which CUDA device in the system to run the search on. -1 indicates that the current thread-local device state (via cudaGetDevice) should be used (can also be set via torch.cuda.set_device in PyTorch) Otherwise, an integer 0 <= device < numDevices indicates the GPU on which the computation should be run vectorsMemoryLimit: int, optional queriesMemoryLimit: int, optional Memory limits for vectors and queries. If not 0, the GPU will use at most this amount of memory for vectors and queries respectively. Vectors are broken up into chunks of size vectorsMemoryLimit, and queries are broken up into chunks of size queriesMemoryLimit, including the memory required for the results. Returns ------- D : array_like Distances of the nearest neighbors, shape (nq, k) I : array_like Labels of the nearest neighbors, shape (nq, k) TFfloat32dtypezxq must be f32 or f16xb must be float32 or float16NzI must be i64 or i32r)&shapeflags c_contiguous f_contiguousTnpascontiguousarrayswig_ptrrrDistanceDataType_F32float16DistanceDataType_F16 TypeErroremptyZint64ZIndicesDataType_I64Zint32ZIndicesDataType_I32GpuDistanceParamsmetrickdimsvectorsvectorsRowMajor vectorType numVectorsqueriesqueriesRowMajor queryType numQueries outDistancesZ outIndicesZoutIndicesTypedeviceuse_cuvsZ bfKnn_tilingbfKnn)r xqxbr,DIr+r7r8ZvectorsMemoryLimitZqueriesMemoryLimitnqd xq_row_majorxq_ptrxq_typenbd2 xb_row_majorxb_ptrxb_typeD_ptrZI_ptrZI_typeargsr r rknn_gpu;s,         " rJcCs|j\}}|jjrd}n|jjr,|j}d}ntdt|} |jtj krNt } n$|jtj kr`t } ntj |dd}d}|j\} } | |ksJ|jjrd} n&|jjr|j}d} ntj |dd}d} t|}|jtj krt }n|jtj krt }ntd|durtj|| ftj d}n&|j|| fks$J|jtj ks6Jt|}t}||_d|_||_||_| |_||_| |_| |_||_| |_||_||_||_t|||S) ak Compute all pairwise distances between xq and xb on one GPU without constructing an index Parameters ---------- res : StandardGpuResources GPU resources to use during computation xq : array_like Query vectors, shape (nq, d) where d is appropriate for the index. `dtype` must be float32. xb : array_like Database vectors, shape (nb, d) where d is appropriate for the index. `dtype` must be float32. D : array_like, optional Output array for all pairwise distances, shape (nq, nb) metric : MetricType, optional Distance measure to use (either METRIC_L2 or METRIC_INNER_PRODUCT) device: int, optional Which CUDA device in the system to run the search on. -1 indicates that the current thread-local device state (via cudaGetDevice) should be used (can also be set via torch.cuda.set_device in PyTorch) Otherwise, an integer 0 <= device < numDevices indicates the GPU on which the computation should be run Returns ------- D : array_like All pairwise distances, shape (nq, nb) TFz5xq matrix should be row (C) or column-major (Fortran)rrrNr)rrrr r!r(r$rr"rr%r&r'r#r)r*r+r,r-r.r/r0r1r2r3r4r5r6r7r9)r r:r;r<r+r7r>r?r@rArBrCrDrErFrGrHrIr r rpairwise_distance_gpush         rK)NN)Nr)NNr) numpyr"Z faiss.loaderrrrZ METRIC_L2rJrKr r r r s