{"id":14964,"date":"2020-02-26T12:31:41","date_gmt":"2020-02-26T04:31:41","guid":{"rendered":"http:\/\/www.zhushiyao.com\/?p=14964"},"modified":"2020-02-26T12:32:21","modified_gmt":"2020-02-26T04:32:21","slug":"03514310a6","status":"publish","type":"post","link":"http:\/\/www.zhushiyao.com\/?p=14964","title":{"rendered":"2019 \u5e74 10 \u5927\u9876\u7ea7 Python \u652f\u6301\u5e93"},"content":{"rendered":"
\n
\n
\n
\n
\u6765\u81ea\u516c\u4f17\u53f7\uff1a<\/span><\/strong>Python\u5927\u5496\u8c08<\/span><\/strong><\/span><\/section>\n

\u539f\u6587\u4f5c\u8005\uff1a<\/span>Fabi\u00e1n Torres<\/span><\/p>\n

\u539f\u6587\u94fe\u63a5\uff1a<\/span>https:\/\/tryolabs.com\/blog\/2019\/12\/10\/top-10-python-libraries-of-2019\/<\/span><\/p>\n<\/section>\n<\/section>\n<\/blockquote>\n

\n

\u6b22\u8fce\u56de\u6765\uff0c\u8fd9\u662f\u6211\u4eec\u7b2c\u4e94\u6b21\u63a8\u51fa\u5e74\u5ea6\u9876\u7ea7 Python \u652f\u6301\u5e93<\/strong><\/span>\u699c\u5355\u3002<\/span>\u5728\u8fd9\u91cc\uff0c\u6211\u4eec\u5217\u51fa\u4e86\u9690\u533f\u4e8e\u5f00\u6e90\u4e16\u754c\u91cc\u7684\u7470\u5b9d\uff0c\u65e2\u53ef\u4ee5\u5e2e\u4f60\u542f\u52a8\u65b0\u7684\u9879\u76ee\uff0c\u4e5f\u53ef\u4ee5\u4e3a\u4f60\u73b0\u6709\u7684\u9879\u76ee\u6dfb\u5149\u52a0\u5f69\u3002<\/span>\u672c\u699c\u5355\u8986\u76d6\u4e86\u5404\u4e2a\u9886\u57df\uff0c\u5305\u62ec\u673a\u5668\u5b66\u4e60\u4e0e\u975e\u673a\u5668\u5b66\u4e60\u7684\u652f\u6301\u5e93\u3002<\/span><\/p>\n

\u5e0c\u671b\u4f60\u80fd\u548c\u6211\u4eec\u4e00\u6837\u559c\u6b22\u672c\u5e74\u5ea6\u7684\u9876\u7ea7\u652f\u6301\u5e93\u699c\u5355\uff0c\u8bdd\u4e0d\u591a\u8bf4\uff0c\u9898\u5f52\u6b63\u6587\uff01<\/span><\/p>\n

1. HTTPX<\/span><\/h2>\n
\"2019<\/figure>\n

\u9879\u76ee\u7f51\u5740\uff1a<\/span>
\nhttps:\/\/www.encode.io\/httpx\/<\/span><\/p>\n

\u7ecf\u5e38\u8fdb\u884c API \u4ea4\u4e92\u7684 Python \u5fe0\u7c89\uff0c\u5bf9 requests \u5e94\u8be5\u662f\u8fd0\u7528\u81ea\u5982\u4e86\u3002<\/span>\u4e0d\u8fc7\uff0c\u5bf9\u4e8e\u65e5\u8d8b\u666e\u53ca\u7684\u73b0\u4ee3\u9ad8\u6027\u80fd\u5e94\u7528\u800c\u8a00\uff0c\u5728\u7528\u5230\u5f02\u6b65\u8303\u5f0f\u65f6\uff0crequests \u5c31\u663e\u5f97\u6709\u70b9\u6349\u895f\u89c1\u8098\u4e86\u3002<\/span><\/p>\n

\u4e3a\u4e86\u89e3\u51b3\u8fd9\u4e00\u95ee\u9898\uff0c\u5927\u5496 Tom Christie \u548c\u4ed6\u7684\u5c0f\u4f19\u4f34\u7ed9\u5927\u5bb6\u5e26\u6765\u4e86\u65b0\u4e16\u4ee3\u5f02\u6b65 HTTP \u5ba2\u6237\u7aef ~ HTTPX\u3002<\/span><\/p>\n

\u7d27\u8ddf requests \u6b65\u4f10\uff0cHTTPX \u5728\u652f\u6301 requests \u6807\u51c6\u529f\u80fd\u7684\u540c\u65f6\uff0c\u8fd8\u652f\u6301 HTTP\/2 \u4e0e HTTP\/1\uff0c\u4ee5\u53ca\u901a\u8fc7 ASGI \u534f\u8bae\u76f4\u63a5\u8c03\u7528 Python \u7f51\u7edc\u5e94\u7528\uff0c\u6b64\u5916\u8fd8\u6709\u5b8c\u5584\u7684\u7c7b\u578b\u6ce8\u91ca\u3002<\/span><\/p>\n

\u8981\u6267\u884c\u5927\u89c4\u6a21\u5e76\u53d1\u8bf7\u6c42\uff1f<\/span>HTTPX \u5c31\u662f\u4f60\u8981\u627e\u7684\u7b54\u6848\u3002<\/span><\/p>\n

\u6ce8\u610f\uff1a<\/span>HTTPX \u4ecd\u5728 alpha \u7248\u5f00\u53d1\u9636\u6bb5\uff0c\u76ee\u524d\u53ea\u5f00\u53d1\u4e86\u5f02\u6b65\u5ba2\u6237\u7aef\uff0c\u5c06\u6765\u8fd8\u4f1a\u5f15\u5165\u540c\u6b65\u5ba2\u6237\u7aef\u3002<\/span><\/p>\n

2. Starlette<\/span><\/h2>\n
\"2019<\/figure>\n

\u9879\u76ee\u7f51\u5740\uff1a<\/span>
\nhttps:\/\/www.starlette.io\/<\/span><\/p>\n

Starlette \u6253\u5305\u4e86 WebSocket\u3001GraphQL\u3001\u8fdb\u7a0b\u5185\u540e\u53f0\u4efb\u52a1\u7b49\u591a\u79cd\u529f\u80fd\uff0c\u662f\u771f\u6b63\u9ad8\u6027\u80fd\u7684\u8f7b\u91cf\u7ea7 ASGI \u6846\u67b6\u4e0e\u5de5\u5177\u7bb1\u3002<\/span>\u6240\u6709\u529f\u80fd\u90fd\u5b8c\u5168\u652f\u6301\u7c7b\u578b\u6ce8\u91ca\u4ee3\u7801\u5e93\uff0c\u4e14\u4e0d\u9700\u8981\u4efb\u4f55\u786c\u6027\u4f9d\u8d56\u9879\u3002<\/span>\u53ef\u4ee5\u628a\u5b83\u5f53\u6210\u975e\u5e38\u8f7b\u91cf\u7ea7\u7684\u73b0\u4ee3\u5f02\u6b65\u7248 Flask\u3002<\/span><\/p>\n

Starlette \u7684\u7075\u6d3b\u6027\u5f88\u5f3a\uff0c\u65e2\u53ef\u4ee5\u7528\u4f5c\u5b8c\u6574\u7684\u7f51\u7edc\u6846\u67b6\uff0c\u4e5f\u53ef\u4ee5\u53ea\u7528\u4f5c ASGI \u5de5\u5177\u7bb1\u3002<\/span><\/p>\n

Starlette \u8fd0\u884c\u5728 uvicorn \u4e4b\u4e0a\uff0c\u8fd9\u4e2a\u652f\u6301\u5e93\u5165\u9009\u4e86\u53bb\u5e74\u7684 10 \u5927\u9876\u7ea7 Python \u652f\u6301\u5e93\u699c\u5355\u3002<\/span><\/p>\n

\u5982\u679c\u60f3\u5f00\u53d1\u65b0\u578b\u7f51\u7edc\u5e94\u7528\uff0c\u5f3a\u70c8\u5efa\u8bae\u60a8\u7ed9 Starlette \u4e00\u4e2a\u53d1\u5149\u53d1\u70ed\u7684\u673a\u4f1a\u3002<\/span><\/p>\n

3. FastAPI<\/span><\/h2>\n
\"2019<\/figure>\n

\u9879\u76ee\u7f51\u5740\uff1a<\/span>
\nhttps:\/\/github.com\/tiangolo\/fastapi<\/span><\/p>\n

Starlette \u5f88\u8d5e\uff0c\u4f46\u5b83\u592a\u8ff7\u4f60\u4e86\uff0c\u867d\u7136\u7ed9\u4e86\u5f00\u53d1\u8005\u5f88\u5927\u7684\u81ea\u7531\uff0c\u4f46\u6709\u65f6\uff0c\u7528\u6237\u8981\u7684\u5176\u5b9e\u53ea\u662f\u4e00\u4e2a\u80fd\u5feb\u901f\u5b8c\u6210\u4efb\u52a1\u7684\u6846\u67b6\u3002<\/span><\/p>\n

Sebasti\u00e1n Ram\u00edrez \u5c31\u662f\u51fa\u4e8e\u8fd9\u4e2a\u76ee\u7684\u5f00\u53d1\u7684 FastAPI\uff0c\u600e\u4e48\u8bf4\u5462\uff1f<\/span>\u8fd9\u4e2a\u652f\u6301\u5e93\u7684\u5404\u4e2a\u65b9\u9762\u90fd\u5f88 Fast\u3002<\/span><\/p>\n

\u8fd9\u4e2a\u7528 Python \u5f00\u53d1 API \u7684\u65b0\u6846\u67b6\u5177\u6709\u8d85\u9ad8\u6027\u80fd\uff0c\u800c\u4e14\u53ef\u4ee5\u57fa\u4e8e OpenAPI \u6807\u51c6\u81ea\u52a8\u751f\u6210\u4ea4\u4e92\u5f0f\u6587\u6863\u3002<\/span>\u9ed8\u8ba4\u652f\u6301 Swagger UI \u4e0e ReDoc\uff0c\u5141\u8bb8\u76f4\u63a5\u4ece\u6d4f\u89c8\u5668\u8c03\u7528\u3001\u6d4b\u8bd5 API\uff0c\u4ece\u800c\u63d0\u9ad8\u5f00\u53d1\u6548\u7387\u3002<\/span>\u7528\u8fd9\u4e2a\u6846\u67b6\u5f00\u53d1 API\uff0c\u771f\u7684\u662f\u53c8\u5feb\u53c8\u7b80\u5355\u3002<\/span><\/p>\n

\u8be5\u652f\u6301\u5e93\u8fd8\u652f\u6301\u73b0\u4ee3 Python \u6700\u4f73\u5b9e\u7528\u529f\u80fd\u4e4b\u4e00\uff1a<\/span>\u7c7b\u578b\u63d0\u793a\u3002<\/span>FastAPI \u5728\u5f88\u591a\u65b9\u9762\u90fd\u4f7f\u7528\u4e86\u7c7b\u578b\u63d0\u793a\uff0c\u5176\u4e2d\u6700\u9177\u7684\u4e00\u4e2a\u529f\u80fd\u662f\u7531 Pydantic \u52a0\u6301\u7684\u81ea\u52a8\u6570\u636e\u9a8c\u8bc1\u4e0e\u8f6c\u6362\u3002<\/span><\/p>\n

FastAPI \u57fa\u4e8e Starlette \u5f00\u53d1\uff0c\u6027\u80fd\u4e0e NodeJS \u548c GO \u76f8\u5f53\uff0c\u8fd8\u81ea\u5e26 WebSocket \u4e0e GraphQL \u539f\u751f\u652f\u6301\u3002<\/span><\/p>\n

\u6700\u540e\uff0c\u5bf9\u4e8e\u5f00\u6e90\u652f\u6301\u5e93\u800c\u8a00\uff0c\u5b83\u7684\u6280\u672f\u652f\u6301\u6587\u6863\u4e5f\u975e\u5e38\u4e0d\u9519\u3002<\/span>\u8bf4\u771f\u7684\uff0c\u53bb\u770b\u4e00\u4e0b\u5427\uff01<\/span><\/p>\n

4. Immutables<\/span><\/h2>\n
\"2019<\/figure>\n

\u9879\u76ee\u7f51\u5740\uff1a<\/span>
\nhttps:\/\/github.com\/magicstack\/immutables<\/span><\/p>\n

MagicStack \u7684\u5144\u5f1f\u56de\u5f52\u4e86\uff0c\u8fd9\u6b21\u4ed6\u4eec\u5e26\u6765\u7684\u662f\u4e00\u4e2a\u7b80\u5355\u3001\u4f18\u96c5\u7684\u4e0d\u53ef\u53d8\u6620\u5c04\u7c7b\u578b\uff08\u201c\u51bb\u7ed3\u5b57\u5178\u201d\uff09\u3002<\/span><\/p>\n

\u8fd9\u4e1c\u897f\u7684\u597d\u5904\u5728\u54ea\u513f\uff1f<\/span>\u5443\uff0c\u5e0c\u671b\u4f60\u80fd\u542c\u5f97\u61c2\u4e0b\u9762\u8fd9\u53e5\u8bdd\uff0cHaskell \u7b49\u51fd\u6570\u7f16\u7a0b\u8bed\u8a00\u91cc\u4f7f\u7528\u7684\u54c8\u5e0c\u6570\u7ec4\u6620\u5c04\u5b57\u5178\u6811\uff08HAMT\uff09\u7684\u5e95\u5c42\u6570\u636e\u7ed3\u6784\u3002<\/span>\u6700\u503c\u5f97\u6ce8\u610f\u7684\u662f\uff0c\u4e0d\u7ba1 set() \u8fd8\u662f get() \u64cd\u4f5c\u7684\u6027\u80fd\u90fd\u4e3a \u00a0O(log N)\uff0c\u4e14\u5bf9\u4e8e\u76f8\u5bf9\u8f83\u5c0f\u7684\u6620\u5c04\uff0c\u53ef\u4ee5\u8fbe\u5230 O(1)\u3002<\/span><\/p>\n

\u5982\u679c\u4f60\u7684\u5e94\u7528\u91cc\u5b57\u5178\u7684\u89c4\u6a21\u8f83\u5927\uff0c\u5e76\u4e14\u60f3\u63d0\u9ad8\u4e00\u4e0b\u6027\u80fd\uff0c\u8fd9\u4e2a\u9177\u9177\u7684\u65b0\u652f\u6301\u5e93\u503c\u5f97\u4e00\u8bd5\u3002<\/span><\/p>\n

5. Pyodide<\/span><\/h2>\n

\u9879\u76ee\u7f51\u5740\uff1a<\/span>
\nhttps:\/\/github.com\/iodide-project\/pyodide\/<\/span><\/p>\n

Pyodide \u662f\u672c\u699c\u5355\u91cc\u6700\u5237\u4e09\u89c2\u7684\u9879\u76ee\u3002<\/span>\u901a\u8fc7 WebAssembly\uff0c\u5b83\u628a Python \u79d1\u5b66\u5f15\u5165\u5230\u4e86\u6d4f\u89c8\u5668\u754c\u9762\uff0c\u628a\u79d1\u5b66\u8ba1\u7b97\u63d0\u5347\u5230\u4e86\u5168\u65b0\u7684\u5c42\u9762\u3002<\/span><\/p>\n

\u60f3\u7528 Numpy \u641e\u5b9a\u4e00\u4e9b\u6570\u5b57\uff1f<\/span>\u7528 Pandas \u5904\u7406\u5927\u578b DataFrame\uff1f<\/span>\u7528 Matplotlib \u4e3a\u8ba1\u7b97\u7ed3\u679c\u8fdb\u884c\u53ef\u89c6\u5316\uff1f<\/span>\u6709\u4e86 Pyodide\uff0c\u8fd9\u4e9b\u64cd\u4f5c\u5728\u6d4f\u89c8\u5668\u91cc\u5c31\u53ef\u4ee5\u8f7b\u8f7b\u677e\u677e\u5730\u5b9e\u73b0\u4e86\u3002<\/span><\/p>\n

\u8fd9\u8fd8\u4e0d\u7b97\u5b8c\uff0c\u8fd8\u6709\u597d\u6d88\u606f\uff1a<\/span>\u8fd9\u4e2a\u5e93\u53ef\u7528\u7684 Python \u652f\u6301\u5e93\u5df2\u7ecf\u8d85\u8fc7\u4e86 35 \u4e2a\uff0c\u60a8\u8fd9\u662f\u8981\u4e0a\u5929\u554a\u3002<\/span><\/p>\n

6. Modin<\/span><\/h2>\n
\"2019<\/figure>\n

\u9879\u76ee\u7f51\u5740\uff1a<\/span>
\n<\/span>https:\/\/github.com\/modin-project\/modin<\/span><\/span><\/p>\n

Modin \u7684\u53e3\u53f7\u662f\u201c\u53ea\u6539\u4e00\u884c\u4ee3\u7801\uff0c\u5c31\u80fd\u52a0\u901f Pandas\u201d\uff0c\u6ca1\u9519\uff0c\u5c31\u662f\u8fd9\u4e48\u7b80\u5355\u3002<\/span>\u53ea\u8981\u5b89\u88c5 Modin\uff0c\u66f4\u6539 import \u8bed\u53e5\uff0c\u5728\u591a\u6838\u7b14\u8bb0\u672c\u4e0a\uff0c\u5904\u7406\u6570\u636e\u7684\u901f\u5ea6\u5c31\u80fd\u63d0\u9ad8 4 \u500d\u3002<\/span><\/p>\n

\u6211\u6ef4\u4e2a\u795e ~~ \u8fd9\u662f\u600e\u4e48\u505a\u5230\u7684\uff1f<\/span>Modin \u6709\u81ea\u5e26\u7684 <\/span>modin.pandas.DataFrame<\/span><\/code> \u5bf9\u8c61\uff0c\u8fd9\u662f\u4e00\u4e2a\u8f7b\u91cf\u7ea7\u7684\u5e76\u884c DataFrame \u5bf9\u8c61\u3002<\/span>\u56e0\u4e3a\u4e0e Pandas API \u517c\u5bb9\uff0c\u4f7f\u7528\u8fd9\u4e2a\u5bf9\u8c61\u5c31\u8ddf\u662f\u900f\u660e\u7684\u4e00\u6837\uff0c\u5728\u540e\u53f0\uff0c\u8be5\u652f\u6301\u5e93\u4f7f\u7528 Ray \u6216 Dask \u4f5c\u4e3a\u8ba1\u7b97\u5f15\u64ce\uff0c\u6267\u884c\u5206\u5e03\u5f0f\u6570\u636e\u8ba1\u7b97\u3002<\/span><\/p>\n

\u6709\u7684\u65f6\u5019\uff0c\u83b7\u5f97\u66f4\u5feb\u7684\u901f\u5ea6\u5c31\u662f\u90a3\u4e48\u7b80\u5355\uff0c\u6539\u4e00\u4e24\u884c\u4ee3\u7801\u5c31\u53ef\u4ee5\u4e86\uff0cModin \u5c31\u662f\u8bc1\u660e\u3002<\/span><\/p>\n

7. Streamlit<\/span><\/h2>\n
\"2019<\/figure>\n

\u9879\u76ee\u7f51\u5740\uff1a<\/span>
\nhttps:\/\/github.com\/streamlit\/streamlit<\/span><\/p>\n

\u4e0d\u8bba\u673a\u5668\u5b66\u4e60\u9879\u76ee\u591a\u4e48\u7684\u8d85\u51e1\u8131\u4fd7\uff0c\u90fd\u514d\u4e0d\u4e86\u8981\u624b\u52a8\u4e0e\u6a21\u578b\u548c\u6570\u636e\u4ea4\u4e92\u7684\u5de5\u4f5c\u3002<\/span><\/p>\n

\u4e0e\u5176\u6295\u5165\u65e0\u6570\u5c0f\u65f6\uff0c\u5f00\u53d1\u6570\u5343\u884c\u4ee3\u7801\u7684\u5e94\u7528\uff0cStreamlit \u8ba9\u4f60\u53ef\u4ee5\u5feb\u901f\u642d\u5efa\u7528\u6765\u5206\u4eab\u6a21\u578b\u4e0e\u5206\u6790\u7684 APP\u3002<\/span>\u6784\u5efa\u4ea4\u4e92\u5f0f UI\u3001\u5b9e\u73b0\u6570\u636e\u53ef\u89c6\u5316\u3001\u8f93\u51fa\u5206\u6790\u6a21\u578b\u73b0\u5728\u5c31\u662f\u5c0f\u83dc\u4e00\u789f\u3002<\/span><\/p>\n

\u6709\u4e86 Streamlit\uff0c\u53ea\u8981\u6dfb\u52a0\u51e0\u884c\u4ee3\u7801\uff0c\u55d6\u7684\u4e00\u4e0b\uff0c\u5c31\u80fd\u4ece Python \u811a\u672c\u4ee3\u7801\u98de\u5347\u4e3a\u4ea7\u54c1\u7ea7 APP\u3002<\/span>TensorFlow\u3001Keras\u3001PyTorch\u3001Pandas\uff0c\u53ea\u8981\u662f\u4f60\u80fd\u60f3\u5230\u7684\u6570\u636e\u79d1\u5b66\u5de5\u5177\uff0cStreamlit \u90fd\u5df2\u7ecf\u652f\u6301\u4e86\u3002<\/span><\/p>\n

8. Transformers<\/span><\/h2>\n
\"2019<\/figure>\n

\u9879\u76ee\u7f51\u5740\uff1a<\/span>
\nhttps:\/\/github.com\/huggingface\/transformers<\/span><\/p>\n

\u4ece\u4e8b\u673a\u5668\u5b66\u4e60\u76f8\u5173\u5de5\u4f5c\u7684\u8bfb\u8005\uff0c\u4e00\u5b9a\u4f1a\u5bf9\u8fc7\u53bb\u7684\u4e00\u5e74\u4e2d\u81ea\u7136\u8bed\u8a00\u5904\u7406\uff08NLP\uff09\u9886\u57df\u91cc\u7684\u91cd\u5927\u7a81\u7834\u6709\u6240\u8033\u95fb\u3002<\/span><\/p>\n

BERT\u3001XLNet\u3001roBERTa \u7b49\u65b0\u6d8c\u73b0\u51fa\u6765\u7684\u9ad8\u6027\u80fd\u6a21\u578b\u5df2\u7ecf\u53d6\u5f97\u4e86\u957f\u8db3\u7684\u53d1\u5c55\uff0c\u5c24\u5176\u662f\u5728\u6587\u672c\u5206\u7c7b\u3001\u673a\u5668\u7ffb\u8bd1\u3001\u547d\u540d\u5b9e\u4f53\u8bc6\u522b\u7b49\u4f17\u591a NLP \u9886\u57df\u91cc\u7684\u9ad8\u7cbe\u5c16\u7a81\u7834\u5c42\u51fa\u4e0d\u7a77\uff0c\u8ba9\u4eba\u773c\u82b1\u7f2d\u4e71\u3002<\/span><\/p>\n

\u5bf9\u4e1a\u754c\u4eba\u58eb\u6765\u8bf4\uff0c\u627e\u5230\u4e00\u6b3e\u65e2\u80fd\u5728\u4ea7\u54c1\u7ea7\u7a0b\u5e8f\u4e0a\u5229\u7528\u8fd9\u4e9b\u6a21\u578b\uff0c\u53c8\u4e0d\u4f1a\u592a\u8fc7\u590d\u6742\u7684\u5de5\u5177\uff0c\u5c24\u4e3a\u91cd\u8981\u3002<\/span>\u5bf9\u7814\u7a76\u8005\u800c\u8a00\uff0c\u4e00\u4e2a\u65e2\u4e0d\u7528\u82b1\u8d39\u592a\u591a\u65f6\u95f4\u7f16\u5199\u6a21\u677f\u4ee3\u7801\uff0c\u53c8\u80fd\u8c03\u6821\u652f\u6301\u5e93\u5185\u6838\uff0c\u4e0d\u65ad\u5f00\u53d1\u3001\u5b9e\u9a8c\u65b0\u6a21\u578b\u7684\u652f\u6301\u5e93\uff0c\u4e5f\u975e\u5e38\u91cd\u8981\u3002<\/span><\/p>\n

Hugging Face \u7684\u5c0f\u4f19\u4f34\u7ed9\u6211\u4eec\u5e26\u6765\u4e86\u53d8\u5f62\u91d1\u521a\uff08Transformers\uff09\uff0c\u8fd9\u4e2a\u652f\u6301\u5e93\u652f\u6301\u6700\u73b0\u4ee3\u7684 NLP \u6a21\u578b\u6253\u5305\u3001\u9884\u8bad\u7ec3\u7b49\u529f\u80fd\uff0c\u800c\u4e14\u8fd8\u5f00\u7bb1\u5373\u7528\u3002<\/span>TensorFlow 2.0 \u4e0e PyTorch \u4e4b\u95f4\u7684\u4e92\u64cd\u4f5c\uff0c\u8ba9\u8be5\u652f\u6301\u5e93\u8fbe\u5230\u4e86\u4e1a\u754c\u6807\u51c6\uff0c\u5b8c\u5168\u53ef\u4ee5\u4e3a\u7814\u7a76\u5de5\u4f5c\u4e0e\u4ea7\u54c1\u7ea7\u5e94\u7528\u7a0b\u5e8f\u52a0\u6301\u3002<\/span>\u8be5\u652f\u6301\u5e93\u7531\u7814\u7a76\u4eba\u5458\u5f00\u53d1\uff0c\u56e0\u6b64\uff0c\u5b83\u7684\u66f4\u65b0\u901f\u5ea6\u975e\u5e38\u5feb\uff0c\u800c\u4e14\u8fd8\u5728\u4e0d\u65ad\u5f15\u5165\u65b0\u5f0f\u6a21\u578b\u3002<\/span><\/p>\n

\u9526\u4e0a\u6dfb\u82b1\uff1a<\/span>Hugging Face \u56e2\u961f\u5f00\u53d1\u4e86 DistilBERT\uff0c\u8fd9\u662f\u57fa\u4e8e BERT \u63d0\u70bc\u51fa\u6765\u7684\uff0c\u66f4\u5c0f\u3001\u66f4\u5feb\u3001\u66f4\u7701\u94b1\u3001\u66f4\u8f7b\u91cf\u7684\u7248\u672c\u3002<\/span><\/p>\n

\u8fd8\u5728\u72b9\u8c6b\u8981\u4e0d\u8981\u7528 Hugging Face \u7684\u53d8\u5f62\u91d1\u521a\u5904\u7406\u73b0\u4ee3 NLP \u4efb\u52a1\uff1f<\/span>\u770b\u5230\u8fd9\u7bc7\u6587\u7ae0\u5c31\u662f\u7f18\u5206\uff0c\u5feb\u6765\u8bd5\u4e0b\u8fd9\u4e2a\u5728\u7ebf\u6f14\u793a\uff0c\u611f\u53d7\u53d8\u5f62\u91d1\u521a\u7684\u9b54\u529b\u5427\u3002<\/span><\/p>\n

\"2019<\/figure>\n

9. Detectron2<\/span><\/h2>\n
\"2019<\/figure>\n

\u9879\u76ee\u7f51\u5740\uff1a<\/span>
\nhttps:\/\/github.com\/facebookresearch\/detectron2<\/span><\/p>\n

Facebook AI \u7814\u7a76\u56e2\u961f\uff08FAIR\uff09\u5728\u5f00\u53d1\u5bf9\u8c61\u63a2\u77e5\u3001\u59ff\u6001\u4f30\u6d4b\u3001\u8bed\u4e49\/\u5b9e\u4f8b\u5206\u9694\u3001\u5168\u666f\u5206\u5272\u7b49\u6a21\u578b\u7684\u540c\u65f6\uff0c\u4e0d\u65ad\u5c06\u8ba1\u7b97\u673a\u89c6\u89c9\uff08CV\uff09\u63a8\u5411\u65b0\u7684\u9ad8\u5ea6\u3002<\/span><\/p>\n

\u5c31\u5728\u77ed\u77ed\u7684\u51e0\u5e74\u524d\uff0c\u4e0a\u8ff0\u8fd9\u4e9b\u6280\u672f\u5c31\u50cf\u662f\u79d1\u5e7b\u5c0f\u8bf4\u3002<\/span>\u6211\u4eec\u4ece\u672a\u5bf9 FAIR \u6709\u8fc7\u4ec0\u4e48\u671f\u671b\uff0c\u4f46\u8fd9\u56de\u4ed6\u4eec\u518d\u4e00\u6b21\u9707\u64bc\u5168\u573a\u3002<\/span><\/p>\n

Detectron2 \u662f Detectron \u7684\u5ef6\u7eed\uff0c\u8be5\u652f\u6301\u5e93\u5efa\u7acb\u5728 PyTorch \u4e4b\u4e0a\uff0c\u6574\u5408\u4e86\u591a\u79cd\u9ad8\u7cbe\u5c16\u7684\u8ba1\u7b97\u673a\u89c6\u89c9\u7b97\u6cd5\u3002<\/span><\/p>\n

\u9274\u4e8e\u652f\u6301\u7528\u4f8b\u7c7b\u578b\u591a\u79cd\u591a\u6837\uff0c\u8fd9\u7c7b\u652f\u6301\u5e93\u7684\u5f00\u53d1\u96be\u5ea6\u7279\u522b\u5927\u3002<\/span>\u4e0e Hugging Face \u7684 Transformers \u7c7b\u4f3c\uff0cFAIR \u56e2\u961f\u7684 Detectron2 \u771f\u7684\u5f88\u5f3a\u5927\uff0c\u5176\u5ef6\u5c55\u6027\u4e0e\u6a21\u5757\u5316\u7684\u8bbe\u8ba1\u98ce\u683c\uff0c\u8ba9\u5b83\u5728\u8ba1\u7b97\u673a\u89c6\u89c9\u7814\u7a76\u5e94\u7528\u4e2d\u72ec\u9886\u98ce\u9a9a\u3002<\/span>\u540c\u65f6\uff0cDetectron2 \u7684\u5e94\u7528\u6781\u5176\u7b80\u5355\uff0c\u975e\u5e38\u9002\u5408\u90a3\u4e9b\u53ea\u60f3\u5feb\u901f\u5f97\u5230\u7ed3\u679c\uff0c\u4e0d\u60f3\u4e86\u89e3\u6838\u5fc3\u7b97\u6cd5\u7684\u4eba\u3002<\/span>\u6ca1\u9519\uff0c\u53ea\u8981\u51e0\u884c Python \u4ee3\u7801\uff0c\u5c31\u53ef\u4ee5\u901a\u8fc7 Detectron2 \u8ba9\u4f60\u5f00\u53d1\u7684\u8f6f\u4ef6\u201c\u7406\u89e3\u201d\u56fe\u7247\u3002<\/span><\/p>\n

Detectron2 \u80fd\u5426\u5f62\u6210\u751f\u673a\u52c3\u52c3\u7684\u793e\u533a\uff0c\u8fd8\u9700\u8981\u4e9b\u65f6\u95f4\u6765\u9a8c\u8bc1\uff0c\u4f46\u5230\u76ee\u524d\u4e3a\u6b62\uff0c\u5b83\u7684\u524d\u9014\u8fd8\u662f\u86ee\u5149\u660e\u7684\u3002<\/span>\u5f88\u591a\u7814\u7a76\u4eba\u5458\u90fd\u5728\u4e3a Detectron2 \u7814\u53d1\u66f4\u65b0\u3001\u66f4\u5feb\u3001\u66f4\u597d\u7684\u6a21\u578b\uff0c\u56e0\u6b64\uff0c\u5b83\u6210\u4e3a\u8ba1\u7b97\u673a\u89c6\u89c9\u5e94\u7528\u201c\u9996\u9009\u201d\u89e3\u51b3\u65b9\u6848\u7684\u53ef\u80fd\u6027\u5f88\u5927\u3002<\/span>\u5982\u679c\u4f60\u4ece\u4e8b\u8ba1\u7b97\u673a\u89c6\u89c9\u5de5\u4f5c\uff0c\u4e00\u5b9a\u8981\u5bf9\u8fd9\u4e2a\u652f\u6301\u5e93\u4fdd\u6301\u5bc6\u5207\u7684\u5173\u6ce8\uff01<\/span><\/p>\n

10. Metaflow<\/span><\/h2>\n

\u9879\u76ee\u7f51\u5740\uff1a
\n<\/span>https:\/\/docs.metaflow.org\/introduction\/what-is-metaflow<\/span><\/p>\n

Metaflow \u8fd9\u4e2a\u652f\u6301\u5e93\u5b9e\u5728\u592a\u65b0\u4e86\uff0c\u5dee\u70b9\u5c31\u6ca1\u8d76\u4e0a 2019 \u5e74\u699c\u5355\u8bc4\u9009\uff01<\/span>\u4f46\u4e0d\u8981\u56e0\u6b64\u5c31\u5c0f\u77a7\u5b83\uff1a<\/span>\u522b\u770b\u5b83 12 \u6708 3 \u65e5\u624d\u6b63\u5f0f\u53d1\u5e03\uff0c\u5176\u5b9e\u5b83\u662f Netfilx \u4e24\u5e74\u7cbe\u96d5\u7ec6\u7422\u7684\u6210\u679c\uff0c\u901a\u8fc7\u4e86\u957f\u671f\u5185\u90e8\u5b9e\u6d4b\uff0cNetflix \u624d\u51b3\u5b9a\u5f00\u6e90\u3002<\/span><\/p>\n

Metaflow \u8fd9\u4e2a Python \u652f\u6301\u5e93\u80fd\u5e2e\u52a9\u6570\u636e\u79d1\u5b66\u5bb6\u4e0e\u6570\u636e\u5de5\u7a0b\u5e08\u5f00\u53d1\u5728\u73b0\u5b9e\u4e16\u754c\u4e2d\u5e94\u7528\u7684\u5b9e\u6218\u9879\u76ee\u3002<\/span>\u5b83\u7684\u76ee\u6807\u662f\uff0c\u51cf\u8f7b\u975e\u6280\u672f\u578b\u6570\u636e\u79d1\u5b66\u5bb6\u5b66\u4e60\u6280\u672f\u7684\u8d1f\u62c5\uff0c\u6bd4\u5982\uff0c\u5982\u4f55\u5229\u7528\u8ba1\u7b97\u8d44\u6e90\u3001\u600e\u4e48\u5b9e\u73b0\u5e76\u884c\u8fd0\u7b97\u3001\u67b6\u6784\u8bbe\u8ba1\u3001\u7248\u672c\u63a7\u5236\u7b49\u3002<\/span>Netflix \u4e0e AWS \u5408\u4f5c\uff0c\u8ba9\u7528\u6237\u901a\u8fc7\u5f00\u7bb1\u5373\u7528\u7684\u5206\u5e03\u5f0f\u8ba1\u7b97\u5373\u53ef\u8f7b\u677e\u5b9a\u4e49\u590d\u6742\u6570\u636e\u6d41\u3002<\/span><\/p>\n

\u6211\u4eec\u5728 Tryolabs \u5185\u90e8\u5df2\u7ecf\u8bc4\u6d4b\u4e86\u51e0\u4e2a Metaflow \u6838\u5fc3\u9879\u76ee\u3002<\/span>\u5982\u679c\u4f60\u4e5f\u5bf9\u8fd9\u4e2a\u5de5\u5177\u611f\u5174\u8da3\uff0c\u8bf7\u53c2\u9605 Netflix \u535a\u5ba2\u7684\u53d1\u5e03\u8d34\u3002<\/span><\/p>\n

\u7ed3\u8bba<\/span><\/h2>\n

\u8fd9\u4e00\u5e74\uff0c\u6216\u8005\u8bf4\u53c8\u4e00\u4e2a10 \u5e74\u9a6c\u4e0a\u5c31\u8981\u8fc7\u53bb\u4e86\uff0c\u5f00\u6e90\u793e\u533a\u8fd9\u4e9b\u5e74\u7684\u5bf9\u4e16\u754c\u7684\u8d21\u732e\u5c06\u5bf9\u4eca\u540e\u51e0\u5e74\u4ea7\u751f\u6df1\u8fdc\u7684\u5f71\u54cd\u3002<\/span>\u60f3\u8981\u4e86\u89e3\u8fd9\u4e9b\u5e74\u7684\u4f18\u79c0\u5f00\u6e90\u5e93\uff0c\u5c31\u6765\u770b\u4e0b\u6211\u4eec\u8fd9\u51e0\u5e74\u7684\u699c\u5355\u5427\uff1a<\/span>2015\u30012016\u30012017\u30012018\u3002<\/span><\/p>\n

\u672c\u6587\u7531 El\u00edas Masquil<\/em><\/span>, <\/em>Alan Descoins<\/em><\/span>\u00a0\u4e0e <\/em>Germ\u00e1n Hoffman<\/em><\/span> \u5408\u4f5c\u7f16\u64b0\u3002<\/span><\/p>\n

\n
\n
\n
\n
\n
\n
\n
\n
\n

\u8f6c\u81ea\uff1ahttps:\/\/mp.weixin.qq.com\/s\/ltty16HXd09P3lCMUB4hHQ<\/p>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/div>\n","protected":false},"excerpt":{"rendered":"

\u6765\u81ea\u516c\u4f17\u53f7\uff1aPython\u5927\u5496\u8c08 \u539f\u6587\u4f5c\u8005\uff1aFabi\u00e1n Torres \u539f\u6587\u94fe\u63a5\uff1ahttps:\/\/tryolab … \u7ee7\u7eed\u9605\u8bfb\u201c2019 \u5e74 10 \u5927\u9876\u7ea7 Python \u652f\u6301\u5e93\u201d<\/span><\/a><\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":[],"categories":[2],"tags":[8],"_links":{"self":[{"href":"http:\/\/www.zhushiyao.com\/index.php?rest_route=\/wp\/v2\/posts\/14964"}],"collection":[{"href":"http:\/\/www.zhushiyao.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/www.zhushiyao.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"http:\/\/www.zhushiyao.com\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"http:\/\/www.zhushiyao.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=14964"}],"version-history":[{"count":2,"href":"http:\/\/www.zhushiyao.com\/index.php?rest_route=\/wp\/v2\/posts\/14964\/revisions"}],"predecessor-version":[{"id":14977,"href":"http:\/\/www.zhushiyao.com\/index.php?rest_route=\/wp\/v2\/posts\/14964\/revisions\/14977"}],"wp:attachment":[{"href":"http:\/\/www.zhushiyao.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=14964"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/www.zhushiyao.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=14964"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/www.zhushiyao.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=14964"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}