Accumulator-Carrying Picard ResNet / results.json

Mechanism confirmed, baseline not beaten

Raw ⬇ ZIP
  1{
  2  "device": "cuda",
  3  "math_check": {
  4    "telescoping_max_error": 8.881784197001252e-16,
  5    "composition_max_error": 0.0,
  6    "bounded_tanh_max_abs": 0.9994070271659857
  7  },
  8  "runs": [
  9    {
 10      "d": 50,
 11      "model": "plain_mlp",
 12      "params": 11649,
 13      "test_mse": 0.09562712162733078,
 14      "checkpoints": [
 15        0.670299232006073,
 16        0.08917469531297684,
 17        0.08228223025798798,
 18        0.09562712162733078
 19      ],
 20      "seconds": 2.0152041912078857
 21    },
 22    {
 23      "d": 50,
 24      "model": "standard_resnet",
 25      "params": 16310,
 26      "test_mse": 0.1040823757648468,
 27      "checkpoints": [
 28        0.6617469787597656,
 29        0.09144451469182968,
 30        0.09496200829744339,
 31        0.1040823757648468
 32      ],
 33      "seconds": 4.086727142333984
 34    },
 35    {
 36      "d": 50,
 37      "model": "accumulator",
 38      "params": 16310,
 39      "test_mse": 0.086046002805233,
 40      "checkpoints": [
 41        0.7085394263267517,
 42        0.08119411766529083,
 43        0.08469437062740326,
 44        0.086046002805233
 45      ],
 46      "seconds": 4.47427225112915
 47    },
 48    {
 49      "d": 100,
 50      "model": "plain_mlp",
 51      "params": 14849,
 52      "test_mse": 0.12415221333503723,
 53      "checkpoints": [
 54        0.7376440763473511,
 55        0.0995437279343605,
 56        0.10381902009248734,
 57        0.12415221333503723
 58      ],
 59      "seconds": 1.5389134883880615
 60    },
 61    {
 62      "d": 100,
 63      "model": "standard_resnet",
 64      "params": 17910,
 65      "test_mse": 0.13711242377758026,
 66      "checkpoints": [
 67        0.7288199663162231,
 68        0.10812899470329285,
 69        0.1201668456196785,
 70        0.13711242377758026
 71      ],
 72      "seconds": 3.8147177696228027
 73    },
 74    {
 75      "d": 100,
 76      "model": "accumulator",
 77      "params": 17910,
 78      "test_mse": 0.1185467317700386,
 79      "checkpoints": [
 80        0.7433508038520813,
 81        0.09851141273975372,
 82        0.10474535822868347,
 83        0.1185467317700386
 84      ],
 85      "seconds": 4.297095537185669
 86    },
 87    {
 88      "d": 200,
 89      "model": "plain_mlp",
 90      "params": 21249,
 91      "test_mse": 0.10115337371826172,
 92      "checkpoints": [
 93        0.7190290093421936,
 94        0.09635581821203232,
 95        0.09813550114631653,
 96        0.10115337371826172
 97      ],
 98      "seconds": 1.7245187759399414
 99    },
100    {
101      "d": 200,
102      "model": "standard_resnet",
103      "params": 21110,
104      "test_mse": 0.1254808008670807,
105      "checkpoints": [
106        0.7280761003494263,
107        0.09675059467554092,
108        0.11149366945028305,
109        0.1254808008670807
110      ],
111      "seconds": 3.7922558784484863
112    },
113    {
114      "d": 200,
115      "model": "accumulator",
116      "params": 21110,
117      "test_mse": 0.11386863887310028,
118      "checkpoints": [
119        0.7529879212379456,
120        0.09412874281406403,
121        0.0989784374833107,
122        0.11386863887310028
123      ],
124      "seconds": 4.683218002319336
125    }
126  ]
127}