让摆杆学会连续控制

CPU · Python 3.12.13 · 6.09 s · 2026-10-22 08:53:43 UTC+8

本次运行结果

{'observation': [-0.14995256066322327, 0.9886931777000427, -0.12224312126636505], 'torque_range': [-2, 2], 'steps': 100000}

Zero torque / untrained mean: -1309.083083455785 -1313.4932457569505

Held-out return before / after: -1313.4932457569505 -1245.1743549325186

Optional exercise: fill answer and enable RUN_EXERCISE.

metrics.json

{
  "zero_torque_returns": [
    -616.5846227403325,
    -942.2967150490249,
    -970.1812460538372,
    -1322.12299706458,
    -1830.6665369115167,
    -1606.7303503800854,
    -1069.423959214765,
    -1651.8971687435771,
    -1374.9455493560688,
    -1705.9816890440609
  ],
  "before_returns": [
    -629.337306154083,
    -958.8697477183122,
    -969.5552225354786,
    -1325.3578921569165,
    -1831.741622901378,
    -1609.063323066726,
    -1070.1215137590723,
    -1654.01809973512,
    -1378.9232341319007,
    -1707.944495410518
  ],
  "after_returns": [
    -1100.886027923237,
    -1160.8816830907372,
    -1137.8037433976367,
    -1124.841653518832,
    -1632.8321858670245,
    -1222.962293879415,
    -1096.4351639798394,
    -1348.2304777619054,
    -1166.695450868538,
    -1460.174869038021
  ]
}