1{
 2  "device": "cuda",
 3  "n": 16,
 4  "forward_max_abs_error": 0.0009072721004486084,
 5  "forward_relative_mean_error": 5.060033149364018e-05,
 6  "gradient_max_abs_error": 0.00116729736328125,
 7  "out_of_range_finite": true,
 8  "out_of_range_outputs": [
 9    0.0,
10    0.0,
11    1.0,
12    1.0,
13    1.0
14  ],
15  "forward_ms_truncated": 2.0198953799748174,
16  "forward_ms_cox_de_boor": 2.6233356533339247,
17  "forward_speedup": 1.298748281391995,
18  "step_ms_truncated": 4.397457100761433,
19  "step_ms_cox_de_boor": 5.538173966730635,
20  "step_speedup": 1.2594037508112774,
21  "float64_max_abs_error": 1.701000451603818e-12,
22  "float64_mean_abs_error": 7.190531822729489e-14
23}