
=== Start adding workers ===
=> Add worker SGDMWorker(index=0, momentum=0.9)
=> Add worker SGDMWorker(index=1, momentum=0.9)
=> Add worker SGDMWorker(index=2, momentum=0.9)
=> Add worker SGDMWorker(index=3, momentum=0.9)
=> Add worker SGDMWorker(index=4, momentum=0.9)
=> Add worker SGDMWorker(index=5, momentum=0.9)
=> Add worker SGDMWorker(index=6, momentum=0.9)
=> Add worker SGDMWorker(index=7, momentum=0.9)
=> Add worker SGDMWorker(index=8, momentum=0.9)
=> Add worker SGDMWorker(index=9, momentum=0.9)
=> Add worker SGDMWorker(index=10, momentum=0.9)
=> Add worker SGDMWorker(index=11, momentum=0.9)
=> Add worker SGDMWorker(index=12, momentum=0.9)
=> Add worker SGDMWorker(index=13, momentum=0.9)
=> Add worker SGDMWorker(index=14, momentum=0.9)
=> Add worker SGDMWorker(index=15, momentum=0.9)
=> Add worker SGDMWorker(index=16, momentum=0.9)
=> Add worker SGDMWorker(index=17, momentum=0.9)
=> Add worker SGDMWorker(index=18, momentum=0.9)
=> Add worker SGDMWorker(index=19, momentum=0.9)
=> Add worker ByzantineWorker(index=20)
=> Add worker ByzantineWorker(index=21)

=== Start adding graph ===
<codes.graph_utils.Dumbbell object at 0x7fecddd4f490>

Train epoch 1
[E 1B0  |    704/60000 (  1%) ] Loss: 2.3066 top1=  9.2188

=== Peeking data label distribution E1B0 ===
Worker 0 has targets: tensor([3, 2, 2, 4, 3], device='cuda:0')
Worker 1 has targets: tensor([1, 0, 0, 4, 0], device='cuda:0')
Worker 2 has targets: tensor([4, 1, 0, 1, 0], device='cuda:0')
Worker 3 has targets: tensor([0, 1, 4, 1, 3], device='cuda:0')
Worker 4 has targets: tensor([0, 4, 1, 2, 4], device='cuda:0')
Worker 5 has targets: tensor([2, 2, 4, 4, 4], device='cuda:0')
Worker 6 has targets: tensor([1, 1, 4, 4, 3], device='cuda:0')
Worker 7 has targets: tensor([4, 4, 1, 3, 0], device='cuda:0')
Worker 8 has targets: tensor([1, 3, 1, 0, 4], device='cuda:0')
Worker 9 has targets: tensor([1, 3, 3, 3, 1], device='cuda:0')
Worker 10 has targets: tensor([8, 9, 7, 7, 9], device='cuda:0')
Worker 11 has targets: tensor([8, 9, 6, 6, 7], device='cuda:0')
Worker 12 has targets: tensor([8, 6, 5, 7, 8], device='cuda:0')
Worker 13 has targets: tensor([7, 6, 9, 6, 5], device='cuda:0')
Worker 14 has targets: tensor([8, 5, 8, 6, 7], device='cuda:0')
Worker 15 has targets: tensor([9, 5, 6, 8, 6], device='cuda:0')
Worker 16 has targets: tensor([7, 7, 8, 5, 8], device='cuda:0')
Worker 17 has targets: tensor([9, 7, 5, 6, 6], device='cuda:0')
Worker 18 has targets: tensor([7, 7, 7, 6, 6], device='cuda:0')
Worker 19 has targets: tensor([5, 7, 9, 9, 7], device='cuda:0')
Worker 20 has targets: tensor([3, 2, 2, 4, 3], device='cuda:0')
Worker 21 has targets: tensor([8, 9, 7, 7, 9], device='cuda:0')


[E 1B10 |   7744/60000 ( 13%) ] Loss: 1.0377 top1= 62.6562
[E 1B20 |  14784/60000 ( 25%) ] Loss: 0.3903 top1= 88.1250

=> Averaged model (Global Average Validation Accuracy) | Eval Loss=0.6714 top1= 80.3986


=> Averaged model (Clique1 Average Validation Accuracy) | Eval Loss=7.5093 top1= 49.3490


=> Averaged model (Clique2 Average Validation Accuracy) | Eval Loss=6.5810 top1= 44.2508

Train epoch 2
[E 2B0  |    704/60000 (  1%) ] Loss: 0.2756 top1= 91.4062
[E 2B10 |   7744/60000 ( 13%) ] Loss: 0.1924 top1= 94.0625
[E 2B20 |  14784/60000 ( 25%) ] Loss: 0.1940 top1= 94.5312

=> Averaged model (Global Average Validation Accuracy) | Eval Loss=0.5209 top1= 86.8590


=> Averaged model (Clique1 Average Validation Accuracy) | Eval Loss=3.7919 top1= 49.9900


=> Averaged model (Clique2 Average Validation Accuracy) | Eval Loss=3.6464 top1= 45.8634

Train epoch 3
[E 3B0  |    704/60000 (  1%) ] Loss: 0.1595 top1= 95.1562
[E 3B10 |   7744/60000 ( 13%) ] Loss: 0.1110 top1= 97.1875
[E 3B20 |  14784/60000 ( 25%) ] Loss: 0.1401 top1= 96.0938

=> Averaged model (Global Average Validation Accuracy) | Eval Loss=0.4255 top1= 87.6002


=> Averaged model (Clique1 Average Validation Accuracy) | Eval Loss=3.3726 top1= 49.7696


=> Averaged model (Clique2 Average Validation Accuracy) | Eval Loss=3.0112 top1= 46.4844

Train epoch 4
[E 4B0  |    704/60000 (  1%) ] Loss: 0.1321 top1= 95.9375
[E 4B10 |   7744/60000 ( 13%) ] Loss: 0.0788 top1= 97.9688
[E 4B20 |  14784/60000 ( 25%) ] Loss: 0.1060 top1= 96.7188

=> Averaged model (Global Average Validation Accuracy) | Eval Loss=0.3713 top1= 88.6118


=> Averaged model (Clique1 Average Validation Accuracy) | Eval Loss=3.1446 top1= 50.7212


=> Averaged model (Clique2 Average Validation Accuracy) | Eval Loss=2.7883 top1= 47.9267

Train epoch 5
[E 5B0  |    704/60000 (  1%) ] Loss: 0.1010 top1= 96.8750
[E 5B10 |   7744/60000 ( 13%) ] Loss: 0.0594 top1= 98.5938
[E 5B20 |  14784/60000 ( 25%) ] Loss: 0.0846 top1= 97.1875

=> Averaged model (Global Average Validation Accuracy) | Eval Loss=0.3352 top1= 89.3229


=> Averaged model (Clique1 Average Validation Accuracy) | Eval Loss=2.9738 top1= 52.0833


=> Averaged model (Clique2 Average Validation Accuracy) | Eval Loss=2.6243 top1= 50.0501

Train epoch 6
[E 6B0  |    704/60000 (  1%) ] Loss: 0.0746 top1= 97.8125
[E 6B10 |   7744/60000 ( 13%) ] Loss: 0.0419 top1= 99.0625
[E 6B20 |  14784/60000 ( 25%) ] Loss: 0.0662 top1= 98.1250

=> Averaged model (Global Average Validation Accuracy) | Eval Loss=0.3079 top1= 89.8738


=> Averaged model (Clique1 Average Validation Accuracy) | Eval Loss=2.8584 top1= 53.5958


=> Averaged model (Clique2 Average Validation Accuracy) | Eval Loss=2.6045 top1= 51.4623

Train epoch 7
[E 7B0  |    704/60000 (  1%) ] Loss: 0.0576 top1= 98.4375
[E 7B10 |   7744/60000 ( 13%) ] Loss: 0.0304 top1= 99.3750
[E 7B20 |  14784/60000 ( 25%) ] Loss: 0.0580 top1= 98.1250

=> Averaged model (Global Average Validation Accuracy) | Eval Loss=0.2999 top1= 90.1242


=> Averaged model (Clique1 Average Validation Accuracy) | Eval Loss=2.8467 top1= 54.5373


=> Averaged model (Clique2 Average Validation Accuracy) | Eval Loss=2.5937 top1= 52.7043

Train epoch 8
[E 8B0  |    704/60000 (  1%) ] Loss: 0.0470 top1= 98.9062
[E 8B10 |   7744/60000 ( 13%) ] Loss: 0.0287 top1= 99.2188
[E 8B20 |  14784/60000 ( 25%) ] Loss: 0.0441 top1= 99.2188

=> Averaged model (Global Average Validation Accuracy) | Eval Loss=0.2993 top1= 90.0641


=> Averaged model (Clique1 Average Validation Accuracy) | Eval Loss=2.7921 top1= 55.8093


=> Averaged model (Clique2 Average Validation Accuracy) | Eval Loss=2.6657 top1= 53.3854

Train epoch 9
[E 9B0  |    704/60000 (  1%) ] Loss: 0.0398 top1= 98.7500
[E 9B10 |   7744/60000 ( 13%) ] Loss: 0.0329 top1= 98.7500
[E 9B20 |  14784/60000 ( 25%) ] Loss: 0.0533 top1= 98.1250

=> Averaged model (Global Average Validation Accuracy) | Eval Loss=0.2997 top1= 90.1042


=> Averaged model (Clique1 Average Validation Accuracy) | Eval Loss=2.7300 top1= 56.3101


=> Averaged model (Clique2 Average Validation Accuracy) | Eval Loss=2.5990 top1= 55.0982

Train epoch 10
[E10B0  |    704/60000 (  1%) ] Loss: 0.0427 top1= 98.2812
[E10B10 |   7744/60000 ( 13%) ] Loss: 0.0284 top1= 99.0625
[E10B20 |  14784/60000 ( 25%) ] Loss: 0.0502 top1= 98.1250

=> Averaged model (Global Average Validation Accuracy) | Eval Loss=0.2873 top1= 90.5248


=> Averaged model (Clique1 Average Validation Accuracy) | Eval Loss=2.5965 top1= 57.7023


=> Averaged model (Clique2 Average Validation Accuracy) | Eval Loss=2.0776 top1= 58.4335

Train epoch 11
[E11B0  |    704/60000 (  1%) ] Loss: 0.0344 top1= 99.0625
[E11B10 |   7744/60000 ( 13%) ] Loss: 0.0285 top1= 99.2188
[E11B20 |  14784/60000 ( 25%) ] Loss: 0.0384 top1= 98.7500

=> Averaged model (Global Average Validation Accuracy) | Eval Loss=0.2728 top1= 90.9255


=> Averaged model (Clique1 Average Validation Accuracy) | Eval Loss=2.5597 top1= 58.9944


=> Averaged model (Clique2 Average Validation Accuracy) | Eval Loss=2.6636 top1= 51.9832

Train epoch 12
[E12B0  |    704/60000 (  1%) ] Loss: 0.0800 top1= 97.5000
[E12B10 |   7744/60000 ( 13%) ] Loss: 0.0325 top1= 98.7500
[E12B20 |  14784/60000 ( 25%) ] Loss: 0.0188 top1= 99.5312

=> Averaged model (Global Average Validation Accuracy) | Eval Loss=0.2655 top1= 91.2460


=> Averaged model (Clique1 Average Validation Accuracy) | Eval Loss=2.5012 top1= 59.7756


=> Averaged model (Clique2 Average Validation Accuracy) | Eval Loss=2.2967 top1= 55.2083

Train epoch 13
[E13B0  |    704/60000 (  1%) ] Loss: 0.0213 top1= 99.3750
[E13B10 |   7744/60000 ( 13%) ] Loss: 0.0190 top1= 99.5312
[E13B20 |  14784/60000 ( 25%) ] Loss: 0.0170 top1= 99.6875

=> Averaged model (Global Average Validation Accuracy) | Eval Loss=0.2497 top1= 91.7368


=> Averaged model (Clique1 Average Validation Accuracy) | Eval Loss=2.1605 top1= 61.1078


=> Averaged model (Clique2 Average Validation Accuracy) | Eval Loss=2.2260 top1= 56.9010

Train epoch 14
[E14B0  |    704/60000 (  1%) ] Loss: 0.0101 top1= 99.8438
[E14B10 |   7744/60000 ( 13%) ] Loss: 0.0112 top1=100.0000
[E14B20 |  14784/60000 ( 25%) ] Loss: 0.0211 top1= 99.3750

=> Averaged model (Global Average Validation Accuracy) | Eval Loss=0.2370 top1= 92.2676


=> Averaged model (Clique1 Average Validation Accuracy) | Eval Loss=2.5836 top1= 59.6354


=> Averaged model (Clique2 Average Validation Accuracy) | Eval Loss=2.0418 top1= 58.8642

Train epoch 15
[E15B0  |    704/60000 (  1%) ] Loss: 0.0099 top1= 99.8438
[E15B10 |   7744/60000 ( 13%) ] Loss: 0.0084 top1= 99.8438
[E15B20 |  14784/60000 ( 25%) ] Loss: 0.0154 top1= 99.8438

=> Averaged model (Global Average Validation Accuracy) | Eval Loss=0.2315 top1= 92.4780


=> Averaged model (Clique1 Average Validation Accuracy) | Eval Loss=2.2474 top1= 61.4784


=> Averaged model (Clique2 Average Validation Accuracy) | Eval Loss=2.0778 top1= 58.7941

Train epoch 16
[E16B0  |    704/60000 (  1%) ] Loss: 0.0064 top1= 99.8438
[E16B10 |   7744/60000 ( 13%) ] Loss: 0.0057 top1=100.0000
[E16B20 |  14784/60000 ( 25%) ] Loss: 0.0083 top1= 99.8438

=> Averaged model (Global Average Validation Accuracy) | Eval Loss=0.2289 top1= 92.6382


=> Averaged model (Clique1 Average Validation Accuracy) | Eval Loss=2.2305 top1= 62.1294


=> Averaged model (Clique2 Average Validation Accuracy) | Eval Loss=1.9431 top1= 60.9575

Train epoch 17
[E17B0  |    704/60000 (  1%) ] Loss: 0.0078 top1=100.0000
[E17B10 |   7744/60000 ( 13%) ] Loss: 0.0062 top1=100.0000
[E17B20 |  14784/60000 ( 25%) ] Loss: 0.0041 top1=100.0000

=> Averaged model (Global Average Validation Accuracy) | Eval Loss=0.2297 top1= 92.7885


=> Averaged model (Clique1 Average Validation Accuracy) | Eval Loss=1.9732 top1= 63.9022


=> Averaged model (Clique2 Average Validation Accuracy) | Eval Loss=1.8692 top1= 63.4916

Train epoch 18
[E18B0  |    704/60000 (  1%) ] Loss: 0.0035 top1=100.0000
[E18B10 |   7744/60000 ( 13%) ] Loss: 0.0050 top1=100.0000
[E18B20 |  14784/60000 ( 25%) ] Loss: 0.0057 top1=100.0000

=> Averaged model (Global Average Validation Accuracy) | Eval Loss=0.2354 top1= 92.6783


=> Averaged model (Clique1 Average Validation Accuracy) | Eval Loss=1.8951 top1= 64.7336


=> Averaged model (Clique2 Average Validation Accuracy) | Eval Loss=1.7916 top1= 64.7536

Train epoch 19
[E19B0  |    704/60000 (  1%) ] Loss: 0.0028 top1=100.0000
[E19B10 |   7744/60000 ( 13%) ] Loss: 0.0035 top1=100.0000
[E19B20 |  14784/60000 ( 25%) ] Loss: 0.0070 top1=100.0000

=> Averaged model (Global Average Validation Accuracy) | Eval Loss=0.2376 top1= 92.6783


=> Averaged model (Clique1 Average Validation Accuracy) | Eval Loss=1.8316 top1= 65.7452


=> Averaged model (Clique2 Average Validation Accuracy) | Eval Loss=1.8044 top1= 65.1042

Train epoch 20
[E20B0  |    704/60000 (  1%) ] Loss: 0.0040 top1=100.0000
[E20B10 |   7744/60000 ( 13%) ] Loss: 0.0034 top1=100.0000
[E20B20 |  14784/60000 ( 25%) ] Loss: 0.0071 top1= 99.6875

=> Averaged model (Global Average Validation Accuracy) | Eval Loss=0.2334 top1= 92.9087


=> Averaged model (Clique1 Average Validation Accuracy) | Eval Loss=1.8154 top1= 66.2861


=> Averaged model (Clique2 Average Validation Accuracy) | Eval Loss=1.7488 top1= 66.0457

Train epoch 21
[E21B0  |    704/60000 (  1%) ] Loss: 0.0024 top1=100.0000
[E21B10 |   7744/60000 ( 13%) ] Loss: 0.0034 top1=100.0000
[E21B20 |  14784/60000 ( 25%) ] Loss: 0.0058 top1=100.0000

=> Averaged model (Global Average Validation Accuracy) | Eval Loss=0.2329 top1= 92.9688


=> Averaged model (Clique1 Average Validation Accuracy) | Eval Loss=1.7938 top1= 66.7368


=> Averaged model (Clique2 Average Validation Accuracy) | Eval Loss=1.6990 top1= 67.3277

Train epoch 22
[E22B0  |    704/60000 (  1%) ] Loss: 0.0023 top1=100.0000
[E22B10 |   7744/60000 ( 13%) ] Loss: 0.0038 top1=100.0000
[E22B20 |  14784/60000 ( 25%) ] Loss: 0.0041 top1=100.0000

=> Averaged model (Global Average Validation Accuracy) | Eval Loss=0.2290 top1= 93.2692


=> Averaged model (Clique1 Average Validation Accuracy) | Eval Loss=1.7148 top1= 67.6082


=> Averaged model (Clique2 Average Validation Accuracy) | Eval Loss=1.5099 top1= 68.3894

Train epoch 23
[E23B0  |    704/60000 (  1%) ] Loss: 0.0020 top1=100.0000
[E23B10 |   7744/60000 ( 13%) ] Loss: 0.0024 top1=100.0000
[E23B20 |  14784/60000 ( 25%) ] Loss: 0.0045 top1=100.0000

=> Averaged model (Global Average Validation Accuracy) | Eval Loss=0.2283 top1= 93.3794


=> Averaged model (Clique1 Average Validation Accuracy) | Eval Loss=1.6821 top1= 68.1190


=> Averaged model (Clique2 Average Validation Accuracy) | Eval Loss=1.3993 top1= 69.3309

Train epoch 24
[E24B0  |    704/60000 (  1%) ] Loss: 0.0033 top1=100.0000
[E24B10 |   7744/60000 ( 13%) ] Loss: 0.0021 top1=100.0000
[E24B20 |  14784/60000 ( 25%) ] Loss: 0.0071 top1= 99.6875

=> Averaged model (Global Average Validation Accuracy) | Eval Loss=0.2247 top1= 93.3494


=> Averaged model (Clique1 Average Validation Accuracy) | Eval Loss=1.5970 top1= 69.2708


=> Averaged model (Clique2 Average Validation Accuracy) | Eval Loss=1.4517 top1= 69.4111

Train epoch 25
[E25B0  |    704/60000 (  1%) ] Loss: 0.0018 top1=100.0000
[E25B10 |   7744/60000 ( 13%) ] Loss: 0.0027 top1=100.0000
[E25B20 |  14784/60000 ( 25%) ] Loss: 0.0022 top1=100.0000

=> Averaged model (Global Average Validation Accuracy) | Eval Loss=0.2281 top1= 93.3694


=> Averaged model (Clique1 Average Validation Accuracy) | Eval Loss=1.5684 top1= 69.8117


=> Averaged model (Clique2 Average Validation Accuracy) | Eval Loss=1.5695 top1= 68.5797

Train epoch 26
[E26B0  |    704/60000 (  1%) ] Loss: 0.0017 top1=100.0000
[E26B10 |   7744/60000 ( 13%) ] Loss: 0.0046 top1= 99.8438
[E26B20 |  14784/60000 ( 25%) ] Loss: 0.0086 top1= 99.8438

=> Averaged model (Global Average Validation Accuracy) | Eval Loss=0.2317 top1= 93.2792


=> Averaged model (Clique1 Average Validation Accuracy) | Eval Loss=1.5313 top1= 70.4127


=> Averaged model (Clique2 Average Validation Accuracy) | Eval Loss=1.7380 top1= 67.4179

Train epoch 27
[E27B0  |    704/60000 (  1%) ] Loss: 0.0017 top1=100.0000
[E27B10 |   7744/60000 ( 13%) ] Loss: 0.0015 top1=100.0000
[E27B20 |  14784/60000 ( 25%) ] Loss: 0.0018 top1=100.0000

=> Averaged model (Global Average Validation Accuracy) | Eval Loss=0.2324 top1= 93.3794


=> Averaged model (Clique1 Average Validation Accuracy) | Eval Loss=1.5235 top1= 70.8734


=> Averaged model (Clique2 Average Validation Accuracy) | Eval Loss=1.6115 top1= 69.3009

Train epoch 28
[E28B0  |    704/60000 (  1%) ] Loss: 0.0016 top1=100.0000
[E28B10 |   7744/60000 ( 13%) ] Loss: 0.0016 top1=100.0000
[E28B20 |  14784/60000 ( 25%) ] Loss: 0.0018 top1=100.0000

=> Averaged model (Global Average Validation Accuracy) | Eval Loss=0.2317 top1= 93.4295


=> Averaged model (Clique1 Average Validation Accuracy) | Eval Loss=1.4941 top1= 71.3642


=> Averaged model (Clique2 Average Validation Accuracy) | Eval Loss=1.4819 top1= 71.0136

Train epoch 29
[E29B0  |    704/60000 (  1%) ] Loss: 0.0012 top1=100.0000
[E29B10 |   7744/60000 ( 13%) ] Loss: 0.0012 top1=100.0000
[E29B20 |  14784/60000 ( 25%) ] Loss: 0.0014 top1=100.0000

=> Averaged model (Global Average Validation Accuracy) | Eval Loss=0.2340 top1= 93.4996


=> Averaged model (Clique1 Average Validation Accuracy) | Eval Loss=1.4642 top1= 71.7949


=> Averaged model (Clique2 Average Validation Accuracy) | Eval Loss=1.3804 top1= 72.7364

Train epoch 30
[E30B0  |    704/60000 (  1%) ] Loss: 0.0013 top1=100.0000
[E30B10 |   7744/60000 ( 13%) ] Loss: 0.0013 top1=100.0000
[E30B20 |  14784/60000 ( 25%) ] Loss: 0.0013 top1=100.0000

=> Averaged model (Global Average Validation Accuracy) | Eval Loss=0.2351 top1= 93.5096


=> Averaged model (Clique1 Average Validation Accuracy) | Eval Loss=1.4309 top1= 72.2055


=> Averaged model (Clique2 Average Validation Accuracy) | Eval Loss=1.3615 top1= 73.2672

