{
 "returncode": 0,
 "stdout": "submission.py:classify at 1.90% error: 15 timed calls, each in a fresh process\n  review flag: embedded data? 1 high-entropy literal(s), 19,775 characters\n  review flag: decodes data: imports base64, lzma\n  review flag: 20,428 bytes, close to the 20,480-byte limit\n  NVIDIA A100-SXM4-80GB, torch 2.12.0+cu130, sandbox on\n  call  1/15: 7,021.333 ms, scorer's clock 7,024.708 ms\n  call  2/15: 7,022.052 ms, scorer's clock 7,024.994 ms\n  call  3/15: 7,021.928 ms, scorer's clock 7,025.472 ms\n  call  4/15: 7,021.150 ms, scorer's clock 7,024.730 ms\n  call  5/15: 7,020.046 ms, scorer's clock 7,023.266 ms\n  call  6/15: 7,020.604 ms, scorer's clock 7,027.402 ms\n  call  7/15: 7,021.410 ms, scorer's clock 7,025.013 ms\n  call  8/15: 7,021.755 ms, scorer's clock 7,027.052 ms\n  call  9/15: 7,022.683 ms, scorer's clock 7,025.325 ms\n  call 10/15: 7,021.345 ms, scorer's clock 7,024.805 ms\n  call 11/15: 7,022.710 ms, scorer's clock 7,028.102 ms\n  call 12/15: 7,021.568 ms, scorer's clock 7,024.847 ms\n  call 13/15: 7,022.360 ms, scorer's clock 7,025.993 ms\n  call 14/15: 7,021.764 ms, scorer's clock 7,025.550 ms\n  call 15/15: 7,021.874 ms, scorer's clock 7,026.374 ms\nMNIST 98.16% (107,977/110,000), 7021.679 ms/call; hold-out (kmnist) 97.28%, 7021.529 ms/call; score 7021.679 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-SXM4-80GB: idle 70.5 W with the method frozen; telemetry reference 18.0 TFLOP/s at 8.45 J/TFLOP above idle (a healthy NVIDIA A100-SXM4-80GB reads 6-11)\n  the round trip alone: 13.8 mJ per call over 1,703 empty calls, subtracted\n  the method: 3 calls in 21.1 s, MNIST 98.05%, 7,027.627 ms per call, 344.8 W on average\nenergy 1931206.167 mJ per call above idle\n",
 "stderr": "tes=8 winner=0 id=6 tile=20 split=1 us=55.7227\nLT tuned m=4096 n=16 k=512 at=0 bt=1 E=4 candidates=6 winner=1 id=21 tile=5 split=1 us=18.1760\nLT tuned m=512 n=16 k=4096 at=1 bt=1 E=4 candidates=4 winner=2 id=31 tile=15 split=1 us=22.6987\nLT tuned m=4096 n=512 k=16 at=0 bt=0 E=4 candidates=4 winner=0 id=5 tile=24 split=1 us=22.8693\nLT tuned m=512 n=512 k=4096 at=1 bt=1 E=4 candidates=5 winner=3 id=31 tile=18 split=1 us=86.1867\nLT tuned m=4096 n=512 k=512 at=0 bt=1 E=4 candidates=8 winner=0 id=6 tile=20 split=1 us=55.2960\nLT tuned m=512 n=128 k=4096 at=1 bt=1 E=4 candidates=5 winner=4 id=31 tile=15 split=1 us=24.5760\nLT tuned m=20000 n=512 k=128 at=0 bt=0 E=4 candidates=8 winner=0 id=6 tile=20 split=1 us=111.3600\nLT tuned m=20000 n=512 k=512 at=0 bt=0 E=4 candidates=8 winner=0 id=6 tile=20 split=1 us=235.7760\nLT tuned m=20000 n=16 k=512 at=0 bt=1 E=4 candidates=8 winner=0 id=6 tile=17 split=1 us=60.3307\n[1/3] /usr/local/cuda/bin/nvcc -MD -MF centered_halfnorm_small_members.cuda.o.d -DTORCH_EXTENSION_NAME=sutro_centered_halfnorm_small_members_bn_stagepack -DTORCH_API_INCLUDE_EXTENSION_H -isystem /usr/local/lib/python3.13/site-packages/torch/include -isystem /usr/local/lib/python3.13/site-packages/torch/include/torch/csrc/api/include -isystem /usr/local/cuda/include -isystem /usr/local/include/python3.13 -D__CUDA_NO_HALF_OPERATORS__ -D__CUDA_NO_HALF_CONVERSIONS__ -D__CUDA_NO_BFLOAT16_CONVERSIONS__ -D__CUDA_NO_HALF2_OPERATORS__ --expt-relaxed-constexpr -gencode=arch=compute_80,code=sm_80 --compiler-options '-fPIC' -O3 -lineinfo --ptxas-options=-v -std=c++20 -c /tmp/sutro-bn-stage-582fe8613daf6be1-65534/kernels/centered_halfnorm_small_members.cu -o centered_halfnorm_small_members.cuda.o \nptxas info    : 0 bytes gmem\nptxas info    : Compiling entry function '_Z9transformILi8EEvPK13__nv_bfloat16S2_PKfPS0_i' for 'sm_80'\nptxas info    : Function properties for _Z9transformILi8EEvPK13__nv_bfloat16S2_PKfPS0_i\n    0 bytes stack frame, 0 bytes spill stores, 0 bytes spill loads\nptxas info    : Used 50 registers, used 1 barriers, 68 bytes smem, 388 bytes cmem[0]\nptxas info    : Compile time = 34.224 ms\nptxas info    : Compiling entry function '_Z9transformILi4EEvPK13__nv_bfloat16S2_PKfPS0_i' for 'sm_80'\nptxas info    : Function properties for _Z9transformILi4EEvPK13__nv_bfloat16S2_PKfPS0_i\n    0 bytes stack frame, 0 bytes spill stores, 0 bytes spill loads\nptxas info    : Used 32 registers, used 1 barriers, 36 bytes smem, 388 bytes cmem[0]\nptxas info    : Compile time = 19.116 ms\nptxas info    : Compiling entry function '_Z9transformILi2EEvPK13__nv_bfloat16S2_PKfPS0_i' for 'sm_80'\nptxas info    : Function properties for _Z9transformILi2EEvPK13__nv_bfloat16S2_PKfPS0_i\n    0 bytes stack frame, 0 bytes spill stores, 0 bytes spill loads\nptxas info    : Used 32 registers, used 1 barriers, 20 bytes smem, 388 bytes cmem[0]\nptxas info    : Compile time = 18.953 ms\nptxas info    : Compiling entry function '_Z9transformILi1EEvPK13__nv_bfloat16S2_PKfPS0_i' for 'sm_80'\nptxas info    : Function properties for _Z9transformILi1EEvPK13__nv_bfloat16S2_PKfPS0_i\n    0 bytes stack frame, 0 bytes spill stores, 0 bytes spill loads\nptxas info    : Used 32 registers, used 1 barriers, 12 bytes smem, 388 bytes cmem[0]\nptxas info    : Compile time = 16.977 ms\nptxas info    : Compiling entry function '_Z11finish_meanPKfPfiii' for 'sm_80'\nptxas info    : Function properties for _Z11finish_meanPKfPfiii\n    0 bytes stack frame, 0 bytes spill stores, 0 bytes spill loads\nptxas info    : Used 32 registers, used 0 barriers, 380 bytes cmem[0]\nptxas info    : Compile time = 11.761 ms\nptxas info    : Compiling entry function '_Z12partial_meanPK13__nv_bfloat16S1_Pfiiii' for 'sm_80'\nptxas info    : Function properties for _Z12partial_meanPK13__nv_bfloat16S1_Pfiiii\n    0 bytes stack frame, 0 bytes spill stores, 0 bytes spill loads\nptxas info    : Used 32 registers, used 1 barriers, 1024 bytes smem, 392 bytes cmem[0]\nptxas info    : Compile time = 9.154 ms\n[2/3] c++ -MMD -MF centered_halfnorm_small_members_bindings.o.d -DTORCH_EXTENSION_NAME=sutro_centered_halfnorm_small_members_bn_stagepack -DTORCH_API_INCLUDE_EXTENSION_H -isystem /usr/local/lib/python3.13/site-packages/torch/include -isystem /usr/local/lib/python3.13/site-packages/torch/include/torch/csrc/api/include -isystem /usr/local/cuda/include -isystem /usr/local/include/python3.13 -fPIC -std=c++20 -O0 -g0 -c /tmp/sutro-bn-stage-582fe8613daf6be1-65534/kernels/centered_halfnorm_small_members_bindings.cpp -o centered_halfnorm_small_members_bindings.o \n[3/3] c++ centered_halfnorm_small_members_bindings.o centered_halfnorm_small_members.cuda.o -shared -L/usr/local/lib/python3.13/site-packages/torch/lib -lc10 -lc10_cuda -ltorch_cpu -ltorch_cuda -ltorch -ltorch_python -L/usr/local/cuda/lib64 -lcudart -o sutro_centered_halfnorm_small_members_bn_stagepack.so\nLT tuned m=10 n=128 k=20000 at=1 bt=1 E=1 candidates=7 winner=0 id=23 tile=11 split=48 us=19.1147\nLT tuned m=4096 n=512 k=128 at=0 bt=0 E=4 candidates=8 winner=4 id=6 tile=18 split=1 us=33.1947\nLT tuned m=4096 n=512 k=512 at=0 bt=0 E=4 candidates=8 winner=0 id=6 tile=20 split=1 us=55.8080\nLT tuned m=4096 n=16 k=512 at=0 bt=1 E=4 candidates=6 winner=2 id=31 tile=15 split=1 us=16.4693\nLT tuned m=512 n=16 k=4096 at=1 bt=1 E=4 candidates=4 winner=2 id=31 tile=15 split=1 us=22.6987\nLT tuned m=4096 n=512 k=16 at=0 bt=0 E=4 candidates=4 winner=0 id=5 tile=24 split=1 us=22.7840\nLT tuned m=512 n=512 k=4096 at=1 bt=1 E=4 candidates=5 winner=3 id=31 tile=18 split=1 us=85.7600\nLT tuned m=4096 n=512 k=512 at=0 bt=1 E=4 candidates=8 winner=0 id=6 tile=20 split=1 us=55.8933\nLT tuned m=512 n=128 k=4096 at=1 bt=1 E=4 candidates=5 winner=0 id=31 tile=15 split=1 us=24.7467\nLT tuned m=20000 n=512 k=128 at=0 bt=0 E=4 candidates=8 winner=0 id=6 tile=20 split=1 us=111.4453\nLT tuned m=20000 n=512 k=512 at=0 bt=0 E=4 candidates=8 winner=0 id=6 tile=20 split=1 us=235.6053\nLT tuned m=20000 n=16 k=512 at=0 bt=1 E=4 candidates=8 winner=0 id=6 tile=17 split=1 us=58.7947\n",
 "record": {
  "version": "mnist-a100/1.2.0 (protections of sutro-mnist-medium/3.0.0, tightened; energy column)",
  "file": "submission.py",
  "function": "classify",
  "source_sha256": "86a1a749f5b12c34be86c2e5025487827bda6b8b20f55fac72bb2911dcaa9066",
  "band_bp": 190,
  "sandboxed": true,
  "holdout": "kmnist",
  "device": "NVIDIA A100-SXM4-80GB",
  "torch": "2.12.0+cu130",
  "calls": [
   {
    "dataset": "kmnist",
    "ms": 7021.33349609375,
    "parent_ms": 7027.664999999999,
    "correct": 9733,
    "total": 10000,
    "overhead_ms": 2.9573449996488037
   },
   {
    "dataset": "kmnist",
    "ms": 7022.05224609375,
    "parent_ms": 7028.137399999991,
    "correct": 9727,
    "total": 10000,
    "overhead_ms": 3.1431605003352843
   },
   {
    "dataset": "mnist",
    "ms": 7021.92822265625,
    "parent_ms": 7028.480134000006,
    "correct": 9827,
    "total": 10000,
    "overhead_ms": 3.008378500472645
   },
   {
    "dataset": "mnist",
    "ms": 7021.150390625,
    "parent_ms": 7027.587672000038,
    "correct": 9816,
    "total": 10000,
    "overhead_ms": 2.8578049996435766
   },
   {
    "dataset": "kmnist",
    "ms": 7020.04638671875,
    "parent_ms": 7026.25498399999,
    "correct": 9727,
    "total": 10000,
    "overhead_ms": 2.9894570003311856
   },
   {
    "dataset": "mnist",
    "ms": 7020.6044921875,
    "parent_ms": 7030.366366999999,
    "correct": 9799,
    "total": 10000,
    "overhead_ms": 2.9647414993937673
   },
   {
    "dataset": "mnist",
    "ms": 7021.41015625,
    "parent_ms": 7027.835771000014,
    "correct": 9803,
    "total": 10000,
    "overhead_ms": 2.8231010023919225
   },
   {
    "dataset": "mnist",
    "ms": 7021.75537109375,
    "parent_ms": 7029.955608000023,
    "correct": 9809,
    "total": 10000,
    "overhead_ms": 2.9031229986212566
   },
   {
    "dataset": "kmnist",
    "ms": 7022.68310546875,
    "parent_ms": 7029.340759999968,
    "correct": 9725,
    "total": 10000,
    "overhead_ms": 4.015281499894627
   },
   {
    "dataset": "mnist",
    "ms": 7021.3447265625,
    "parent_ms": 7027.5677920000135,
    "correct": 9818,
    "total": 10000,
    "overhead_ms": 2.7624364971065916
   },
   {
    "dataset": "mnist",
    "ms": 7022.7099609375,
    "parent_ms": 7031.252012999971,
    "correct": 9821,
    "total": 10000,
    "overhead_ms": 3.149972000335765
   },
   {
    "dataset": "mnist",
    "ms": 7021.56787109375,
    "parent_ms": 7027.875109999968,
    "correct": 9832,
    "total": 10000,
    "overhead_ms": 3.027984500716684
   },
   {
    "dataset": "mnist",
    "ms": 7022.3603515625,
    "parent_ms": 7028.855835999821,
    "correct": 9801,
    "total": 10000,
    "overhead_ms": 2.8623955018556444
   },
   {
    "dataset": "mnist",
    "ms": 7021.763671875,
    "parent_ms": 7028.1952429998,
    "correct": 9820,
    "total": 10000,
    "overhead_ms": 2.645194999638079
   },
   {
    "dataset": "mnist",
    "ms": 7021.8740234375,
    "parent_ms": 7029.062399000168,
    "correct": 9831,
    "total": 10000,
    "overhead_ms": 2.688191500562425
   }
  ],
  "problems": [],
  "summary": "MNIST 98.16% (107,977/110,000), 7021.679 ms/call; hold-out (kmnist) 97.28%, 7021.529 ms/call",
  "ranked_ms": 7021.679021661932,
  "energy": {
   "mj_per_call": 1931206.1669086744,
   "problems": [],
   "net_mj_before_control": 1931219.9320106541,
   "control_mj_per_call": 13.765101979628302,
   "control_calls": 1703,
   "gross_mj_per_call": 2427587.0,
   "idle_w": 70.49109802600958,
   "active_w": 344.7514636231659,
   "calls": 3,
   "seconds": 21.124670286999844,
   "ms_median": 7027.626953125,
   "timed_mnist_ms_median": 7021.75537109375,
   "correct": 29416,
   "total": 30000,
   "reference": {
    "tflops_per_s": 18.01372078101125,
    "j_per_tflop": 8.445815257049489,
    "exact": true,
    "band": [
     [
      6.0,
      11.0
     ],
     [
      15.0,
      23.0
     ]
    ]
   },
   "device": {
    "name": "NVIDIA A100-SXM4-80GB",
    "uuid": "GPU-8a18e100-7e85-0e50-473f-78efce7e77ee",
    "vbios": "92.00.9E.00.01",
    "driver": "580.95.05",
    "power_limit_w": 400.0
   },
   "windows": [
    {
     "name": "idle 1",
     "seconds": 5.006141637999917,
     "joules": 344.237,
     "utilization_max": 0,
     "contexts": 1
    },
    {
     "name": "reference",
     "seconds": 5.24922092099996,
     "joules": 1163.046,
     "matmuls": 688,
     "dim": 4096,
     "exact": true
    },
    {
     "name": "idle 2",
     "seconds": 5.004547964000039,
     "joules": 350.753,
     "utilization_max": 0,
     "contexts": 1
    },
    {
     "name": "control",
     "seconds": 5.00481783400005,
     "joules": 374.095,
     "calls": 1703
    },
    {
     "name": "idle 3",
     "seconds": 5.005015604000164,
     "joules": 350.548,
     "utilization_max": 0,
     "contexts": 1
    },
    {
     "name": "method",
     "seconds": 21.124670286999844,
     "joules": 7282.761,
     "calls": 3,
     "ms": [
      7045.89306640625,
      7026.45849609375,
      7027.626953125
     ],
     "correct": [
      9808,
      9805,
      9803
     ]
    },
    {
     "name": "idle 4",
     "seconds": 5.005634537000105,
     "joules": 355.114,
     "utilization_max": 0,
     "contexts": 1
    }
   ]
  }
 }
}
