Parallel Quadratic Tree Layer / results.json

✓✓ Beats tuned baseline

Raw ⬇ ZIP
  1{
  2  "device": "cuda",
  3  "math_checks": {
  4    "quadratic_scaling": {
  5      "gammas": [
  6        0.05,
  7        0.1,
  8        0.2,
  9        0.4,
 10        0.8
 11      ],
 12      "observed_delta_H": [
 13        -0.0010000000000000002,
 14        -0.004000000000000001,
 15        -0.016000000000000004,
 16        -0.06400000000000002,
 17        -0.25600000000000006
 18      ],
 19      "predicted_delta_H": [
 20        -0.0010000000000000002,
 21        -0.004000000000000001,
 22        -0.016000000000000004,
 23        -0.06400000000000002,
 24        -0.25600000000000006
 25      ],
 26      "max_relative_error": 0.0,
 27      "delta_H_over_gamma2_spread": 0.0
 28    },
 29    "damping_scaling": {
 30      "epsilons": [
 31        0.0,
 32        0.1,
 33        0.3,
 34        0.8
 35      ],
 36      "observed_magnitude": [
 37        0.1,
 38        0.09615384615384615,
 39        0.08928571428571429,
 40        0.07575757575757576
 41      ],
 42      "predicted_magnitude": [
 43        0.1,
 44        0.09615384615384615,
 45        0.08928571428571429,
 46        0.07575757575757576
 47      ],
 48      "max_absolute_error": 0.0,
 49      "strictly_decreasing": true
 50    },
 51    "balanced_depth": {
 52      "rows": [
 53        {
 54          "n": 3,
 55          "observed_levels": 1,
 56          "predicted_levels": 1
 57        },
 58        {
 59          "n": 7,
 60          "observed_levels": 2,
 61          "predicted_levels": 2
 62        },
 63        {
 64          "n": 15,
 65          "observed_levels": 3,
 66          "predicted_levels": 3
 67        },
 68        {
 69          "n": 31,
 70          "observed_levels": 4,
 71          "predicted_levels": 4
 72        },
 73        {
 74          "n": 63,
 75          "observed_levels": 5,
 76          "predicted_levels": 5
 77        }
 78      ],
 79      "prediction_confirmed": true
 80    }
 81  },
 82  "agreement": {
 83    "max_root_coefficient_difference": 4.440892098500626e-16,
 84    "max_input_gradient_difference": 8.881784197001252e-16,
 85    "reconstruction_finite": true,
 86    "levels": 6
 87  },
 88  "benchmark": [
 89    {
 90      "n": 127,
 91      "sequential_ms": 224.8403503326699,
 92      "rake_ms": 223.14003400000124,
 93      "rake_over_sequential": 0.9924376726412634
 94    },
 95    {
 96      "n": 511,
 97      "sequential_ms": 927.3020416658255,
 98      "rake_ms": 905.5352006677518,
 99      "rake_over_sequential": 0.9765266978610645
100    },
101    {
102      "n": 1023,
103      "sequential_ms": 2069.1302990004865,
104      "rake_ms": 1907.6914530002493,
105      "rake_over_sequential": 0.9219774385024366
106    }
107  ],
108  "cuda_fallback_error": null
109}