{ "returncode": 0, "stdout": "ladder_d5.py:custom_kernel at 1.90% error: 15 timed calls, each in a fresh process\n review flag: 18,670 bytes, close to the 20,480-byte limit\n NVIDIA A100 80GB PCIe, torch 2.12.0+cu130, sandbox on\n call 1/15: 13,270.663 ms, scorer's clock 13,272.947 ms\n call 2/15: 13,259.937 ms, scorer's clock 13,262.602 ms\n call 3/15: 13,256.392 ms, scorer's clock 13,258.749 ms\n call 4/15: 13,257.523 ms, scorer's clock 13,260.214 ms\n call 5/15: 13,258.721 ms, scorer's clock 13,261.232 ms\n call 6/15: 13,261.265 ms, scorer's clock 13,263.695 ms\n call 7/15: 13,262.532 ms, scorer's clock 13,264.764 ms\n call 8/15: 13,256.176 ms, scorer's clock 13,258.655 ms\n call 9/15: 13,259.264 ms, scorer's clock 13,261.527 ms\n call 10/15: 13,272.152 ms, scorer's clock 13,274.448 ms\n call 11/15: 13,260.620 ms, scorer's clock 13,262.844 ms\n call 12/15: 13,259.605 ms, scorer's clock 13,261.932 ms\n call 13/15: 13,257.160 ms, scorer's clock 13,259.646 ms\n call 14/15: 13,252.795 ms, scorer's clock 13,254.983 ms\n call 15/15: 13,271.446 ms, scorer's clock 13,273.971 ms\nMNIST 98.28% (108,109/110,000), 13259.912 ms/call; hold-out (kmnist) 97.56%, 13264.305 ms/call; score 13264.305 ms\nenergy: one more fresh process runs the method back to back on fresh MNIST draws for 20 s, between idle windows in which it is frozen\n NVIDIA A100 80GB PCIe: idle 85.2 W with the method frozen; telemetry reference 18.4 TFLOP/s at 7.69 J/TFLOP above idle (a healthy NVIDIA A100 80GB PCIe reads 6-11)\n the round trip alone: 21.6 mJ per call over 3,461 empty calls, subtracted\n the method: 3 calls in 39.7 s, MNIST 98.22%, 13,241.146 ms per call, 222.9 W on average\nenergy 1824759.863 mJ per call above idle\n", "stderr": "", "record": { "version": "mnist-a100/1.2.0 (protections of sutro-mnist-medium/3.0.0, tightened; energy column)", "file": "ladder_d5.py", "function": "custom_kernel", "source_sha256": "3c0ff0d43ead0c17654116886da37ba345e4e7d72a04fd96c5255061a9dc0c9e", "band_bp": 190, "sandboxed": true, "holdout": "kmnist", "device": "NVIDIA A100 80GB PCIe", "torch": "2.12.0+cu130", "calls": [ { "dataset": "kmnist", "ms": 13270.6630859375, "parent_ms": 13275.189149999998, "correct": 9760, "total": 10000, "overhead_ms": 2.2420374990401655 }, { "dataset": "mnist", "ms": 13259.9365234375, "parent_ms": 13264.576349000003, "correct": 9855, "total": 10000, "overhead_ms": 1.9740355001248417 }, { "dataset": "kmnist", "ms": 13256.3916015625, "parent_ms": 13260.679865, "correct": 9750, "total": 10000, "overhead_ms": 1.9311914992973698 }, { "dataset": "mnist", "ms": 13257.5234375, "parent_ms": 13261.809886999998, "correct": 9822, "total": 10000, "overhead_ms": 1.5962760000335265 }, { "dataset": "kmnist", "ms": 13258.720703125, "parent_ms": 13263.071127000017, "correct": 9761, "total": 10000, "overhead_ms": 1.8391720001163776 }, { "dataset": "mnist", "ms": 13261.2646484375, "parent_ms": 13265.824814000012, "correct": 9820, "total": 10000, "overhead_ms": 2.130279500040956 }, { "dataset": "mnist", "ms": 13262.5322265625, "parent_ms": 13267.027748999993, "correct": 9841, "total": 10000, "overhead_ms": 2.2633554987291973 }, { "dataset": "mnist", "ms": 13256.17578125, "parent_ms": 13260.600381000018, "correct": 9816, "total": 10000, "overhead_ms": 1.9455344987944727 }, { "dataset": "mnist", "ms": 13259.263671875, "parent_ms": 13263.471319000018, "correct": 9840, "total": 10000, "overhead_ms": 1.944169500292901 }, { "dataset": "mnist", "ms": 13272.15234375, "parent_ms": 13276.41960599999, "correct": 9834, "total": 10000, "overhead_ms": 1.9720255014638042 }, { "dataset": "mnist", "ms": 13260.6201171875, "parent_ms": 13264.792996999971, "correct": 9824, "total": 10000, "overhead_ms": 1.949172499484348 }, { "dataset": "mnist", "ms": 13259.60546875, "parent_ms": 13263.871594999955, "correct": 9811, "total": 10000, "overhead_ms": 1.9399299984152094 }, { "dataset": "mnist", "ms": 13257.16015625, "parent_ms": 13261.631287, "correct": 9832, "total": 10000, "overhead_ms": 1.9848044991294955 }, { "dataset": "mnist", "ms": 13252.794921875, "parent_ms": 13256.97217300001, "correct": 9814, "total": 10000, "overhead_ms": 1.989317000455685 }, { "dataset": "kmnist", "ms": 13271.4462890625, "parent_ms": 13275.869709000062, "correct": 9753, "total": 10000, "overhead_ms": 1.8988460010305062 } ], "problems": [], "summary": "MNIST 98.28% (108,109/110,000), 13259.912 ms/call; hold-out (kmnist) 97.56%, 13264.305 ms/call", "ranked_ms": 13264.305419921875, "energy": { "mj_per_call": 1824759.8632064017, "problems": [], "net_mj_before_control": 1824781.470068497, "control_mj_per_call": 21.606862095203226, "control_calls": 3461, "gross_mj_per_call": 2952713.0, "idle_w": 85.15201885984206, "active_w": 222.91199987907922, "calls": 3, "seconds": 39.738277907, "ms_median": 13241.1455078125, "timed_mnist_ms_median": 13259.60546875, "correct": 29465, "total": 30000, "reference": { "tflops_per_s": 18.408059123782344, "j_per_tflop": 7.689257501091745, "exact": true, "band": [ [ 6.0, 11.0 ], [ 15.0, 23.0 ] ] }, "device": { "name": "NVIDIA A100 80GB PCIe", "uuid": "GPU-ed96ace1-e726-e430-b430-bc2b637af6d5", "vbios": "92.00.A0.00.05", "driver": "580.95.05", "power_limit_w": 300.0 }, "windows": [ { "name": "idle 1", "seconds": 5.005440483000029, "joules": 442.381, "utilization_max": 0, "contexts": 1 }, { "name": "reference", "seconds": 5.196501759000057, "joules": 1191.806, "matmuls": 696, "dim": 4096, "exact": true }, { "name": "idle 2", "seconds": 5.003653812000039, "joules": 436.453, "utilization_max": 0, "contexts": 1 }, { "name": "control", "seconds": 5.00383822099991, "joules": 503.594, "calls": 3461 }, { "name": "idle 3", "seconds": 5.003918832999943, "joules": 421.163, "utilization_max": 0, "contexts": 1 }, { "name": "method", "seconds": 39.738277907, "joules": 8858.139, "calls": 3, "ms": [ 13241.267578125, 13241.1455078125, 13239.6435546875 ], "correct": [ 9815, 9842, 9808 ] }, { "name": "idle 4", "seconds": 5.003424487000075, "joules": 430.982, "utilization_max": 0, "contexts": 1 } ] } } }