{"id":66593,"date":"2026-11-14T19:27:03","date_gmt":"2026-11-14T11:27:03","guid":{"rendered":"https:\/\/www.biocloudservice.com\/wordpress\/?p=66593"},"modified":"2026-11-14T19:27:03","modified_gmt":"2026-11-14T11:27:03","slug":"rapids-singlecell-gpu-neighbors-guide","status":"publish","type":"post","link":"https:\/\/www.biocloudservice.com\/wordpress\/rapids-singlecell-gpu-neighbors-guide\/","title":{"rendered":"\u5355\u7ec6\u80de\u5206\u6790\u5361\u5728\u5efa\u56fe\uff0c\u503c\u5f97\u6362 GPU \u5417\uff1f\u8bfb\u61c2 rapids-singlecell"},"content":{"rendered":"
\n

\u5047\u8bbe\u624b\u91cc\u5df2\u6709\u4e00\u4efd\u5355\u7ec6\u80de\u8868\u8fbe\u77e9\u9635\u3002\u7b2c\u4e00\u6b21\u8dd1\u5b8c\u805a\u7c7b\u540e\uff0c\u4f60\u60f3\u628a Leiden \u7684\u5206\u8fa8\u7387\u8c03\u4f4e\u4e00\u70b9\uff0c\u770b\u770b\u4e00\u4e2a\u7ec6\u80de\u7fa4\u8fd8\u4f1a\u4e0d\u4f1a\u62c6\u6210\u4e24\u7c07\u3002\u5982\u679c\u90bb\u63a5\u56fe\u8fd8\u4fdd\u5b58\u5728\u5bf9\u8c61\u91cc\uff0c\u53ea\u8981\u5728\u539f\u56fe\u4e0a\u91cd\u8dd1\u805a\u7c7b\u5c31\u80fd\u56de\u7b54\u8fd9\u4e2a\u95ee\u9898\u3002\u8981\u662f\u4f60\u6539\u7684\u662f\u53c2\u4e0e\u5efa\u56fe\u7684\u4e3b\u6210\u5206\u6570\uff0c\u7ec6\u80de\u4e4b\u95f4\u7684\u8ddd\u79bb\u4f1a\u53d8\uff0c\u624d\u9700\u8981\u91cd\u5efa\u90bb\u63a5\u56fe\uff0c\u518d\u91cd\u8dd1\u4f9d\u8d56\u5b83\u7684\u805a\u7c7b\u6216\u53ef\u89c6\u5316\u3002<\/p>\n

\u8fd9\u4e00\u533a\u522b\u51b3\u5b9a\u4e86 GPU \u503c\u4e0d\u503c\u5f97\u7528\u3002\u5148\u627e\u51fa\u8fd9\u6b21\u5206\u6790\u771f\u6b63\u9700\u8981\u91cd\u7b97\u7684\u6b65\u9aa4\uff0c\u518d\u770b\u5b83\u5360\u4e86\u591a\u5c11\u7b49\u5f85\u65f6\u95f4\u3002<\/strong> rapids-singlecell \u662f scverse \u751f\u6001\u4e2d\u7684 Python \u4ed3\u5e93\uff0c\u4e3a AnnData \u5bf9\u8c61\u63d0\u4f9b\u4e00\u6279\u53ef\u7528 NVIDIA GPU \u6267\u884c\u7684\u5355\u7ec6\u80de\u7b97\u6cd5\u3002\u4e0b\u9762\u4ee5 v0.18.1 \u7684\u5e38\u7528\u63a5\u53e3\u4e3a\u4f8b\uff0c\u4ece\u5df2\u7ecf\u5f97\u5230\u7684\u201c\u7ec6\u80de \u00d7 \u57fa\u56e0\u201d\u77e9\u9635\u8bb2\u8d77\u3002\u6bd4\u5bf9 FASTQ\u3001\u8bc6\u522b\u6761\u5f62\u7801\u548c\u751f\u6210\u8ba1\u6570\u77e9\u9635\u5c5e\u4e8e\u66f4\u4e0a\u6e38\u7684\u4efb\u52a1\u3002rapids-singlecell v0.18.1<\/a><\/p>\n

\"\u5355\u7ec6\u80de\u8ba1\u7b97\u4f9d\u8d56\u793a\u610f\uff1a\u8868\u8fbe\u77e9\u9635\u7ecf\u9884\u5904\u7406\u548c<\/p>\n

\u4ece\u8868\u8fbe\u77e9\u9635\u5230 UMAP\uff0c\u54ea\u4e9b\u6b65\u9aa4\u80fd\u5728 GPU \u4e0a\u505a\uff1f<\/h2>\n

\u5728 AnnData \u4e2d\uff0cadata.X<\/code> \u901a\u5e38\u5b58\u653e\u5f53\u524d\u4f7f\u7528\u7684\u8868\u8fbe\u77e9\u9635\u3002\u5bf9\u6bcf\u4e2a\u7ec6\u80de\u6c42\u548c\u3001\u5bf9\u6bcf\u4e2a\u57fa\u56e0\u7edf\u8ba1\u53d8\u5f02\uff0c\u5305\u542b\u5927\u91cf\u53ef\u5e76\u884c\u7684\u6570\u503c\u8fd0\u7b97\u3002\u7b49\u5230\u5bfb\u627e\u90bb\u5c45\u65f6\uff0c\u8ba1\u7b97\u5bf9\u8c61\u5df2\u7ecf\u53d8\u4e86\uff1a\u7528\u4e00\u7ec4\u4e3b\u6210\u5206\u4ee3\u8868\u6bcf\u4e2a\u7ec6\u80de\uff0c\u5728\u8fd9\u4e2a\u4f4e\u7ef4\u7a7a\u95f4\u91cc\u5bfb\u627e\u76f8\u4f3c\u7684\u7ec6\u80de\uff0c\u518d\u628a\u5b83\u4eec\u8fde\u6210\u56fe\u3002\u805a\u7c7b\u548c UMAP \u90fd\u80fd\u4f7f\u7528\u8fd9\u5f20\u56fe\uff0c\u5404\u81ea\u56de\u7b54\u4e0d\u540c\u7684\u95ee\u9898\u3002\u5b98\u65b9 GPU \u51fd\u6570\u76ee\u5f55<\/a><\/p>\n

\n\n\n\n\n\n\n\n\n\n
\u5206\u6790\u73af\u8282<\/th>\nGPU \u63a5\u53e3\u793a\u4f8b<\/th>\n\u53ef\u4ee5\u5f97\u5230\u4ec0\u4e48\uff0c\u4f55\u65f6\u9700\u8981\u91cd\u7b97<\/th>\n<\/tr>\n<\/thead>\n
\u8d28\u63a7\u3001\u5f52\u4e00\u5316<\/td>\nrsc.pp.calculate_qc_metrics<\/code>\u3001normalize_total<\/code>\u3001log1p<\/code><\/td>\n\u7edf\u8ba1\u6bcf\u4e2a\u7ec6\u80de\u7684\u8ba1\u6570\u4e0e\u68c0\u51fa\u57fa\u56e0\uff0c\u6216\u5bf9\u8868\u8fbe\u91cf\u505a\u5f52\u4e00\u5316\u548c\u5bf9\u6570\u53d8\u6362\u3002\u8d28\u63a7\u9608\u503c\u4ecd\u8981\u7ed3\u5408\u6837\u672c\u3001\u7ec4\u7ec7\u548c\u5efa\u5e93\u65b9\u5f0f\u5224\u65ad\u3002<\/td>\n<\/tr>\n
\u9ad8\u53d8\u57fa\u56e0\u3001PCA<\/td>\nrsc.pp.highly_variable_genes<\/code>\u3001rsc.pp.pca<\/code><\/td>\n\u524d\u8005\u7b5b\u9009\u7279\u5f81\uff0c\u540e\u8005\u628a\u8868\u8fbe\u77e9\u9635\u538b\u7f29\u4e3a\u6bcf\u4e2a\u7ec6\u80de\u7684\u4e00\u7ec4\u5750\u6807\u3002\u9ad8\u53d8\u57fa\u56e0\u65b9\u6cd5\u5bf9\u8f93\u5165\u6709\u8981\u6c42\uff0c\u4f8b\u5982 seurat_v3<\/code> \u4f7f\u7528\u539f\u59cb\u8ba1\u6570\uff1b\u8981\u5148\u5f04\u6e05 X<\/code> \u548c counts<\/code> \u5c42\u5404\u5b58\u4e86\u4ec0\u4e48\u3002<\/td>\n<\/tr>\n
\u8fd1\u90bb\u641c\u7d22\u3001\u6784\u56fe<\/td>\nrsc.pp.neighbors<\/code><\/td>\n\u7528 PCA \u5750\u6807\u7b49\u8868\u793a\u5bfb\u627e\u90bb\u5c45\uff0c\u4fdd\u5b58\u8ddd\u79bb\u4e0e\u8fde\u63a5\u5173\u7cfb\u3002\u6539\u7528\u53e6\u4e00\u4efd\u8868\u793a\u3001\u8c03\u6574\u4e3b\u6210\u5206\u6570\u6216\u90bb\u5c45\u6570\uff0c\u90fd\u5e94\u91cd\u65b0\u5efa\u56fe\u3002<\/td>\n<\/tr>\n
\u56fe\u805a\u7c7b<\/td>\nrsc.tl.leiden<\/code><\/td>\n\u5728\u90bb\u63a5\u56fe\u4e0a\u5212\u5206\u7ec6\u80de\u7fa4\uff0c\u8f93\u51fa\u6bcf\u4e2a\u7ec6\u80de\u7684\u7c07\u6807\u7b7e\u3002\u53ea\u6bd4\u8f83\u4e0d\u540c\u5206\u8fa8\u7387\uff0c\u53ef\u56fa\u5b9a\u540c\u4e00\u5f20\u56fe\uff0c\u4fdd\u5b58\u591a\u7ec4\u6807\u7b7e\u3002<\/td>\n<\/tr>\n
UMAP \u5d4c\u5165<\/td>\nrsc.tl.umap<\/code><\/td>\n\u6839\u636e\u90bb\u63a5\u5173\u7cfb\u6c42\u4e8c\u7ef4\u5750\u6807\u3002\u5df2\u6709\u5750\u6807\u65f6\uff0c\u6309\u53e6\u4e00\u4e2a\u57fa\u56e0\u3001\u6837\u672c\u6216\u7c07\u6807\u7b7e\u91cd\u65b0\u7740\u8272\uff0c\u53ea\u9700\u91cd\u65b0\u7ed8\u56fe\u3002<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n

\u8868\u4e2d\u51fd\u6570\u90fd\u6709 GPU \u5b9e\u73b0\u3002\u4e00\u6b21\u8c03\u7528\u8fd8\u53ef\u80fd\u6d89\u53ca CPU \u8c03\u5ea6\u3001\u6570\u7ec4\u8f6c\u6362\u4e0e\u7ed3\u679c\u56de\u5b58\u3002\u8bfb\u53d6 h5ad\u3001\u5199\u6587\u4ef6\u3001\u7528 Scanpy \u753b\u56fe\u4e5f\u5e94\u5355\u72ec\u8ba1\u65f6\uff0c\u5c24\u5176\u662f\u6570\u636e\u653e\u5728\u5171\u4eab\u7f51\u7edc\u76d8\u4e0a\u65f6\u3002\u5b98\u65b9\u7528\u6cd5\u4e0e\u6570\u636e\u8f6c\u6362\u793a\u4f8b<\/a><\/p>\n

\u770b\u4e00\u6bb5\u4ee3\u7801\uff1a\u540c\u4e00\u5f20\u56fe\u4e0a\u8bd5\u4e24\u79cd\u805a\u7c7b\u5206\u8fa8\u7387<\/h2>\n

\u4e0b\u9762\u8fd9\u4e2a\u793a\u4f8b\u4ece\u8f83\u9760\u540e\u7684\u73af\u8282\u5207\u5165\u3002\u5047\u8bbe cells_pca.h5ad<\/code> \u5df2\u5b8c\u6210\u8d28\u63a7\u3001\u7279\u5f81\u9009\u62e9\u548c PCA<\/strong>\uff0cobsm[\"X_pca\"]<\/code> \u4fdd\u5b58\u4e86\u53ef\u7528\u7684\u4e3b\u6210\u5206\u5750\u6807\uff0c\u7ec6\u80de\u6570\u5927\u4e8e\u793a\u4f8b\u4e2d\u7684\u90bb\u5c45\u6570\u3002\u5728\u5df2\u914d\u7f6e\u517c\u5bb9 NVIDIA \u9a71\u52a8\u3001CUDA \u4e0e rapids-singlecell \u7684\u73af\u5883\u4e2d\uff0c\u53ef\u4ee5\u76f4\u63a5\u6307\u5b9a\u8fd9\u4efd\u8868\u793a\u5efa\u56fe\u3002\u53c2\u6570\u7528\u4e8e\u6f14\u793a\uff0c\u53ef\u6309\u7814\u7a76\u8bbe\u8ba1\u8c03\u6574\u3002\u8fd1\u90bb\u63a5\u53e3\u6e90\u7801<\/a> \u00b7 \u5b89\u88c5\u8bf4\u660e<\/a><\/p>\n

import scanpy as sc\r\nimport rapids_singlecell as rsc\r\n\r\nadata = sc.read_h5ad("cells_pca.h5ad")\r\nrsc.pp.neighbors(adata, use_rep="X_pca", n_neighbors=15)\r\n\r\nrsc.tl.leiden(adata, resolution=0.5, key_added="leiden_05")\r\nrsc.tl.leiden(adata, resolution=1.0, key_added="leiden_10")\r\n\r\nrsc.tl.umap(adata)\r\nsc.pl.umap(adata, color=["leiden_05", "leiden_10"])\r\n<\/code><\/pre>\n
\u4e24\u5e45\u56fe\u5171\u7528\u4e00\u5957 UMAP \u5750\u6807\u3002\u989c\u8272\u8fb9\u754c\u79fb\u52a8\u65f6\uff0c\u4f60\u770b\u5230\u7684\u662f\u805a\u7c7b\u7c92\u5ea6\u7684\u5f71\u54cd\u3002\u5982\u679c\u5bf9\u8c61\u91cc\u672c\u6765\u5c31\u6709\u6309\u76ee\u6807\u53c2\u6570\u6784\u5efa\u7684\u90bb\u63a5\u56fe\uff0c\u8fd8\u53ef\u4ee5\u8df3\u8fc7 neighbors<\/code>\u3002\u82e5\u7814\u7a76\u7684\u662f\u5904\u7406\u7ec4\u548c\u5bf9\u7167\u7ec4\u7684\u5dee\u5f02\uff0c\u63a5\u4e0b\u6765\u4ecd\u8981\u6838\u5bf9\u5206\u7fa4\u662f\u5426\u7531\u67d0\u4e2a\u6837\u672c\u4e3b\u5bfc\u3001\u6807\u8bb0\u57fa\u56e0\u662f\u5426\u652f\u6301\u8be5\u6ce8\u91ca\uff1b\u6362\u8ba1\u7b97\u8bbe\u5907\u65e0\u6cd5\u89e3\u51b3\u6279\u6b21\u6df7\u6742\u3002Leiden \u5b9e\u73b0<\/a> \u00b7 UMAP \u5b9e\u73b0<\/a><\/p>\n
\u663e\u5b58\u600e\u4e48\u7b97\uff1f\u5148\u770b\u77e9\u9635\u5728\u5185\u5b58\u4e2d\u957f\u4ec0\u4e48\u6837<\/h2>\n
\u4e0a\u4f8b\u5efa\u56fe\u4f7f\u7528 PCA \u5750\u6807\u3002\u5982\u679c\u4ece\u8868\u8fbe\u77e9\u9635\u5f00\u59cb\u505a\u5f52\u4e00\u5316\uff0c\u5e38\u89c1\u7528\u6cd5\u662f rsc.get.anndata_to_GPU(adata)<\/code>\uff0c\u9ed8\u8ba4\u8f6c\u6362 adata.X<\/code>\u3002\u539f\u59cb\u8ba1\u6570\u5982\u679c\u53e6\u5b58\u5728 adata.layers[\"counts\"]<\/code>\uff0c\u53ef\u4ee5\u4e3a\u786e\u5b9e\u8981\u7528\u5b83\u7684\u6b65\u9aa4\u5355\u72ec\u6307\u5b9a\u8be5\u5c42\u3002\u4fdd\u7559\u5728\u4e3b\u673a\u5185\u5b58\u7684\u5907\u7528\u77e9\u9635\u8d8a\u591a\uff0c\u8d8a\u8981\u7559\u610f\u4e3b\u673a\u5185\u5b58\uff1b\u628a\u591a\u4e2a\u5927\u77e9\u9635\u540c\u65f6\u642c\u4e0a\u5361\uff0c\u53c8\u4f1a\u62ac\u9ad8\u663e\u5b58\u9700\u6c42\u3002v0.18.1 \u6570\u636e\u8fc1\u79fb\u63a5\u53e3<\/a><\/p>\n
\u4e00\u4e2a\u7b97\u4f8b\u80fd\u8bf4\u660e\u4e3a\u4ec0\u4e48 h5ad \u6587\u4ef6\u5927\u5c0f\u5f88\u96be\u76f4\u63a5\u6362\u7b97\u6210\u663e\u5b58\u3002\u5047\u8bbe\u6709 10 \u4e07\u4e2a\u7ec6\u80de\u30012 \u4e07\u4e2a\u57fa\u56e0<\/strong>\uff0c\u6bcf\u4e2a\u5143\u7d20\u7528 4 \u5b57\u8282\u7684 float32 \u8868\u793a\uff0c\u90a3\u4e48\u4e00\u4efd\u5b8c\u6574\u7684\u7a20\u5bc6\u77e9\u9635\u5c31\u662f 8 GB<\/strong>\u3002\u540c\u6837\u7ec6\u80de\u6570\u4e0b\uff0c\u82e5\u6bcf\u4e2a\u7ec6\u80de\u53ea\u4fdd\u7559 50 \u4e2a\u4e3b\u6210\u5206\uff0c\u5355\u4efd\u7a20\u5bc6\u5750\u6807\u6570\u7ec4\u5c31\u662f 20 MB<\/strong>\u3002\u8fd9\u91cc\u6309\u5341\u8fdb\u5236\u5355\u4f4d\u8ba1\u7b97\uff0c\u4e24\u9879\u90fd\u662f\u6570\u7ec4\u5927\u5c0f\u7684\u7b97\u672f\u793a\u4f8b\u3002\u771f\u5b9e\u7a00\u758f\u8868\u8fbe\u77e9\u9635\u8fd8\u8981\u5b58\u975e\u96f6\u503c\u548c\u7d22\u5f15\uff0c\u8fd0\u884c\u65f6\u4e5f\u8981\u653e\u90bb\u63a5\u56fe\u3001\u4e34\u65f6\u6570\u7ec4\uff0c\u67d0\u4e9b\u53d8\u6362\u8fd8\u4f1a\u4ea7\u751f\u7a20\u5bc6\u526f\u672c\u3002\u9009\u5361\u8981\u770b\u76ee\u6807\u6b65\u9aa4\u7684\u5cf0\u503c\u663e\u5b58<\/strong>\u3002<\/p>\n
\u4ec0\u4e48\u65f6\u5019\u503c\u5f97\u7528 GPU\uff1f\u7528\u81ea\u5df1\u7684\u4e00\u4e2a\u5206\u6790\u95ee\u9898\u8bd5\u51fa\u6765<\/h2>\n
\n\n\n\n\n\n\n\n\n
\u4f60\u5b9e\u9645\u5728\u7b49\u4ec0\u4e48<\/th>\n\u5148\u505a\u7684\u5224\u65ad<\/th>\n<\/tr>\n<\/thead>\n
\u5408\u5e76\u591a\u4e2a\u6837\u672c\u540e\uff0c\u53cd\u590d\u8c03\u6574\u8868\u793a\u3001\u4e3b\u6210\u5206\u6216\u90bb\u5c45\u6570\uff0c\u6bcf\u8f6e\u90fd\u8981\u91cd\u5efa\u56fe<\/td>\n\u4f18\u5148\u6d4b\u8bd5 GPU \u5efa\u56fe\u3002\u628a\u540c\u4e00\u4efd PCA \u5750\u6807\u4ea4\u7ed9 CPU \u548c GPU \u4e24\u6761\u8def\u7ebf\uff0c\u6bd4\u8f83\u5b8c\u6574\u4e00\u8f6e\u82b1\u591a\u4e45\u3002<\/td>\n<\/tr>\n
\u56fe\u548c\u4e8c\u7ef4\u5750\u6807\u5df2\u7ecf\u4fdd\u5b58\uff0c\u4e3b\u8981\u662f\u770b\u6807\u8bb0\u57fa\u56e0\u3001\u6539\u914d\u8272\u6216\u6838\u5bf9\u6837\u672c\u6807\u7b7e<\/td>\n\u5148\u590d\u7528\u5df2\u6709\u56fe\u548c\u5750\u6807\u3002\u53ea\u6709\u805a\u7c7b\u672c\u8eab\u4e5f\u5360\u53bb\u660e\u663e\u65f6\u95f4\u65f6\uff0c\u518d\u5355\u6d4b\u8fd9\u4e00\u6bb5\u3002<\/td>\n<\/tr>\n
\u6bcf\u8f6e\u90fd\u8981\u4ece\u8fdc\u7aef\u8bfb\u6587\u4ef6\u3001\u4f20\u5230\u663e\u5361\uff0c\u53ea\u7b97\u4e00\u6b65\u5c31\u4f20\u56de\u6765<\/td>\n\u628a\u8bfb\u76d8\u3001\u4f20\u8f93\u3001\u8ba1\u7b97\u548c\u5199\u56de\u5206\u5f00\u8ba1\u65f6\u3002\u82e5\u7b49\u5f85\u96c6\u4e2d\u5728\u8bfb\u5199\uff0c\u5148\u5904\u7406\u6570\u636e\u4f4d\u7f6e\u6216\u91cd\u590d\u52a0\u8f7d\u3002<\/td>\n<\/tr>\n
\u5b50\u96c6\u53ef\u4ee5\u8dd1\uff0c\u5168\u91cf\u4e00\u5230\u9884\u5904\u7406\u5c31\u663e\u5b58\u4e0d\u8db3<\/td>\n\u67e5\u660e\u662f\u54ea\u4e00\u5c42\u3001\u54ea\u4e2a\u64cd\u4f5c\u4ea7\u751f\u526f\u672c\u6216\u7a20\u5bc6\u77e9\u9635\u3002\u5206\u522b\u6d4b\u4e0d\u540c\u89c4\u6a21\u5b50\u96c6\uff0c\u5e76\u5728\u76ee\u6807\u89c4\u6a21\u590d\u6838\uff1b\u4e00\u6b21\u5c0f\u6837\u672c\u8dd1\u901a\u8fd8\u4e0d\u8db3\u4ee5\u5b9a\u5361\u578b\u3002<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n
\u505a\u4e00\u6b21\u516c\u5e73\u7684\u8bd5\u8dd1\u3002<\/strong> \u56fa\u5b9a\u7ec6\u80de\u4e0e\u57fa\u56e0\u3001\u8f93\u5165\u8868\u793a\u3001\u4e3b\u6210\u5206\u6570\u3001\u90bb\u5c45\u6570\u3001\u8ddd\u79bb\u5ea6\u91cf\u548c\u968f\u673a\u79cd\u5b50\uff0c\u540c\u65f6\u8bb0\u5f55 CPU \u7ebf\u7a0b\u6570\u3001GPU \u578b\u53f7\u3001\u8f6f\u4ef6\u7248\u672c\uff0c\u4ee5\u53ca\u8fd1\u90bb\u641c\u7d22\u7528\u7684\u662f\u7cbe\u786e\u7b97\u6cd5\u8fd8\u662f\u8fd1\u4f3c\u7b97\u6cd5\u3002\u8bb0\u4e24\u7b14\u65f6\u95f4\uff1a\u7528\u6237\u771f\u6b63\u7b49\u7684\u201c\u8bfb\u5165\u5230\u4fdd\u5b58\u201d\uff0c\u4ee5\u53ca\u8981\u4f18\u5316\u7684\u201c\u5efa\u56fe\u5230\u7ed3\u679c\u5c31\u7eea\u201d\u3002GPU \u6267\u884c\u53ef\u80fd\u5f02\u6b65\uff0c\u5355\u5361\u5355\u8fdb\u7a0b\u7684\u5206\u6bb5\u8ba1\u65f6\u5e94\u5728\u8fb9\u754c\u540c\u6b65\u8bbe\u5907\uff1b\u9996\u6b21\u521d\u59cb\u5316\u4e0e\u540e\u7eed\u91cd\u590d\u8fd0\u884c\u5206\u5f00\u8bb0\u5f55\u3002\u82e5\u5e73\u65f6\u53ea\u8dd1\u4e00\u6b21\uff0c\u9996\u6b21\u5f00\u9500\u4e5f\u7b97\u8fdb\u8fd9\u6b21\u51b3\u5b9a\u3002CuPy \u5b98\u65b9\u8ba1\u65f6\u6307\u5357<\/a><\/p>\n
\u518d\u770b\u7ed3\u679c\u662f\u5426\u8db3\u4ee5\u56de\u7b54\u539f\u95ee\u9898\u3002<\/strong> \u6bd4\u5982\u8981\u6bd4\u8f83\u8fd1\u90bb\u56fe\uff0c\u5c31\u5bf9\u540c\u4e00\u7ec6\u80de\u53d6\u4e24\u6761\u8def\u7ebf\u5404\u81ea\u7684\u8fd1\u90bb\u540d\u5355\uff0c\u7528\u5171\u540c\u90bb\u5c45\u6570\u9664\u4ee5\u90bb\u5c45\u6570\uff0c\u518d\u5206\u522b\u67e5\u770b\u4e3b\u8981\u7ec6\u80de\u7fa4\u548c\u7a00\u6709\u7fa4\u3002\u82e5\u67d0\u4e2a\u5c0f\u7fa4\u53ea\u5728\u4e00\u79cd\u8bbe\u7f6e\u4e0b\u51fa\u73b0\uff0c\u56de\u770b\u5b83\u7684\u6807\u8bb0\u57fa\u56e0\u3001\u6837\u672c\u6765\u6e90\u548c\u90bb\u5c45\u6784\u6210\u3002UMAP \u7684\u6574\u4f53\u65cb\u8f6c\u6216\u7ffb\u8f6c\u672c\u8eab\u4e0d\u8bf4\u660e\u751f\u7269\u5b66\u7ed3\u8bba\u53d8\u4e86\u3002<\/p>\n
\u5982\u679c\u4e3b\u8981\u7b49\u5f85\u786e\u5b9e\u6765\u81ea\u53ef\u91cd\u590d\u7684 GPU \u8ba1\u7b97\uff0c\u800c\u73b0\u6709\u673a\u5668\u7f3a\u5c11\u5408\u9002\u663e\u5361\uff0c\u518d\u8003\u8651\u79df\u7528\u5c31\u6709\u4e86\u4f9d\u636e\u3002\u54a8\u8be2\u65f6\u5e26\u4e0a\u7ec6\u80de\u6570\u3001\u53c2\u4e0e\u8ba1\u7b97\u7684\u7279\u5f81\u6570\u3001\u77e9\u9635\u683c\u5f0f\u3001\u6700\u6162\u7684\u6b65\u9aa4<\/strong>\uff1b\u6709\u8bd5\u8dd1\u8bb0\u5f55\uff0c\u518d\u8865\u4e0a\u5cf0\u503c\u663e\u5b58\u3001\u6574\u8f6e\u8017\u65f6\u548c\u9884\u8ba1\u91cd\u590d\u6b21\u6570\u3002\u53ef\u4ee5\u636e\u6b64\u548c\u4e91\u751f\u4fe1\u786e\u8ba4\u663e\u5361\u3001\u4e3b\u673a\u5185\u5b58\u3001\u5de5\u4f5c\u76d8\u53ca\u9a71\u52a8\u73af\u5883\uff0c\u628a\u6570\u636e\u4e0a\u4f20\u548c\u73af\u5883\u51c6\u5907\u4e5f\u7b97\u8fdb\u5b9e\u9645\u79df\u671f\u3002\u4e91\u751f\u4fe1 GPU \u79d1\u7814\u8ba1\u7b97\u73af\u5883\u4e0e\u670d\u52a1\u5668\u79df\u8d41<\/a><\/p>\n<\/div>\n","protected":false},"excerpt":{"rendered":"
\u7528\u540c\u4e00\u5f20\u8fd1\u90bb\u56fe\u6bd4\u8f83\u4e24\u79cd Leiden \u5206\u8fa8\u7387\uff0c\u8bfb\u61c2 rapids-singlecell \u7684 GPU \u63a5\u53e3\u3001\u663e\u5b58\u5f00\u9500\u4e0e\u7aef\u5230\u7aef\u8bd5\u8dd1\u65b9\u6cd5\u3002<\/p>\n","protected":false},"author":1,"featured_media":66592,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_jetpack_newsletter_access":"","_jetpack_dont_email_post_to_subs":false,"_jetpack_newsletter_tier_id":0,"_jetpack_memberships_contains_paywalled_content":false,"_jetpack_feature_clip_id":0,"_jetpack_memberships_contains_paid_content":false,"footnotes":"","jetpack_post_was_ever_published":false},"categories":[1,75],"tags":[508,340,556,240,90],"class_list":["post-66593","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-part","category-compute-environment","tag-anndata","tag-gpu","tag-rapids-singlecell","tag-scanpy","tag-90"],"jetpack_sharing_enabled":true,"jetpack_featured_media_url":"https:\/\/i0.wp.com\/www.biocloudservice.com\/wordpress\/wp-content\/uploads\/2026\/11\/repo-guide-20261114-rapids.png?fit=1080%2C860&ssl=1","_links":{"self":[{"href":"https:\/\/www.biocloudservice.com\/wordpress\/wp-json\/wp\/v2\/posts\/66593","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.biocloudservice.com\/wordpress\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.biocloudservice.com\/wordpress\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.biocloudservice.com\/wordpress\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.biocloudservice.com\/wordpress\/wp-json\/wp\/v2\/comments?post=66593"}],"version-history":[{"count":1,"href":"https:\/\/www.biocloudservice.com\/wordpress\/wp-json\/wp\/v2\/posts\/66593\/revisions"}],"predecessor-version":[{"id":66594,"href":"https:\/\/www.biocloudservice.com\/wordpress\/wp-json\/wp\/v2\/posts\/66593\/revisions\/66594"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.biocloudservice.com\/wordpress\/wp-json\/wp\/v2\/media\/66592"}],"wp:attachment":[{"href":"https:\/\/www.biocloudservice.com\/wordpress\/wp-json\/wp\/v2\/media?parent=66593"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.biocloudservice.com\/wordpress\/wp-json\/wp\/v2\/categories?post=66593"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.biocloudservice.com\/wordpress\/wp-json\/wp\/v2\/tags?post=66593"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}