{"id":63015,"date":"2024-03-28T12:48:59","date_gmt":"2024-03-28T04:48:59","guid":{"rendered":"https:\/\/www.ioiotimes.com\/?p=63015"},"modified":"2024-03-28T12:49:00","modified_gmt":"2024-03-28T04:49:00","slug":"nvidia-hopper-%e5%9c%a8mlperf%e7%9a%84%e7%94%9f%e6%88%90%e5%bc%8f%e4%ba%ba%e5%b7%a5%e6%99%ba%e6%85%a7%e9%a0%98%e5%9f%9f%e5%8f%96%e5%be%97%e9%a3%9b%e8%ba%8d%e6%80%a7%e9%80%b2%e5%b1%95","status":"publish","type":"post","link":"https:\/\/www.ioiotimes.com\/?p=63015","title":{"rendered":"NVIDIA Hopper \u5728MLPerf\u7684\u751f\u6210\u5f0f\u4eba\u5de5\u667a\u6167\u9818\u57df\u53d6\u5f97\u98db\u8e8d\u6027\u9032\u5c55"},"content":{"rendered":"\n

\u696d\u754c\u6a19\u6e96\u6e2c\u8a66\u8868\u660e\uff0c\u57fa\u65bcNVIDIA Hopper\u7684\u7cfb\u7d71\u904b\u884cTensorRT-LLM\u8edf\u9ad4\uff0c\u70ba\u751f\u6210\u5f0fAI\u63d0\u4f9b\u4e86\u4e16\u754c\u4e0a\u6700\u5f37\u5927\u7684\u5e73\u53f0<\/h3>\n\n\n\n
\"\"<\/figure>\n\n\n\n

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

NVIDIA \u6b63\u5f0f\u5ba3\u5e03\u5728\u696d\u754c\u6a19\u6e96\u6e2c\u8a66\u4e2d\u63d0\u4f9b\u4e86\u4e16\u754c\u4e0a\u6700\u5feb\u7684\u751f\u6210\u5f0f\u4eba\u5de5\u667a\u6167\uff08AI\uff09<\/a>\u63a8\u8ad6\u5e73\u53f0\u3002<\/p>\n\n\n\n

\u5728\u6700\u65b0\u7684 MLPerf \u57fa\u6e96\u6e2c\u8a66\u4e2d\uff0cNVIDIA TensorRT-LLM<\/a> \u9019\u500b\u53ef\u52a0\u901f\u548c\u7c21\u5316\u5927\u578b\u8a9e\u8a00\u6a21\u578b<\/a>\u7684\u8907\u96dc\u63a8\u8ad6\u5de5\u4f5c\u7684\u8edf\u9ad4\u5c07 GPT-J LLM \u4e0a\u7684 NVIDIA Hopper \u67b6\u69cb GPU \u6548\u80fd\u8f03\u516d\u500b\u6708\u524d\u63d0\u9ad8\u4e86\u8fd1 3 \u500d\u3002<\/p>\n\n\n\n

\u901f\u5ea6\u7684\u5927\u5e45\u63d0\u5347\u5c55\u793a\u4e86 NVIDIA \u7684\u6676\u7247\u3001\u7cfb\u7d71\u548c\u8edf\u9ad4\u5168\u7aef\u5e73\u53f0\u5728\u6eff\u8db3\u904b\u884c\u751f\u6210\u5f0fAI\u56b4\u82db\u8981\u6c42\u65b9\u9762\u7684\u5f37\u5927\u80fd\u529b\u3002<\/p>\n\n\n\n

\u8af8\u591a\u9818\u5148\u7684\u516c\u53f8\u6b63\u5728\u4f7f\u7528<\/a> TensorRT-LLM \u6700\u4f73\u5316\u4ed6\u5011\u7684\u6a21\u578b\u3002\u800c NVIDIA NIM <\/a>\u662f\u4e00\u5957\u63a8\u8ad6\u5fae\u670d\u52d9\uff0c\u5176\u4e2d\u5305\u542b TensorRT-LLM \u7b49\u63a8\u8ad6\u5f15\u64ce\uff0c\u8b93\u4f01\u696d\u6bd4\u4ee5\u5f80\u80fd\u66f4\u8f15\u9b06\u5730\u90e8\u7f72 NVIDIA \u63a8\u8ad6\u5e73\u53f0\u3002<\/p>\n\n\n

\n
\"\"<\/figure><\/div>\n\n\n

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

\u63d0\u9ad8\u751f\u6210\u5f0f<\/strong>AI<\/strong>\u7684\u6a19\u6e96<\/strong><\/strong><\/p>\n\n\n\n

\u5728 NVIDIA H200 Tensor\u6838\u5fc3GPU<\/a>\uff08\u6700\u65b0\u7684\u8a18\u61b6\u9ad4\u589e\u5f37\u578bHopper GPU\uff09\u4e0a\u904b\u884c\u7684 TensorRT-LLM\uff0c\u5728 MLPerf \u8fc4\u4eca\u70ba\u6b62\u6700\u5927\u898f\u6a21\u7684\u751f\u6210\u5f0f AI \u6e2c\u8a66\u4e2d\u63d0\u4f9b\u4e86\u6700\u5feb\u7684\u904b\u884c\u63a8\u8ad6\u6548\u80fd\u3002<\/p>\n\n\n\n

\u65b0\u7684\u57fa\u6e96\u6e2c\u8a66\u4f7f\u7528 Llama 2 \u7684\u6700\u5927\u7248\u672c\uff0cLlama 2 \u662f\u6700\u5148\u9032\u7684\u5927\u578b\u8a9e\u8a00\u6a21\u578b\uff0c\u5305\u542b 700 \u5104\u500b\u53c3\u6578\u3002\u8a72\u6a21\u578b\u6bd4 9 \u6708\u57fa\u6e96\u6e2c\u8a66<\/a>\u4e2d\u9996\u6b21\u4f7f\u7528\u7684 GPT-J \u5927\u578b\u8a9e\u8a00\u6a21\u578b\u5927 10 \u500d\u4ee5\u4e0a\u3002<\/p>\n\n\n\n

\u8a18\u61b6\u9ad4\u589e\u5f37\u578b H200 GPU \u5728 MLPerf \u9996\u6b21\u4eae\u76f8\u6642\uff0c\u4f7f\u7528 TensorRT-LLM \u6bcf\u79d2\u7522\u751f\u9ad8\u9054 31,000 \u500b\u8a5e\u5143\uff0c\u5275\u4e0b\u4e86 MLPerf \u7684 Llama 2 \u57fa\u6e96\u6e2c\u8a66\u7684\u7d00\u9304\u3002<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

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

H200 GPU \u7684\u7d50\u679c\u5305\u62ec\u5ba2\u88fd\u5316\u6563\u71b1\u89e3\u6c7a\u65b9\u6848\u5e36\u4f86\u7684\u9ad8\u9054 14% \u7684\u589e\u76ca\u3002\u9019\u662f\u6a19\u6e96\u7a7a\u6c23\u51b7\u537b\u4ee5\u5916\u7684\u5275\u65b0\u7bc4\u4f8b\u4e4b\u4e00\uff0c\u7cfb\u7d71\u88fd\u9020\u5546\u6b63\u5728\u5c07\u5176\u61c9\u7528\u5230 NVIDIA MGX<\/a>\u8a2d\u8a08\u4e2d\uff0c\u4ee5\u5c07 Hopper GPU \u7684\u6548\u80fd\u63d0\u5347\u5230\u65b0\u7684\u9ad8\u5ea6\u3002<\/p>\n\n\n\n

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

NVIDIA Hopper GPU <\/strong>\u7684\u8a18\u61b6\u9ad4\u63d0\u5347<\/strong><\/strong><\/p>\n\n\n\n

NVIDIA \u73fe\u5728\u5df2\u63d0\u4f9b H200 GPU \u4f9b\u5ba2\u6236\u6e2c\u8a66\uff0c\u4e26\u5c07\u65bc\u7b2c\u4e8c\u5b63\u51fa\u8ca8\u3002H200 GPU \u5f88\u5feb\u5c07\u7531\u8fd1 20 \u5bb6\u9818\u5148\u7684\u7cfb\u7d71\u88fd\u9020\u5546\u548c\u96f2\u7aef\u670d\u52d9\u4f9b\u61c9\u5546\u4f86\u63d0\u4f9b\u3002<\/p>\n\n\n\n

H200 GPU \u5305\u542b 141GB \u9ad8\u983b\u5bec\u8a18\u61b6\u9ad4 HBM3e\uff0c\u904b\u8f49\u901f\u5ea6\u70ba 4.8TB\/s\u3002\u8207 H100 GPU \u76f8\u6bd4\uff0c\u8a18\u61b6\u9ad4\u589e\u52a0\u4e86 76%\uff0c\u904b\u884c\u901f\u5ea6\u63d0\u9ad8\u4e86 43%\u3002\u9019\u4e9b\u52a0\u901f\u5668\u53ef\u63d2\u5165\u8207 H100 GPU \u76f8\u540c\u7684\u4e3b\u6a5f\u677f\u548c\u7cfb\u7d71\uff0c\u4e26\u4f7f\u7528\u76f8\u540c\u7684\u8edf\u9ad4\u3002<\/p>\n\n\n\n

\u501f\u52a9 HBM3e \u8a18\u61b6\u9ad4\uff0c\u55ae\u500b H200 GPU \u80fd\u4ee5\u6700\u9ad8\u541e\u5410\u91cf\u904b\u884c\u6574\u500b Llama 2 70B \u6a21\u578b\uff0c\u5f9e\u800c\u7c21\u5316\u4e26\u52a0\u901f\u63a8\u8ad6\u3002<\/p>\n\n\n\n

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

GH200\u914d\u5099\u66f4\u591a\u7684\u8a18\u61b6\u9ad4<\/strong><\/p>\n\n\n\n

NVIDIA GH200 \u8d85\u7d1a\u6676\u7247<\/a>\u4e2d\u914d\u5099\u66f4\u591a\u8a18\u61b6\u9ad4\uff0c\u6700\u9ad8\u53ef\u9054 624GB \u9ad8\u901f\u8a18\u61b6\u9ad4\uff0c\u5176\u4e2d\u5305\u542b 144GB \u7684 HBM3e \u8a18\u61b6\u9ad4\uff0c\u6b64\u8d85\u7d1a\u6676\u7247\u5c07 Hopper \u67b6\u69cb GPU \u548c\u7bc0\u80fd\u7684 NVIDIA Grace CPU<\/a> \u7d50\u5408\u5728\u4e00\u500b\u6a21\u7d44\u4e0a\u3002NVIDIA \u52a0\u901f\u5668\u662f\u9996\u6279\u4f7f\u7528 HBM3e \u8a18\u61b6\u9ad4\u6280\u8853\u7684\u52a0\u901f\u5668\u3002<\/p>\n\n\n\n

\u6191\u85c9\u5c07\u8fd1 5 TB\/s \u7684\u8a18\u61b6\u9ad4\u983b\u5bec\uff0cGH200 \u8d85\u7d1a\u6676\u7247\u5728\u5982\u63a8\u85a6\u7cfb\u7d71<\/a>\u7b49\u8a18\u61b6\u9ad4\u5bc6\u96c6\u578b\u7684 MLPerf \u6e2c\u8a66\u4e2d\u63d0\u4f9b\u4e86\u51fa\u8272\u7684\u6548\u80fd\u3002<\/p>\n\n\n\n

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

\u6a6b\u6383\u6bcf\u4e00\u500b<\/strong> MLPerf <\/strong>\u6e2c\u8a66<\/strong><\/strong><\/p>\n\n\n\n

\u4ee5\u6bcf\u500b\u52a0\u901f\u5668\u70ba\u57fa\u790e\uff0cHopper GPU \u5728\u6700\u65b0\u4e00\u8f2a MLPerf \u7522\u696d\u57fa\u6e96\u6e2c\u8a66\u4e2d\uff0c\u6a6b\u6383\u4e86\u6240\u6709 AI \u63a8\u8ad6\u6e2c\u8a66\u3002<\/p>\n\n\n\n

\u9019\u4e9b\u57fa\u6e96\u6e2c\u8a66\u6db5\u84cb\u7576\u4eca\u6700\u53d7\u6b61\u8fce\u7684 AI \u5de5\u4f5c\u8ca0\u8f09\u548c\u5834\u666f\uff0c\u5305\u62ec\u751f\u6210\u5f0f AI\u3001\u63a8\u85a6\u7cfb\u7d71\u3001\u81ea\u7136\u8a9e\u8a00\u8655\u7406\u3001\u8a9e\u97f3\u548c\u96fb\u8166\u8996\u89ba\u3002NVIDIA \u662f\u552f\u4e00\u4e00\u5bb6\u5728\u6700\u65b0\u4e00\u8f2a\u4ee5\u53ca\u81ea 2020 \u5e74 10 \u6708\u958b\u59cb MLPerf \u8cc7\u6599\u4e2d\u5fc3\u63a8\u8ad6\u57fa\u6e96\u6e2c\u8a66\u4ee5\u4f86\uff0c\u6bcf\u4e00\u8f2a\u90fd\u63d0\u4ea4\u6240\u6709\u5de5\u4f5c\u8ca0\u8f09\u7d50\u679c\u7684\u516c\u53f8\u3002<\/p>\n\n\n\n

\u6301\u7e8c\u7684\u6548\u80fd\u63d0\u5347\u610f\u5473\u8457\u63a8\u8ad6\u6210\u672c\u7684\u964d\u4f4e\uff0c\u5c0d\u65bc\u5168\u7403\u90e8\u7f72\u7684\u6578\u767e\u842c\u500b NVIDIA GPU \u4f86\u8aaa\uff0c\u63a8\u8ad6\u5df2\u6210\u70ba\u65e5\u5e38\u5de5\u4f5c\u4e2d\u7684\u4e00\u5927\u90e8\u5206\uff0c\u800c\u4e14\u9084\u5728\u4e0d\u65b7\u589e\u9577\u3002<\/p>\n\n\n\n

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

\u63a8\u9032\u4e00\u5207\u53ef\u80fd<\/strong><\/strong><\/p>\n\n\n\n

NVIDIA \u5728\u57fa\u6e96\u6e2c\u8a66\u4e2d\u4e00\u500b\u540d\u70ba\u300c\u958b\u653e\u7d44\u300d\u7684\u7279\u5225\u90e8\u5206\u4e2d\u5c55\u793a\u4e86\u4e09\u7a2e\u5275\u65b0\u6280\u8853\uff0c\u9019\u90e8\u5206\u662f\u70ba\u4e86\u6e2c\u8a66\u5148\u9032\u7684AI\u65b9\u6cd5\u800c\u5275\u5efa\u3002<\/p>\n\n\n\n

NVIDIA \u5de5\u7a0b\u5e2b\u4f7f\u7528\u4e86\u4e00\u7a2e\u7a31\u70ba\u7d50\u69cb\u5316\u7a00\u758f\u6027\uff08structured sparsity\uff09<\/a>\u7684\u6280\u8853\uff0c\u4f7f Llama 2 \u7684\u63a8\u8ad6\u901f\u5ea6\u63d0\u9ad8\u4e86 33%\u3002\u7d50\u69cb\u5316\u7a00\u758f\u6027<\/a>\u662f\u4e00\u7a2e\u6e1b\u5c11\u8a08\u7b97\u7684\u65b9\u6cd5\uff0c\u9996\u6b21\u5728 NVIDIA A100 Tensor\u6838\u5fc3GPU<\/a> \u4e2d\u5f15\u5165\u3002<\/p>\n\n\n\n

\u7b2c\u4e8c\u500b\u958b\u653e\u7d44\u6e2c\u8a66\u767c\u73fe\uff0c\u4f7f\u7528\u526a\u679d\u6280\u8853\uff08pruning\uff09\u53ef\u4ee5\u5c07\u63a8\u8ad6\u901f\u5ea6\u63d0\u9ad8\u9ad8\u9054 40%\uff0c\u9019\u662f\u7c21\u5316 AI \u6a21\u578b\uff08\u6b64\u4f8b\u70ba\u5927\u578b\u8a9e\u8a00\u6a21\u578b\uff09\u4ee5\u589e\u52a0\u63a8\u8ad6\u541e\u5410\u91cf\u7684\u4e00\u7a2e\u65b9\u5f0f\u3002<\/p>\n\n\n\n

\u6700\u5f8c\uff0c\u4e00\u7a2e\u540d\u70ba DeepCache \u7684\u6700\u4f73\u5316\u65b9\u6cd5\u6e1b\u5c11\u4e86\u5c0d Stable Diffusion XL \u6a21\u578b\u63a8\u8ad6\u6240\u9700\u7684\u6578\u5b78\u904b\u7b97\uff0c\u5c07\u6548\u80fd\u63d0\u5347\u4e86\u9a5a\u4eba\u7684 74%\u3002<\/p>\n\n\n\n

\u6240\u6709\u9019\u4e9b\u7d50\u679c\u90fd\u662f\u5728 NVIDIA H100 Tensor\u6838\u5fc3GPU <\/a>\u4e0a\u904b\u884c\u7684\u3002<\/p>\n\n\n\n

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

\u4f7f\u7528\u8005\u503c\u5f97\u4fe1\u8cf4\u7684\u4f86\u6e90<\/strong><\/strong><\/p>\n\n\n\n

MLPerf \u7684\u6e2c\u8a66\u900f\u660e\u4e14\u5ba2\u89c0\uff0c\u56e0\u6b64\u4f7f\u7528\u8005\u53ef\u4ee5\u4f9d\u9760\u7d50\u679c\u505a\u51fa\u660e\u667a\u7684\u8cfc\u8cb7\u6c7a\u5b9a\u3002<\/p>\n\n\n\n

NVIDIA \u7684\u5408\u4f5c\u5925\u4f34\u53c3\u8207 MLPerf \u662f\u56e0\u70ba\u4ed6\u5011\u77e5\u9053\u9019\u5c0d\u5ba2\u6236\u8a55\u4f30 AI \u7cfb\u7d71\u548c\u670d\u52d9\u4f86\u8aaa\u662f\u4e00\u500b\u5f88\u6709\u50f9\u503c\u7684\u5de5\u5177\u3002<\/p>\n\n\n\n

\u672c\u8f2a\u5728 NVIDIA AI \u5e73\u53f0\u4e0a\u63d0\u4ea4\u7d50\u679c\u7684\u5408\u4f5c\u5925\u4f34\u5305\u62ec\u83ef\u78a9\u96fb\u8166\u3001\u601d\u79d1\u3001\u6234\u723e\u79d1\u6280\u96c6\u5718\u3001\u5bcc\u58eb\u901a\u3001\u6280\u5609\u79d1\u6280\u3001Google\u3001\u6167\u8207\u79d1\u6280\u3001\u806f\u60f3\u3001Microsoft Azure\u3001\u7532\u9aa8\u6587\u3001\u96f2\u9054\u79d1\u6280\u3001\u7f8e\u8d85\u5fae\u3001VMware\uff08\u6700\u8fd1\u7531\u535a\u901a\u6536\u8cfc\uff09\u548c\u7def\u7a4e\u79d1\u6280\u3002<\/p>\n\n\n\n

NVIDIA \u5728\u672c\u6b21\u6e2c\u8a66\u4e2d\u4f7f\u7528\u7684\u6240\u6709\u8edf\u9ad4\u90fd\u53ef\u4ee5\u5f9e MLPerf \u8cc7\u6e90\u5eab\u4e2d\u53d6\u5f97\uff0cNVIDIA \u4e0d\u65b7\u5c07\u8edf\u9ad4\u6700\u4f73\u5316\u7d50\u679c\u653e\u5165 NVIDIA \u7684 GPU \u61c9\u7528\u8edf\u9ad4\u4e2d\u5fc3 NGC<\/a> \u4ee5\u53ca NVIDIA AI Enterprise<\/a>\u7684\u5bb9\u5668\u4e2d\u3002NVIDIA AI Enterprise \u70ba\u4e00\u500b\u5b89\u5168\u3001\u53d7\u652f\u63f4\u7684\u5e73\u53f0\uff0c\u5176\u4e2d\u5305\u542b NIM \u63a8\u8ad6\u5fae\u670d\u52d9\u3002<\/p>\n\n\n\n

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

\u4e0b\u4e00\u4ef6\u5927\u4e8b<\/strong><\/strong><\/p>\n\n\n\n

\u751f\u6210\u5f0f AI \u7684\u7528\u4f8b\u3001\u6a21\u578b\u5927\u5c0f\u548c\u8cc7\u6599\u96c6\u4e0d\u65b7\u64f4\u5927\u3002\u9019\u5c31\u662f MLPerf \u4e0d\u65b7\u767c\u5c55\u7684\u539f\u56e0\uff0c\u589e\u52a0\u4e86 Llama 2 70B \u548c Stable Diffusion XL \u7b49\u4e3b\u6d41\u6a21\u578b\u7684\u771f\u5be6\u6e2c\u8a66\u3002<\/p>\n\n\n\n

\u70ba\u4e86\u8ddf\u4e0a\u5927\u578b\u8a9e\u8a00\u6a21\u578b\u898f\u6a21\u7684\u7206\u70b8\u6027\u589e\u9577\uff0cNVIDIA \u5275\u8fa6\u4eba\u66a8\u57f7\u884c\u9577\u9ec3\u4ec1\u52f3\u4e0a\u9031\u5728 GTC \u4e0a\u5ba3\u5e03\uff0cNVIDIA Blackwell \u67b6\u69cb GPU<\/a> \u5c07\u63d0\u4f9b\u5146\u7d1a\u53c3\u6578 AI \u6a21\u578b\u6240\u9700\u7684\u65b0\u6548\u80fd\u6c34\u5e73\u3002<\/p>\n\n\n\n

\u5927\u578b\u8a9e\u8a00\u6a21\u578b\u7684\u63a8\u8ad6\u975e\u5e38\u56f0\u96e3\uff0c\u9700\u8981\u5c08\u696d\u77e5\u8b58\u548c NVIDIA \u4f7f\u7528 Hopper \u67b6\u69cb GPU \u548c TensorRT-LLM \u5728 MLPerf \u4e0a\u5c55\u793a\u7684\u5168\u7aef\u67b6\u69cb\u3002\u672a\u4f86\u9084\u6703\u6709\u66f4\u591a\u3002<\/p>\n\n\n\n

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

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

\ud83d\udfe6\u73fe\u5728\u5c31\u52a0\u5165 ioioTIMES \u81c9\u66f8\u7c89\u7d72\u5718<\/a> \u66f4\u591a\u4e92\u52d5\u3001\u66f4\u591a\u597d\u5eb7\u650f\u62b5\u52a0!!<\/strong>
\ud83d\udfe6\u6211\u5011\u6709LINE TODAY<\/a>\u983b\u9053\u4e86\uff0c\u5feb\u4f86\u8ffd\u8e2a\u6211\u5011\u5427!!–\u6700\u65b0\u79d1\u6280\u65b0\u805e \u76e1\u5728\u4f60\u624b<\/strong><\/h4>\n","protected":false},"excerpt":{"rendered":"

\u696d\u754c\u6a19\u6e96\u6e2c\u8a66\u8868\u660e\uff0c\u57fa\u65bcNVIDIA <\/p>\n","protected":false},"author":3,"featured_media":63023,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"rank_math_lock_modified_date":false,"footnotes":""},"categories":[13],"tags":[580,1287,6360,71],"class_list":["post-63015","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-news","tag-focus","tag-hopper","tag-mlperf","tag-nvidia"],"_links":{"self":[{"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=\/wp\/v2\/posts\/63015"}],"collection":[{"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=\/wp\/v2\/users\/3"}],"replies":[{"embeddable":true,"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=63015"}],"version-history":[{"count":2,"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=\/wp\/v2\/posts\/63015\/revisions"}],"predecessor-version":[{"id":63024,"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=\/wp\/v2\/posts\/63015\/revisions\/63024"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=\/wp\/v2\/media\/63023"}],"wp:attachment":[{"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=63015"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=63015"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=63015"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}