{"id":795,"date":"2024-06-25T13:44:08","date_gmt":"2024-06-25T05:44:08","guid":{"rendered":"http:\/\/ai.gitpp.com\/?p=795"},"modified":"2024-06-25T13:44:08","modified_gmt":"2024-06-25T05:44:08","slug":"pytorch-%e5%88%9d%e5%ad%a6%e8%80%85%e6%95%99%e7%a8%8b%ef%bc%9a%e6%89%80%e6%9c%89%e5%9f%ba%e7%a1%80%e7%9f%a5%e8%af%86","status":"publish","type":"post","link":"http:\/\/ai.gitpp.com\/index.php\/2024\/06\/25\/pytorch-%e5%88%9d%e5%ad%a6%e8%80%85%e6%95%99%e7%a8%8b%ef%bc%9a%e6%89%80%e6%9c%89%e5%9f%ba%e7%a1%80%e7%9f%a5%e8%af%86\/","title":{"rendered":"PyTorch \u521d\u5b66\u8005\u6559\u7a0b\uff1a\u6240\u6709\u57fa\u7840\u77e5\u8bc6"},"content":{"rendered":"\n

<\/p>\n\n\n\n

<\/p>\n\n\n\n

\u4e16\u754c\u5728\u4e0d\u65ad\u53d1\u5c55\uff0c\u670d\u52a1\u4e8e\u4e16\u754c\u7684\u6280\u672f\u4e5f\u5728\u4e0d\u65ad\u53d1\u5c55\u3002\u6bcf\u4e2a\u4eba\u90fd\u5fc5\u987b\u8ddf\u4e0a\u6280\u672f\u7684\u5feb\u901f\u53d8\u5316\u3002\u4eba\u5de5\u667a\u80fd\u662f\u53d1\u5c55\u901f\u5ea6\u6700\u5feb\u3001\u89c4\u6a21\u6700\u5927\u7684\u9886\u57df\u4e4b\u4e00\u3002<\/p>\n\n\n\n

\u6211\u4eec\u6b63\u5728\u8bad\u7ec3\u673a\u5668\u5b66\u4e60\uff0c\u7ed3\u679c\u8d8a\u6765\u8d8a\u597d\u3002\u6709\u53ef\u4ee5\u751f\u6210\u65b0\u56fe\u50cf\u7684\u751f\u6210\u7f51\u7edc (GAN)\u3001\u7528\u4e8e\u5c06\u624b\u8bed\u7ffb\u8bd1\u6210\u6587\u672c\u7684\u6df1\u5ea6\u5b66\u4e60\u6a21\u578b\u7b49\u7b49\uff01\u5728\u8fd9\u4e2a\u5feb\u901f\u53d1\u5c55\u7684\u9886\u57df\uff0cPyTorch \u662f\u6784\u5efa\u8fd9\u4e9b\u6a21\u578b\u7684\u66ff\u4ee3\u65b9\u6848\u3002<\/p>\n\n\n\n

\u8003\u8651\u5230\u4e86\u89e3 PyTorch \u7684\u6240\u6709\u4f18\u70b9\uff0c\u6211\u4eec\u64b0\u5199\u4e86\u4e00\u7cfb\u5217\u5173\u4e8e\u4f7f\u7528 PyTorch \u8fdb\u884c\u6df1\u5ea6\u5b66\u4e60\u7684\u6587\u7ae0\u2014\u2014\u9762\u5411\u521d\u5b66\u8005\u7684 Pytorch \u6559\u7a0b\u3002\u5728\u672c\u5165\u95e8\u8bfe\u7a0b\u4e2d\uff0c\u6211\u4eec\u5c06\u4ecb\u7ecd\u4ee5\u4e0b\u4e3b\u9898\u3002<\/p>\n\n\n\n

PyTorch \u521d\u5b66\u8005\u6307\u5357<\/th><\/tr><\/thead>
PyTorch \u521d\u5b66\u8005\u6307\u5357\uff1a\u57fa\u7840\u77e5\u8bc6<\/td><\/tr>
PyTorch \u521d\u5b66\u8005\u6307\u5357\uff1a\u4f7f\u7528\u9884\u8bad\u7ec3\u6a21\u578b\u8fdb\u884c\u56fe\u50cf\u5206\u7c7b<\/td><\/tr>
\u5728 PyTorch \u4e2d\u4f7f\u7528\u8fc1\u79fb\u5b66\u4e60\u8fdb\u884c\u56fe\u50cf\u5206\u7c7b<\/td><\/tr>
\u4f7f\u7528 ONNX \u548c Caffe2 \u7684 PyTorch \u6a21\u578b\u63a8\u7406<\/td><\/tr>
PyTorch \u521d\u5b66\u8005\u6307\u5357\uff1a\u4f7f\u7528 torchvision \u8fdb\u884c\u8bed\u4e49\u5206\u5272<\/td><\/tr>
\u7269\u4f53\u68c0\u6d4b<\/td><\/tr>
\u5b9e\u4f8b\u5206\u5272<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

\u4ec0\u4e48\u662f PyTorch\uff1f<\/h2>\n\n\n\n

PyTorch \u662f\u4e00\u4e2a\u57fa\u4e8e Python \u7684\u5e93\uff0c\u5b83\u6709\u52a9\u4e8e\u6784\u5efa\u6df1\u5ea6\u5b66\u4e60\u6a21\u578b\u5e76\u5c06\u5176\u7528\u4e8e\u5404\u79cd\u5e94\u7528\u7a0b\u5e8f\u3002\u4f46\u8fd9\u4e0d\u4ec5\u4ec5\u662f\u53e6\u4e00\u4e2a\u6df1\u5ea6\u5b66\u4e60\u5e93\u3002\u5b83\u662f\u4e00\u4e2a\u79d1\u5b66\u8ba1\u7b97\u5305\uff08\u6b63\u5982\u5b98\u65b9 PyTorch \u6587\u6863\u6240\u8ff0\uff09\u3002<\/p>\n\n\n\n

\n

\u5b83\u662f\u4e00\u4e2a\u57fa\u4e8e Python \u7684\u79d1\u5b66\u8ba1\u7b97\u5305\uff0c\u9488\u5bf9\u4e24\u7c7b\u53d7\u4f17\uff1a
1. \u66ff\u4ee3 NumPy \u4ee5\u5229\u7528 GPU \u7684\u529f\u80fd
2. \u63d0\u4f9b\u6700\u5927\u7075\u6d3b\u6027\u548c\u901f\u5ea6\u7684\u6df1\u5ea6\u5b66\u4e60\u7814\u7a76\u5e73\u53f0\u4f7f\u7528 PyTorch \u8fdb\u884c\u6df1\u5ea6\u5b66\u4e60\uff1a60 \u5206\u949f\u901f\u6210<\/p>\n<\/blockquote>\n\n\n\n

PyTorch \u4f7f\u7528Tensor\u4f5c\u4e3a\u5176\u6838\u5fc3\u6570\u636e\u7ed3\u6784\uff0c\u7c7b\u4f3c\u4e8e Numpy \u6570\u7ec4\u3002\u60a8\u53ef\u80fd\u5bf9\u8fd9\u79cd\u7279\u5b9a\u7684\u6570\u636e\u7ed3\u6784\u9009\u62e9\u611f\u5230\u7591\u60d1\u3002\u7b54\u6848\u5728\u4e8e\uff0c\u6709\u4e86\u9002\u5f53\u7684\u8f6f\u4ef6\u548c\u786c\u4ef6\uff0c\u5f20\u91cf\u53ef\u4ee5\u52a0\u901f\u5404\u79cd\u6570\u5b66\u8fd0\u7b97\u3002\u5f53\u8fd9\u4e9b\u8fd0\u7b97\u5728\u6df1\u5ea6\u5b66\u4e60\u4e2d\u5927\u91cf\u6267\u884c\u65f6\uff0c\u901f\u5ea6\u4f1a\u6709\u5de8\u5927\u7684\u5dee\u5f02\u3002<\/p>\n\n\n\n

PyTorch \u4e0e Python \u7c7b\u4f3c\uff0c\u6ce8\u91cd\u6613\u7528\u6027\uff0c\u5373\u4f7f\u662f\u7f16\u7a0b\u77e5\u8bc6\u975e\u5e38\u57fa\u7840\u7684\u7528\u6237\u4e5f\u53ef\u4ee5\u5728\u9879\u76ee\u4e2d\u4f7f\u7528\u6df1\u5ea6\u5b66\u4e60\u3002\u5982\u679c\u60a8\u8fd8\u4e0d\u4e86\u89e3\u6df1\u5ea6\u5b66\u4e60\u5e93\uff0c\u90a3\u4e48 PyTorch \u5c31\u662f\u5b8c\u7f8e\u7684\u201c\u7b2c\u4e00\u4e2a\u8981\u5b66\u4e60\u7684\u6df1\u5ea6\u5b66\u4e60\u5e93\u201d\u3002<\/em><\/p>\n\n\n\n

\"\u56fe\u7247\"<\/figure>\n\n\n\n

\u4e3a\u4ec0\u4e48\u8981\u5b66\u4e60 PyTorch\uff1f\u2013 Pytorch \u521d\u5b66\u8005\u6559\u7a0b<\/h2>\n\n\n\n

\u5728\u4e0a\u4e00\u8282\u4e2d\uff0c\u6211\u4eec\u63d0\u5230 PyTorch \u662f\u4f60\u5e94\u8be5\u5b66\u4e60\u7684\u7b2c\u4e00\u4e2a\u6df1\u5ea6\u5b66\u4e60\u5e93\u7684\u5b8c\u7f8e\u9009\u62e9\u3002\u5728\u672c\u8282\u4e2d\uff0c\u6211\u4eec\u5c06\u8be6\u7ec6\u8bf4\u660e\u4e3a\u4ec0\u4e48\u5b83\u5982\u6b64\u3002<\/p>\n\n\n\n

\u6df1\u5ea6\u5b66\u4e60\u5e93\u5e76\u4e0d\u7f3a\u4e4f\uff1aKeras\u3001Tensorflow\u3001Caffe\u3001Theano\uff08RIP\uff09\u7b49\u7b49\u3002PyTorch \u6709\u4f55\u4e0d\u540c\uff1f<\/p>\n\n\n\n

\u7406\u60f3\u7684\u6df1\u5ea6\u5b66\u4e60\u5e93\u5e94\u8be5\u6613\u4e8e\u5b66\u4e60\u548c\u4f7f\u7528\uff0c\u8db3\u591f\u7075\u6d3b\u4ee5\u7528\u4e8e\u5404\u79cd\u5e94\u7528\u7a0b\u5e8f\uff0c\u9ad8\u6548\u4ee5\u4fbf\u6211\u4eec\u53ef\u4ee5\u5904\u7406\u5e9e\u5927\u7684\u73b0\u5b9e\u6570\u636e\u96c6\uff0c\u5e76\u4e14\u8db3\u591f\u51c6\u786e\u4ee5\u4fbf\u5728\u8f93\u5165\u6570\u636e\u5b58\u5728\u4e0d\u786e\u5b9a\u6027\u7684\u60c5\u51b5\u4e0b\u4e5f\u80fd\u63d0\u4f9b\u6b63\u786e\u7684\u7ed3\u679c\u3002<\/p>\n\n\n\n

PyTorch \u5728\u4e0a\u8ff0\u6240\u6709\u6307\u6807\u4e0a\u90fd\u8868\u73b0\u5f97\u975e\u5e38\u51fa\u8272\u3002\u201c pythonic \u201d\u7f16\u7801\u98ce\u683c\u4f7f\u5176\u6613\u4e8e\u5b66\u4e60\u548c\u4f7f\u7528\u3002GPU\u52a0\u901f\u3001\u5bf9\u5206\u5e03\u5f0f\u8ba1\u7b97\u7684\u652f\u6301\u548c\u81ea\u52a8\u68af\u5ea6\u8ba1\u7b97\u6709\u52a9\u4e8e\u4ece\u524d\u5411\u8868\u8fbe\u5f0f\u81ea\u52a8\u6267\u884c\u540e\u5411\u4f20\u9012\u3002<\/p>\n\n\n\n

\u5f53\u7136\uff0c\u7531\u4e8e\u4f7f\u7528 Python\uff0c\u5b83\u9762\u4e34\u8fd0\u884c\u901f\u5ea6\u6162\u7684\u98ce\u9669\uff0c\u4f46\u9ad8\u6027\u80fd C++ API\uff08libtorch\uff09\u6d88\u9664\u4e86\u8fd9\u79cd\u5f00\u9500\u3002\u8fd9\u4f7f\u5f97\u4ece\u7814\u53d1\u5230\u751f\u4ea7\u7684\u8fc7\u6e21\u975e\u5e38\u987a\u5229\u3002\u8fd9\u662f\u4f7f\u7528 PyTorch \u7684\u53e6\u4e00\u4e2a\u7406\u7531\uff01<\/p>\n\n\n\n

\u8ba9\u6211\u4eec\u770b\u4e00\u4e0b\u4f7f\u7528 PyTorch \u5e94\u7528\u7a0b\u5e8f\u53ef\u4ee5\u83b7\u5f97\u7684\u4e00\u4e9b\u6709\u8da3\u7684\u7ed3\u679c\u3002<\/p>\n\n\n\n

\"\u56fe\u7247\"\/<\/figure>\n\n\n\n
\"\u56fe\u7247\"\/<\/figure>\n\n\n\n

\u4e3a\u4e86\u8ba9\u60a8\u5bf9 PyTorch \u66f4\u52a0\u7740\u8ff7\uff0c\u8fd9\u91cc\u5217\u51fa\u4e86\u6d89\u53ca PyTorch \u7684\u975e\u5e38\u9177\u7684\u9879\u76ee\u3002<\/p>\n\n\n\n

PyTorch \u5e93\u6982\u8ff0<\/h2>\n\n\n\n

\u73b0\u5728\u6211\u4eec\u5df2\u7ecf\u4e86\u89e3\u4e86 PyTorch \u53ca\u5176\u72ec\u7279\u4e4b\u5904\uff0c\u8ba9\u6211\u4eec\u6765\u770b\u770b PyTorch \u9879\u76ee\u7684\u57fa\u672c\u6d41\u7a0b\u3002\u4e0b\u56fe\u63cf\u8ff0\u4e86\u5178\u578b\u7684\u5de5\u4f5c\u6d41\u7a0b\u4ee5\u53ca\u4e0e\u6bcf\u4e2a\u6b65\u9aa4\u76f8\u5173\u7684\u91cd\u8981\u6a21\u5757\u3002<\/p>\n\n\n\n

\"\u56fe\u7247\"
\u57fa\u672c PyTorch \u5de5\u4f5c\u6d41\u7a0b<\/figcaption><\/figure>\n\n\n\n

\u6211\u4eec\u5c06\u5728\u8fd9\u91cc\u7b80\u8981\u8ba8\u8bba\u7684\u91cd\u8981 PyTorch \u6a21\u5757\u662f\uff1atorch.nn\u3001torch.optim\u3001torch.utils \u548c torch.autograd\u3002<\/p>\n\n\n\n

1. \u6570\u636e\u52a0\u8f7d\u548c\u5904\u7406<\/h3>\n\n\n\n

\u4efb\u4f55\u6df1\u5ea6\u5b66\u4e60\u9879\u76ee\u7684\u7b2c\u4e00\u6b65\u90fd\u662f\u5904\u7406\u6570\u636e\u52a0\u8f7d\u548c\u5904\u7406\u3002PyTorch \u901a\u8fc7torch.utils.data \u63d0\u4f9b\u76f8\u5173\u5b9e\u7528\u7a0b\u5e8f\u3002<\/p>\n\n\n\n

\u8be5\u6a21\u5757\u4e2d\u7684\u4e24\u4e2a\u91cd\u8981\u7c7b\u662fDataset\u548cDataLoader\u3002<\/p>\n\n\n\n

    \n
  • \u6570\u636e\u96c6\u5efa\u7acb\u5728Tensor\u6570\u636e\u7c7b\u578b\u4e4b\u4e0a\uff0c\u4e3b\u8981\u7528\u4e8e\u81ea\u5b9a\u4e49\u6570\u636e\u96c6\u3002<\/li>\n\n\n\n
  • \u5f53\u60a8\u62e5\u6709\u5927\u578b\u6570\u636e\u96c6\u5e76\u4e14\u60f3\u8981\u4ece\u540e\u53f0\u7684\u6570\u636e\u96c6\u52a0\u8f7d\u6570\u636e\u4ee5\u4fbf\u5176\u51c6\u5907\u5c31\u7eea\u5e76\u7b49\u5f85\u8bad\u7ec3\u5faa\u73af\u65f6\uff0c\u5c06\u4f7f\u7528DataLoader \u3002<\/li>\n<\/ul>\n\n\n\n

    \u5982\u679c\u6211\u4eec\u53ef\u4ee5\u8bbf\u95ee\u591a\u53f0\u673a\u5668\u6216 GPU\uff0c\u6211\u4eec\u8fd8\u53ef\u4ee5\u4f7f\u7528torch.nn.DataParallel\u548ctorch.distributed\u3002<\/p>\n\n\n\n

    2. \u5efa\u7acb\u795e\u7ecf\u7f51\u7edc<\/h3>\n\n\n\n

    torch.nn\u6a21\u5757\u7528\u4e8e\u521b\u5efa\u795e\u7ecf\u7f51\u7edc\u3002\u5b83\u63d0\u4f9b\u6240\u6709\u5e38\u89c1\u7684\u795e\u7ecf\u7f51\u7edc\u5c42\uff0c\u4f8b\u5982\u5168\u8fde\u63a5\u5c42\u3001\u5377\u79ef\u5c42\u3001\u6fc0\u6d3b\u548c\u635f\u5931\u51fd\u6570\u7b49\u3002<\/p>\n\n\n\n

    \u4e00\u65e6\u521b\u5efa\u4e86\u7f51\u7edc\u67b6\u6784\u5e76\u4e14\u6570\u636e\u5df2\u51c6\u5907\u597d\u8f93\u5165\u5230\u7f51\u7edc\uff0c\u6211\u4eec\u5c31\u9700\u8981\u4e00\u4e9b\u6280\u672f\u6765\u66f4\u65b0\u6743\u91cd\u548c\u504f\u5dee\uff0c\u4ee5\u4fbf\u7f51\u7edc\u5f00\u59cb\u5b66\u4e60\u3002\u8fd9\u4e9b\u5b9e\u7528\u7a0b\u5e8f\u5728torch.optim\u6a21\u5757\u4e2d\u63d0\u4f9b\u3002\u540c\u6837\uff0c\u5bf9\u4e8e\u53cd\u5411\u4f20\u9012\u8fc7\u7a0b\u4e2d\u6240\u9700\u7684\u81ea\u52a8\u5fae\u5206\uff0c\u6211\u4eec\u4f7f\u7528torch.autograd\u6a21\u5757\u3002<\/p>\n\n\n\n

    3. \u6a21\u578b\u63a8\u7406\u4e0e\u517c\u5bb9\u6027<\/h3>\n\n\n\n

    \u6a21\u578b\u8bad\u7ec3\u5b8c\u6210\u540e\uff0c\u53ef\u4ee5\u7528\u4e8e\u9884\u6d4b\u6d4b\u8bd5\u7528\u4f8b\u751a\u81f3\u65b0\u6570\u636e\u96c6\u7684\u8f93\u51fa\u3002\u8fd9\u4e2a\u8fc7\u7a0b\u79f0\u4e3a\u6a21\u578b\u63a8\u7406\u3002<\/p>\n\n\n\n

    PyTorch \u8fd8\u63d0\u4f9bTorchScript\uff0c\u53ef\u7528\u4e8e\u72ec\u7acb\u4e8e Python \u8fd0\u884c\u65f6\u8fd0\u884c\u6a21\u578b\u3002\u8fd9\u53ef\u4ee5\u88ab\u8ba4\u4e3a\u662f\u4e00\u4e2a\u865a\u62df\u673a\uff0c\u5176\u6307\u4ee4\u4e3b\u8981\u9488\u5bf9\u5f20\u91cf\u3002<\/p>\n\n\n\n

    \u60a8\u8fd8\u53ef\u4ee5\u5c06\u4f7f\u7528 PyTorch \u8bad\u7ec3\u7684\u6a21\u578b\u8f6c\u6362\u4e3aONNX \u7b49\u683c\u5f0f\uff0c\u8fd9\u6837\u60a8\u5c31\u53ef\u4ee5\u5728\u5176\u4ed6 DL \u6846\u67b6\uff08\u5982 MXNet\u3001CNTK\u3001Caffe2\uff09\u4e2d\u4f7f\u7528\u8fd9\u4e9b\u6a21\u578b\u3002\u60a8\u8fd8\u53ef\u4ee5\u5c06onnx\u6a21\u578b\u8f6c\u6362\u4e3a Tensorflow\u3002<\/p>\n\n\n\n

    \u5f20\u91cf\u7b80\u4ecb<\/h2>\n\n\n\n

    \u5230\u76ee\u524d\u4e3a\u6b62\uff0c\u5728\u8fd9\u7bc7\u6587\u7ae0\u4e2d\uff0c\u6211\u4eec\u5df2\u7ecf\u8ba8\u8bba\u4e86\u4e3a\u4ec0\u4e48\u4f60\u5e94\u8be5\u5b66\u4e60 PyTorch\u3002\u73b0\u5728\u662f\u65f6\u5019\u5f00\u59cb\u540c\u6837\u7684\u65c5\u7a0b\u4e86\u3002\u6211\u4eec\u5c06\u4ece Tensors\uff08PyTorch \u4e2d\u4f7f\u7528\u7684\u6838\u5fc3\u6570\u636e\u7ed3\u6784\uff09\u5f00\u59cb\u3002<\/p>\n\n\n\n

    \u5f20\u91cf\u53ea\u662f\u77e9\u9635\u7684\u4e00\u4e2a\u82b1\u54e8\u540d\u79f0\u3002\u5982\u679c\u60a8\u719f\u6089 NumPy \u6570\u7ec4\uff0c\u90a3\u4e48\u7406\u89e3\u548c\u4f7f\u7528 PyTorch \u5f20\u91cf\u5c06\u975e\u5e38\u5bb9\u6613\u3002\u6807\u91cf\u503c\u7531 0 \u7ef4\u5f20\u91cf\u8868\u793a\u3002\u7c7b\u4f3c\u5730\uff0c\u5217\/\u884c\u77e9\u9635\u4f7f\u7528 1-D \u5f20\u91cf\u8868\u793a\uff0c\u4f9d\u6b64\u7c7b\u63a8\u3002\u4ee5\u4e0b\u663e\u793a\u4e86\u5177\u6709\u4e0d\u540c\u7ef4\u5ea6\u7684\u5f20\u91cf\u7684\u4e00\u4e9b\u793a\u4f8b\uff0c\u4ee5\u4fbf\u60a8\u76f4\u89c2\u5730\u4e86\u89e3\u3002<\/p>\n\n\n\n

    \"\u56fe\u7247\"
    PyTorch \u4e2d\u5f20\u91cf\u7684\u7ef4\u5ea6<\/figcaption><\/figure>\n\n\n\n

    \u5728\u6211\u4eec\u5f00\u59cb\u4ecb\u7ecd Tensors \u4e4b\u524d\uff0c\u8ba9\u6211\u4eec\u901a\u8fc7\u8fd0\u884c\u4e0b\u9762\u7ed9\u51fa\u7684\u547d\u4ee4\u6765\u5b89\u88c5 PyTorch 1.1.0\u3002<\/p>\n\n\n\n

    1<\/td>conda <\/code>install<\/code> -c pytorch pytorch-cpu<\/code><\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

    \u5c31\u8fd9\u6837\uff0c\u60a8\u73b0\u5728\u53ef\u4ee5\u5f00\u59cb\u4f7f\u7528 PyTorch \u4e86\uff01\u73b0\u5728\u8ba9\u6211\u4eec\u5f00\u59cb\u5427\u3002\u6211\u4eec\u5efa\u8bae\u60a8\u4f7f\u7528Google Colab\u5e76\u8ddf\u7740\u505a\u3002\u4ece\u83dc\u5355\u4e2d\u9009\u62e9 GPU \u8fd0\u884c\u65f6\u7c7b\u578b\u3002<\/p>\n\n\n\n

    \u6784\u5efa\u4f60\u7684\u7b2c\u4e00\u4e2a\u5f20\u91cf<\/h3>\n\n\n\n

    \u8ba9\u6211\u4eec\u770b\u770b\u5982\u4f55\u521b\u5efa PyTorch Tensor\u3002\u9996\u5148\uff0c\u6211\u4eec\u5c06\u5bfc\u5165 PyTorch\u3002<\/p>\n\n\n\n

    \u4e0b\u8f7d\u4ee3\u7801 \u4e3a\u4e86\u8f7b\u677e\u5b66\u4e60\u672c\u6559\u7a0b\uff0c\u8bf7\u70b9\u51fb\u4e0b\u9762\u7684\u6309\u94ae\u4e0b\u8f7d\u4ee3\u7801\u3002\u514d\u8d39\uff01
    <\/p>\n\n\n\n

    \u4e0b\u8f7d\u4ee3\u7801<\/p>\n\n\n\n

    123456789101112\u5341\u4e091415<\/td>import<\/code> torch<\/code> # Create a Tensor with just ones in a column<\/code>a <\/code>=<\/code> torch.ones(<\/code>5<\/code>)<\/code> # Print the tensor we created<\/code>print<\/code>(a)<\/code> # tensor([1., 1., 1., 1., 1.])<\/code> # Create a Tensor with just zeros in a column<\/code>b <\/code>=<\/code> torch.zeros(<\/code>5<\/code>)<\/code>print<\/code>(b)<\/code> # tensor([0., 0., 0., 0., 0.])<\/code><\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

    \u6211\u4eec\u53ef\u4ee5\u7c7b\u4f3c\u5730\u521b\u5efa\u5177\u6709\u81ea\u5b9a\u4e49\u503c\u7684\u5f20\u91cf\uff0c\u5982\u4e0b\u6240\u793a\u3002<\/p>\n\n\n\n

    1234<\/td>c <\/code>=<\/code> torch.tensor([<\/code>1.0<\/code>, <\/code>2.0<\/code>, <\/code>3.0<\/code>, <\/code>4.0<\/code>, <\/code>5.0<\/code>])<\/code>print<\/code>(c)<\/code> # tensor([1., 2., 3., 4., 5.])<\/code><\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

    \u5728\u4e0a\u8ff0\u6240\u6709\u60c5\u51b5\u4e0b\uff0c\u6211\u4eec\u90fd\u521b\u5efa\u4e86\u5355\u7ef4\u5411\u91cf\u6216\u5f20\u91cf\u3002\u73b0\u5728\uff0c\u8ba9\u6211\u4eec\u521b\u5efa\u4e00\u4e9b\u66f4\u9ad8\u7ef4\u5ea6\u7684\u5f20\u91cf\u3002<\/p>\n\n\n\n

    123456789101112\u5341\u4e091415161718192021222324\u4e8c\u5341\u4e94\u4e8c\u5341\u516d\u4e8c\u5341\u4e03\u4e8c\u5341\u516b\u4e8c\u5341\u4e5d<\/td>d <\/code>=<\/code> torch.zeros(<\/code>3<\/code>,<\/code>2<\/code>)<\/code>print<\/code>(d)<\/code> # tensor([[0., 0.],<\/code>#        [0., 0.],<\/code>#        [0., 0.]])<\/code> e <\/code>=<\/code> torch.ones(<\/code>3<\/code>,<\/code>2<\/code>)<\/code>print<\/code>(e)<\/code> # tensor([[1., 1.],<\/code>#        [1., 1.],<\/code>#        [1., 1.]])<\/code> f <\/code>=<\/code> torch.tensor([[<\/code>1.0<\/code>, <\/code>2.0<\/code>],[<\/code>3.0<\/code>, <\/code>4.0<\/code>]])<\/code>print<\/code>(f)<\/code> # tensor([[1., 2.],<\/code>#        [3., 4.]])<\/code> # 3D Tensor<\/code>g <\/code>=<\/code> torch.tensor([[[<\/code>1.<\/code>, <\/code>2.<\/code>], [<\/code>3.<\/code>, <\/code>4.<\/code>]], [[<\/code>5.<\/code>, <\/code>6.<\/code>], [<\/code>7.<\/code>, <\/code>8.<\/code>]]])<\/code>print<\/code>(g)<\/code> # tensor([[[1., 2.],<\/code>#         [3., 4.]],<\/code>#<\/code>#        [[5., 6.],<\/code>#         [7., 8.]]])<\/code><\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

    \u6211\u4eec\u8fd8\u53ef\u4ee5\u4f7f\u7528shape\u65b9\u6cd5\u627e\u51fa\u5f20\u91cf\u7684\u5f62\u72b6\u3002<\/p>\n\n\n\n

    12345678<\/td>print<\/code>(f.shape)<\/code># torch.Size([2, 2])<\/code> print<\/code>(e.shape)<\/code># torch.Size([3, 2])<\/code> print<\/code>(g.shape)<\/code># torch.Size([2, 2, 2])<\/code><\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

    \u8bbf\u95ee Tensor \u4e2d\u7684\u4e00\u4e2a\u5143\u7d20<\/h3>\n\n\n\n

    \u73b0\u5728\u6211\u4eec\u5df2\u7ecf\u521b\u5efa\u4e86\u4e00\u4e9b\u5f20\u91cf\uff0c\u8ba9\u6211\u4eec\u770b\u770b\u5982\u4f55\u8bbf\u95ee\u5f20\u91cf\u4e2d\u7684\u5143\u7d20\u3002\u9996\u5148\u8ba9\u6211\u4eec\u770b\u770b\u5982\u4f55\u5bf9 1D \u5f20\u91cf\uff08\u53c8\u79f0\u5411\u91cf\uff09\u6267\u884c\u6b64\u64cd\u4f5c\u3002<\/p>\n\n\n\n

    1234<\/td># Get element at index 2<\/code>print<\/code>(c[<\/code>2<\/code>])<\/code> # tensor(3.)<\/code><\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

    \u90a3\u4e48 2D \u6216 3D \u5f20\u91cf\u5462\uff1f\u56de\u60f3\u4e00\u4e0b\u6211\u4eec\u5728\u4e0a\u4e00\u8282\u4e2d\u63d0\u5230\u7684\u5f20\u91cf\u7ef4\u5ea6\u3002\u8981\u8bbf\u95ee\u5f20\u91cf\u4e2d\u7684\u4e00\u4e2a\u7279\u5b9a\u5143\u7d20\uff0c\u6211\u4eec\u9700\u8981\u6307\u5b9a\u7b49\u4e8e\u5f20\u91cf\u7ef4\u5ea6\u7684\u7d22\u5f15\u3002\u8fd9\u5c31\u662f\u4e3a\u4ec0\u4e48\u5bf9\u4e8e\u5f20\u91cfc\u6211\u4eec\u53ea\u9700\u8981\u6307\u5b9a\u4e00\u4e2a\u7d22\u5f15\u3002<\/p>\n\n\n\n

    123456789101112\u5341\u4e0914<\/td># All indices starting from 0<\/code> # Get element at row 1, column 0<\/code>print<\/code>(f[<\/code>1<\/code>,<\/code>0<\/code>])<\/code># We can also use the following<\/code>print<\/code>(f[<\/code>1<\/code>][<\/code>0<\/code>])<\/code> # tensor(3.)<\/code> # Similarly for 3D Tensor<\/code>print<\/code>(g[<\/code>1<\/code>,<\/code>0<\/code>,<\/code>0<\/code>])<\/code>print<\/code>(g[<\/code>1<\/code>][<\/code>0<\/code>][<\/code>0<\/code>])<\/code> # tensor(5.)<\/code><\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

    \u4f46\u662f\u5982\u679c\u4f60\u60f3\u8bbf\u95ee 2D \u5f20\u91cf\u4e2d\u7684\u6574\u884c\u600e\u4e48\u529e\uff1f\u6211\u4eec\u53ef\u4ee5\u4f7f\u7528\u4e0e NumPy \u6570\u7ec4\u4e2d\u76f8\u540c\u7684\u8bed\u6cd5\u3002<\/p>\n\n\n\n

    123456789101112\u5341\u4e0914<\/td># All elements<\/code>print<\/code>(f[:])<\/code> # All elements from index 1 to 2 (inclusive)<\/code>print<\/code>(c[<\/code>1<\/code>:<\/code>3<\/code>])<\/code> # All elements till index 4 (exclusive)<\/code>print<\/code>(c[:<\/code>4<\/code>])<\/code> # First row<\/code>print<\/code>(f[<\/code>0<\/code>,:])<\/code> # Second column<\/code>print<\/code>(f[:,<\/code>1<\/code>])<\/code><\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

    \u6307\u5b9a\u5143\u7d20\u7684\u6570\u636e\u7c7b\u578b<\/h3>\n\n\n\n

    \u6bcf\u5f53\u6211\u4eec\u521b\u5efa\u4e00\u4e2a\u5f20\u91cf\u65f6\uff0cPyTorch \u90fd\u4f1a\u51b3\u5b9a\u5f20\u91cf\u5143\u7d20\u7684\u6570\u636e\u7c7b\u578b\uff0c\u4ee5\u4fbf\u6570\u636e\u7c7b\u578b\u53ef\u4ee5\u8986\u76d6\u6240\u6709\u5f20\u91cf\u5143\u7d20\u3002\u6211\u4eec\u53ef\u4ee5\u5728\u521b\u5efa\u5f20\u91cf\u65f6\u901a\u8fc7\u6307\u5b9a\u6570\u636e\u7c7b\u578b\u6765\u8986\u76d6\u5b83\u3002<\/p>\n\n\n\n

    123456789101112\u5341\u4e091415161718192021222324\u4e8c\u5341\u4e94\u4e8c\u5341\u516d<\/td>int_tensor <\/code>=<\/code> torch.tensor([[<\/code>1<\/code>,<\/code>2<\/code>,<\/code>3<\/code>],[<\/code>4<\/code>,<\/code>5<\/code>,<\/code>6<\/code>]])<\/code>print<\/code>(int_tensor.dtype)<\/code> # torch.int64<\/code> # What if we changed any one element to floating point number?<\/code>int_tensor <\/code>=<\/code> torch.tensor([[<\/code>1<\/code>,<\/code>2<\/code>,<\/code>3<\/code>],[<\/code>4.<\/code>,<\/code>5<\/code>,<\/code>6<\/code>]])<\/code>print<\/code>(int_tensor.dtype)<\/code> # torch.float32<\/code> print<\/code>(int_tensor)<\/code> # tensor([[1., 2., 3.],<\/code>#        [4., 5., 6.]])<\/code>  # This can be overridden as follows<\/code>int_tensor <\/code>=<\/code> torch.tensor([[<\/code>1<\/code>,<\/code>2<\/code>,<\/code>3<\/code>],[<\/code>4.<\/code>,<\/code>5<\/code>,<\/code>6<\/code>]], dtype<\/code>=<\/code>torch.int32)<\/code>print<\/code>(int_tensor.dtype)<\/code> # torch.int32<\/code>print<\/code>(int_tensor)<\/code> # tensor([[1, 2, 3],<\/code>#        [4, 5, 6]], dtype=torch.int32)<\/code><\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

    \u5f20\u91cf\u5230\/\u6765\u81ea NumPy \u6570\u7ec4<\/h3>\n\n\n\n

    \u6211\u4eec\u66fe\u591a\u6b21\u63d0\u5230 PyTorch \u5f20\u91cf\u548c NumPy \u6570\u7ec4\u975e\u5e38\u76f8\u4f3c\u3002\u8fd9\u5f15\u53d1\u4e86\u4e00\u4e2a\u95ee\u9898\uff1a\u662f\u5426\u53ef\u4ee5\u5c06\u4e00\u79cd\u6570\u636e\u7ed3\u6784\u8f6c\u6362\u4e3a\u53e6\u4e00\u79cd\u6570\u636e\u7ed3\u6784\u3002\u8ba9\u6211\u4eec\u770b\u770b\u5982\u4f55\u505a\u5230\u8fd9\u4e00\u70b9\u3002<\/p>\n\n\n\n

    123456789101112\u5341\u4e0914151617<\/td># Import NumPy<\/code>import<\/code> numpy as np<\/code> # Tensor to Array<\/code>f_numpy <\/code>=<\/code> f.numpy()<\/code>print<\/code>(f_numpy)<\/code> # array([[1., 2.],<\/code>#       [3., 4.]], dtype=float32)<\/code> # Array to Tensor<\/code>h <\/code>=<\/code> np.array([[<\/code>8<\/code>,<\/code>7<\/code>,<\/code>6<\/code>,<\/code>5<\/code>],[<\/code>4<\/code>,<\/code>3<\/code>,<\/code>2<\/code>,<\/code>1<\/code>]])<\/code>h_tensor <\/code>=<\/code> torch.from_numpy(h)<\/code>print<\/code>(h_tensor)<\/code> # tensor([[8, 7, 6, 5],<\/code>#        [4, 3, 2, 1]])<\/code><\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

    \u5f20\u91cf\u7684\u7b97\u672f\u8fd0\u7b97<\/h3>\n\n\n\n

    \u73b0\u5728\u5230\u4e86\u4e0b\u4e00\u6b65\u7684\u65f6\u5019\u4e86\u3002\u8ba9\u6211\u4eec\u770b\u770b\u5982\u4f55\u5bf9 PyTorch \u5f20\u91cf\u6267\u884c\u7b97\u672f\u8fd0\u7b97\u3002<\/p>\n\n\n\n

    123456789101112\u5341\u4e091415161718192021222324\u4e8c\u5341\u4e94\u4e8c\u5341\u516d\u4e8c\u5341\u4e03\u4e8c\u5341\u516b\u4e8c\u5341\u4e5d\u4e09\u534131\u4e09\u5341\u4e8c33\u4e09\u5341\u56db\u4e09\u5341\u4e94\u4e09\u5341\u516d\u4e09\u5341\u4e03\u4e09\u5341\u516b\u4e09\u5341\u4e5d4041\u56db\u5341\u4e8c43\u56db\u5341\u56db\u56db\u5341\u4e94\u56db\u5341\u516d\u56db\u5341\u4e03\u56db\u5341\u516b49<\/td># Create tensor<\/code>tensor1 <\/code>=<\/code> torch.tensor([[<\/code>1<\/code>,<\/code>2<\/code>,<\/code>3<\/code>],[<\/code>4<\/code>,<\/code>5<\/code>,<\/code>6<\/code>]])<\/code>tensor2 <\/code>=<\/code> torch.tensor([[<\/code>-<\/code>1<\/code>,<\/code>2<\/code>,<\/code>-<\/code>3<\/code>],[<\/code>4<\/code>,<\/code>-<\/code>5<\/code>,<\/code>6<\/code>]])<\/code> # Addition<\/code>print<\/code>(tensor1<\/code>+<\/code>tensor2)<\/code># We can also use<\/code>print<\/code>(torch.add(tensor1,tensor2))<\/code> # tensor([[ 0,  4,  0],<\/code>#        [ 8,  0, 12]])<\/code> # Subtraction<\/code>print<\/code>(tensor1<\/code>-<\/code>tensor2)<\/code># We can also use<\/code>print<\/code>(torch.sub(tensor1,tensor2))<\/code> # tensor([[ 2,  0,  6],<\/code>#        [ 0, 10,  0]])<\/code> # Multiplication<\/code># Tensor with Scalar<\/code>print<\/code>(tensor1 <\/code>*<\/code> 2<\/code>)<\/code># tensor([[ 2,  4,  6],<\/code>#        [ 8, 10, 12]])<\/code> # Tensor with another tensor<\/code># Elementwise Multiplication<\/code>print<\/code>(tensor1 <\/code>*<\/code> tensor2)<\/code># tensor([[ -1,   4,  -9],<\/code>#        [ 16, -25,  36]])<\/code> # Matrix multiplication<\/code>tensor3 <\/code>=<\/code> torch.tensor([[<\/code>1<\/code>,<\/code>2<\/code>],[<\/code>3<\/code>,<\/code>4<\/code>],[<\/code>5<\/code>,<\/code>6<\/code>]])<\/code>print<\/code>(torch.mm(tensor1,tensor3))<\/code># tensor([[22, 28],<\/code>#        [49, 64]])<\/code> # Division<\/code># Tensor with scalar<\/code>print<\/code>(tensor1<\/code>\/<\/code>2<\/code>)<\/code># tensor([[0, 1, 1],<\/code>#        [2, 2, 3]])<\/code> # Tensor with another tensor<\/code># Elementwise division<\/code>print<\/code>(tensor1<\/code>\/<\/code>tensor2)<\/code># tensor([[-1,  1, -1],<\/code>#        [ 1, -1,  1]])<\/code><\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

    CPU \u4e0e GPU \u5f20\u91cf<\/h3>\n\n\n\n

    PyTorch \u9488\u5bf9 CPU \u548c GPU \u63d0\u4f9b\u4e86\u4e0d\u540c\u7684 Tensor \u5b9e\u73b0\u3002\u6bcf\u4e2a\u5f20\u91cf\u90fd\u53ef\u4ee5\u8f6c\u6362\u4e3a GPU\uff0c\u4ee5\u4fbf\u6267\u884c\u5927\u89c4\u6a21\u5e76\u884c\u3001\u5feb\u901f\u8ba1\u7b97\u3002\u5bf9\u5f20\u91cf\u6267\u884c\u7684\u6240\u6709\u64cd\u4f5c\u90fd\u5c06\u4f7f\u7528 PyTorch \u9644\u5e26\u7684 GPU \u7279\u5b9a\u4f8b\u7a0b\u6267\u884c\u3002<\/p>\n\n\n\n

    \u5982\u679c\u60a8\u65e0\u6cd5\u4f7f\u7528 GPU\uff0c\u5219\u53ef\u4ee5\u5728 Google Colab \u4e0a\u6267\u884c\u8fd9\u4e9b\u793a\u4f8b\u3002\u9009\u62e9 GPU \u4f5c\u4e3a\u8fd0\u884c\u65f6\u3002<\/p>\n\n\n\n

    \u8ba9\u6211\u4eec\u9996\u5148\u770b\u770b\u5982\u4f55\u4e3a GPU \u521b\u5efa\u5f20\u91cf\u3002<\/p>\n\n\n\n

    1234567<\/td># Create a tensor for CPU<\/code># This will occupy CPU RAM<\/code>tensor_cpu <\/code>=<\/code> torch.tensor([[<\/code>1.0<\/code>, <\/code>2.0<\/code>], [<\/code>3.0<\/code>, <\/code>4.0<\/code>], [<\/code>5.0<\/code>, <\/code>6.0<\/code>]], device<\/code>=<\/code>'cpu'<\/code>)<\/code> # Create a tensor for GPU<\/code># This will occupy GPU RAM<\/code>tensor_gpu <\/code>=<\/code> torch.tensor([[<\/code>1.0<\/code>, <\/code>2.0<\/code>], [<\/code>3.0<\/code>, <\/code>4.0<\/code>], [<\/code>5.0<\/code>, <\/code>6.0<\/code>]], device<\/code>=<\/code>'cuda'<\/code>)<\/code><\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

    \u5982\u679c\u60a8\u6b63\u5728\u4f7f\u7528 Google Colab\uff0c\u8bf7\u5173\u6ce8\u53f3\u4e0a\u89d2\u7684 RAM \u6d88\u8017\u8ba1\uff0c\u60a8\u4f1a\u770b\u5230\u5728\u521b\u5efatensor_gpu\u540e GPU RAM \u6d88\u8017\u5c31\u4f1a\u589e\u52a0\u3002<\/p>\n\n\n\n

    \u4e0e\u5f20\u91cf\u521b\u5efa\u4e00\u6837\uff0c\u5bf9 CPU \u548c GPU \u5f20\u91cf\u6267\u884c\u7684\u64cd\u4f5c\u4e5f\u4e0d\u540c\uff0c\u5e76\u4e14\u4f1a\u6d88\u8017\u4e0e\u6307\u5b9a\u8bbe\u5907\u76f8\u5bf9\u5e94\u7684 RAM\u3002<\/p>\n\n\n\n

    123456<\/td># This uses CPU RAM<\/code>tensor_cpu <\/code>=<\/code> tensor_cpu <\/code>*<\/code> 5<\/code> # This uses GPU RAM<\/code># Focus on GPU RAM Consumption<\/code>tensor_gpu <\/code>=<\/code> tensor_gpu <\/code>*<\/code> 5<\/code><\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

    \u8fd9\u91cc\u8981\u6ce8\u610f\u7684\u5173\u952e\u70b9\u662f\uff0c\u5728 GPU \u5f20\u91cf\u64cd\u4f5c\u4e2d\u6ca1\u6709\u4fe1\u606f\u6d41\u5411 CPU\uff08\u9664\u975e\u6211\u4eec\u6253\u5370\u6216\u8bbf\u95ee\u5f20\u91cf\uff09\u3002<\/p>\n\n\n\n

    \u6211\u4eec\u53ef\u4ee5\u5c06 GPU \u5f20\u91cf\u79fb\u52a8\u5230 CPU \u6216\u53cd\u4e4b\u4ea6\u7136\uff0c\u5982\u4e0b\u6240\u793a\u3002<\/p>\n\n\n\n

    12345<\/td># Move GPU tensor to CPU<\/code>tensor_gpu_cpu <\/code>=<\/code> tensor_gpu.to(device<\/code>=<\/code>'cpu'<\/code>)<\/code> # Move CPU tensor to GPU<\/code>tensor_cpu_gpu <\/code>=<\/code> tensor_cpu.to(device<\/code>=<\/code>'cuda'<\/code>)<\/code><\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n
    \"\u56fe\u7247\"<\/figure>\n\n\n\n

    \u5c31\u8fd9\u4e9b\u4e86\uff0c\u670b\u53cb\u4eec\uff01<\/p>\n\n\n\n

    \u7b80\u5355\u56de\u987e\u4e00\u4e0b\uff0c\u5728\u8fd9\u7bc7\u6587\u7ae0\u4e2d\uff0c\u6211\u4eec\u8ba8\u8bba\u4e86 PyTorch\u3001\u5b83\u7684\u72ec\u7279\u4e4b\u5904\u4ee5\u53ca\u4e3a\u4ec0\u4e48\u8981\u5b66\u4e60\u5b83\u3002\u6211\u4eec\u8fd8\u6df1\u5165\u8ba8\u8bba\u4e86 PyTorch \u5de5\u4f5c\u6d41\u7a0b\u548c PyTorch Tensor \u6570\u636e\u7c7b\u578b\u3002<\/p>\n\n\n\n

    \u6e90\u4ee3\u7801\uff1a<\/p>\n\n\n\n

    http:\/\/www.gitpp.com\/datasets\/learnopencv-cn\/blob\/master\/PyTorch-for-Beginners\/PyTorch_for_Beginners.ipynb<\/p>\n","protected":false},"excerpt":{"rendered":"

    \u4e16\u754c\u5728\u4e0d\u65ad\u53d1\u5c55\uff0c\u670d\u52a1\u4e8e\u4e16\u754c\u7684\u6280\u672f\u4e5f\u5728\u4e0d\u65ad\u53d1\u5c55\u3002\u6bcf\u4e2a\u4eba\u90fd\u5fc5\u987b\u8ddf\u4e0a\u6280\u672f\u7684\u5feb\u901f\u53d8\u5316\u3002\u4eba\u5de5\u667a\u80fd\u662f\u53d1\u5c55\u901f\u5ea6\u6700\u5feb\u3001\u89c4\u6a21\u6700\u5927 […]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[26],"tags":[],"class_list":["post-795","post","type-post","status-publish","format-standard","hentry","category-pytorch"],"blocksy_meta":"","_links":{"self":[{"href":"http:\/\/ai.gitpp.com\/index.php\/wp-json\/wp\/v2\/posts\/795","targetHints":{"allow":["GET"]}}],"collection":[{"href":"http:\/\/ai.gitpp.com\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/ai.gitpp.com\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"http:\/\/ai.gitpp.com\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"http:\/\/ai.gitpp.com\/index.php\/wp-json\/wp\/v2\/comments?post=795"}],"version-history":[{"count":1,"href":"http:\/\/ai.gitpp.com\/index.php\/wp-json\/wp\/v2\/posts\/795\/revisions"}],"predecessor-version":[{"id":796,"href":"http:\/\/ai.gitpp.com\/index.php\/wp-json\/wp\/v2\/posts\/795\/revisions\/796"}],"wp:attachment":[{"href":"http:\/\/ai.gitpp.com\/index.php\/wp-json\/wp\/v2\/media?parent=795"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/ai.gitpp.com\/index.php\/wp-json\/wp\/v2\/categories?post=795"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/ai.gitpp.com\/index.php\/wp-json\/wp\/v2\/tags?post=795"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}