Gokulapriyan's picture
End of training
6582283
{
"best_metric": 0.9782978378816098,
"best_model_checkpoint": "swinv2-tiny-patch4-window8-256-finetuned-og-dataset-10e-finetuned-og-dataset-10e/checkpoint-1092",
"epoch": 1.9986282578875172,
"global_step": 1092,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.02,
"learning_rate": 4.5454545454545455e-06,
"loss": 0.2055,
"step": 10
},
{
"epoch": 0.04,
"learning_rate": 9.090909090909091e-06,
"loss": 0.2267,
"step": 20
},
{
"epoch": 0.05,
"learning_rate": 1.3636363636363637e-05,
"loss": 0.2416,
"step": 30
},
{
"epoch": 0.07,
"learning_rate": 1.8181818181818182e-05,
"loss": 0.206,
"step": 40
},
{
"epoch": 0.09,
"learning_rate": 2.272727272727273e-05,
"loss": 0.2383,
"step": 50
},
{
"epoch": 0.11,
"learning_rate": 2.7272727272727273e-05,
"loss": 0.2187,
"step": 60
},
{
"epoch": 0.13,
"learning_rate": 3.181818181818182e-05,
"loss": 0.2272,
"step": 70
},
{
"epoch": 0.15,
"learning_rate": 3.6363636363636364e-05,
"loss": 0.2259,
"step": 80
},
{
"epoch": 0.16,
"learning_rate": 4.0909090909090915e-05,
"loss": 0.2384,
"step": 90
},
{
"epoch": 0.18,
"learning_rate": 4.545454545454546e-05,
"loss": 0.2495,
"step": 100
},
{
"epoch": 0.2,
"learning_rate": 5e-05,
"loss": 0.2547,
"step": 110
},
{
"epoch": 0.22,
"learning_rate": 4.9490835030549896e-05,
"loss": 0.2233,
"step": 120
},
{
"epoch": 0.24,
"learning_rate": 4.89816700610998e-05,
"loss": 0.2539,
"step": 130
},
{
"epoch": 0.26,
"learning_rate": 4.84725050916497e-05,
"loss": 0.2629,
"step": 140
},
{
"epoch": 0.27,
"learning_rate": 4.79633401221996e-05,
"loss": 0.2284,
"step": 150
},
{
"epoch": 0.29,
"learning_rate": 4.745417515274949e-05,
"loss": 0.2647,
"step": 160
},
{
"epoch": 0.31,
"learning_rate": 4.694501018329939e-05,
"loss": 0.2756,
"step": 170
},
{
"epoch": 0.33,
"learning_rate": 4.643584521384929e-05,
"loss": 0.2573,
"step": 180
},
{
"epoch": 0.35,
"learning_rate": 4.592668024439919e-05,
"loss": 0.2583,
"step": 190
},
{
"epoch": 0.37,
"learning_rate": 4.541751527494909e-05,
"loss": 0.253,
"step": 200
},
{
"epoch": 0.38,
"learning_rate": 4.490835030549899e-05,
"loss": 0.2678,
"step": 210
},
{
"epoch": 0.4,
"learning_rate": 4.439918533604888e-05,
"loss": 0.2518,
"step": 220
},
{
"epoch": 0.42,
"learning_rate": 4.3890020366598776e-05,
"loss": 0.2511,
"step": 230
},
{
"epoch": 0.44,
"learning_rate": 4.338085539714868e-05,
"loss": 0.2686,
"step": 240
},
{
"epoch": 0.46,
"learning_rate": 4.287169042769857e-05,
"loss": 0.2609,
"step": 250
},
{
"epoch": 0.48,
"learning_rate": 4.236252545824848e-05,
"loss": 0.237,
"step": 260
},
{
"epoch": 0.49,
"learning_rate": 4.185336048879837e-05,
"loss": 0.2512,
"step": 270
},
{
"epoch": 0.51,
"learning_rate": 4.134419551934827e-05,
"loss": 0.2511,
"step": 280
},
{
"epoch": 0.53,
"learning_rate": 4.083503054989817e-05,
"loss": 0.2532,
"step": 290
},
{
"epoch": 0.55,
"learning_rate": 4.032586558044807e-05,
"loss": 0.262,
"step": 300
},
{
"epoch": 0.57,
"learning_rate": 3.981670061099796e-05,
"loss": 0.285,
"step": 310
},
{
"epoch": 0.59,
"learning_rate": 3.930753564154787e-05,
"loss": 0.2432,
"step": 320
},
{
"epoch": 0.6,
"learning_rate": 3.879837067209776e-05,
"loss": 0.2498,
"step": 330
},
{
"epoch": 0.62,
"learning_rate": 3.8289205702647656e-05,
"loss": 0.2676,
"step": 340
},
{
"epoch": 0.64,
"learning_rate": 3.778004073319756e-05,
"loss": 0.2773,
"step": 350
},
{
"epoch": 0.66,
"learning_rate": 3.727087576374745e-05,
"loss": 0.2349,
"step": 360
},
{
"epoch": 0.68,
"learning_rate": 3.676171079429735e-05,
"loss": 0.257,
"step": 370
},
{
"epoch": 0.7,
"learning_rate": 3.625254582484725e-05,
"loss": 0.2374,
"step": 380
},
{
"epoch": 0.71,
"learning_rate": 3.574338085539715e-05,
"loss": 0.2531,
"step": 390
},
{
"epoch": 0.73,
"learning_rate": 3.523421588594705e-05,
"loss": 0.237,
"step": 400
},
{
"epoch": 0.75,
"learning_rate": 3.472505091649695e-05,
"loss": 0.2559,
"step": 410
},
{
"epoch": 0.77,
"learning_rate": 3.421588594704684e-05,
"loss": 0.2477,
"step": 420
},
{
"epoch": 0.79,
"learning_rate": 3.370672097759674e-05,
"loss": 0.2202,
"step": 430
},
{
"epoch": 0.8,
"learning_rate": 3.319755600814664e-05,
"loss": 0.2655,
"step": 440
},
{
"epoch": 0.82,
"learning_rate": 3.268839103869654e-05,
"loss": 0.2348,
"step": 450
},
{
"epoch": 0.84,
"learning_rate": 3.217922606924644e-05,
"loss": 0.2322,
"step": 460
},
{
"epoch": 0.86,
"learning_rate": 3.167006109979633e-05,
"loss": 0.227,
"step": 470
},
{
"epoch": 0.88,
"learning_rate": 3.116089613034623e-05,
"loss": 0.2518,
"step": 480
},
{
"epoch": 0.9,
"learning_rate": 3.065173116089613e-05,
"loss": 0.2472,
"step": 490
},
{
"epoch": 0.91,
"learning_rate": 3.014256619144603e-05,
"loss": 0.2826,
"step": 500
},
{
"epoch": 0.93,
"learning_rate": 2.9633401221995927e-05,
"loss": 0.2387,
"step": 510
},
{
"epoch": 0.95,
"learning_rate": 2.9124236252545828e-05,
"loss": 0.2321,
"step": 520
},
{
"epoch": 0.97,
"learning_rate": 2.8615071283095725e-05,
"loss": 0.2018,
"step": 530
},
{
"epoch": 0.99,
"learning_rate": 2.8105906313645626e-05,
"loss": 0.2237,
"step": 540
},
{
"epoch": 1.0,
"eval_accuracy": 0.9735201230868896,
"eval_loss": 0.07291990518569946,
"eval_runtime": 204.6202,
"eval_samples_per_second": 60.351,
"eval_steps_per_second": 1.886,
"step": 546
},
{
"epoch": 1.01,
"learning_rate": 2.759674134419552e-05,
"loss": 0.2748,
"step": 550
},
{
"epoch": 1.03,
"learning_rate": 2.708757637474542e-05,
"loss": 0.2414,
"step": 560
},
{
"epoch": 1.04,
"learning_rate": 2.6578411405295317e-05,
"loss": 0.2247,
"step": 570
},
{
"epoch": 1.06,
"learning_rate": 2.606924643584521e-05,
"loss": 0.2446,
"step": 580
},
{
"epoch": 1.08,
"learning_rate": 2.5560081466395115e-05,
"loss": 0.2291,
"step": 590
},
{
"epoch": 1.1,
"learning_rate": 2.505091649694501e-05,
"loss": 0.221,
"step": 600
},
{
"epoch": 1.12,
"learning_rate": 2.454175152749491e-05,
"loss": 0.2218,
"step": 610
},
{
"epoch": 1.14,
"learning_rate": 2.403258655804481e-05,
"loss": 0.2289,
"step": 620
},
{
"epoch": 1.15,
"learning_rate": 2.3523421588594704e-05,
"loss": 0.1879,
"step": 630
},
{
"epoch": 1.17,
"learning_rate": 2.30142566191446e-05,
"loss": 0.2292,
"step": 640
},
{
"epoch": 1.19,
"learning_rate": 2.2505091649694502e-05,
"loss": 0.244,
"step": 650
},
{
"epoch": 1.21,
"learning_rate": 2.19959266802444e-05,
"loss": 0.2235,
"step": 660
},
{
"epoch": 1.23,
"learning_rate": 2.1486761710794297e-05,
"loss": 0.2208,
"step": 670
},
{
"epoch": 1.25,
"learning_rate": 2.0977596741344197e-05,
"loss": 0.2306,
"step": 680
},
{
"epoch": 1.26,
"learning_rate": 2.0468431771894095e-05,
"loss": 0.2214,
"step": 690
},
{
"epoch": 1.28,
"learning_rate": 1.9959266802443992e-05,
"loss": 0.1851,
"step": 700
},
{
"epoch": 1.3,
"learning_rate": 1.9450101832993893e-05,
"loss": 0.2152,
"step": 710
},
{
"epoch": 1.32,
"learning_rate": 1.894093686354379e-05,
"loss": 0.1935,
"step": 720
},
{
"epoch": 1.34,
"learning_rate": 1.8431771894093687e-05,
"loss": 0.2138,
"step": 730
},
{
"epoch": 1.35,
"learning_rate": 1.7922606924643584e-05,
"loss": 0.221,
"step": 740
},
{
"epoch": 1.37,
"learning_rate": 1.741344195519348e-05,
"loss": 0.2085,
"step": 750
},
{
"epoch": 1.39,
"learning_rate": 1.6904276985743382e-05,
"loss": 0.1815,
"step": 760
},
{
"epoch": 1.41,
"learning_rate": 1.639511201629328e-05,
"loss": 0.2142,
"step": 770
},
{
"epoch": 1.43,
"learning_rate": 1.5885947046843177e-05,
"loss": 0.2027,
"step": 780
},
{
"epoch": 1.45,
"learning_rate": 1.5376782077393077e-05,
"loss": 0.1703,
"step": 790
},
{
"epoch": 1.46,
"learning_rate": 1.4867617107942975e-05,
"loss": 0.1879,
"step": 800
},
{
"epoch": 1.48,
"learning_rate": 1.4358452138492872e-05,
"loss": 0.2016,
"step": 810
},
{
"epoch": 1.5,
"learning_rate": 1.3849287169042771e-05,
"loss": 0.1797,
"step": 820
},
{
"epoch": 1.52,
"learning_rate": 1.334012219959267e-05,
"loss": 0.2204,
"step": 830
},
{
"epoch": 1.54,
"learning_rate": 1.2830957230142567e-05,
"loss": 0.1896,
"step": 840
},
{
"epoch": 1.56,
"learning_rate": 1.2321792260692464e-05,
"loss": 0.1661,
"step": 850
},
{
"epoch": 1.57,
"learning_rate": 1.1812627291242363e-05,
"loss": 0.1973,
"step": 860
},
{
"epoch": 1.59,
"learning_rate": 1.1303462321792262e-05,
"loss": 0.2137,
"step": 870
},
{
"epoch": 1.61,
"learning_rate": 1.079429735234216e-05,
"loss": 0.2134,
"step": 880
},
{
"epoch": 1.63,
"learning_rate": 1.0285132382892057e-05,
"loss": 0.2033,
"step": 890
},
{
"epoch": 1.65,
"learning_rate": 9.775967413441956e-06,
"loss": 0.1814,
"step": 900
},
{
"epoch": 1.67,
"learning_rate": 9.266802443991853e-06,
"loss": 0.2123,
"step": 910
},
{
"epoch": 1.68,
"learning_rate": 8.757637474541752e-06,
"loss": 0.1829,
"step": 920
},
{
"epoch": 1.7,
"learning_rate": 8.248472505091651e-06,
"loss": 0.1747,
"step": 930
},
{
"epoch": 1.72,
"learning_rate": 7.739307535641548e-06,
"loss": 0.1861,
"step": 940
},
{
"epoch": 1.74,
"learning_rate": 7.230142566191446e-06,
"loss": 0.1823,
"step": 950
},
{
"epoch": 1.76,
"learning_rate": 6.7209775967413446e-06,
"loss": 0.2024,
"step": 960
},
{
"epoch": 1.78,
"learning_rate": 6.211812627291243e-06,
"loss": 0.2012,
"step": 970
},
{
"epoch": 1.79,
"learning_rate": 5.702647657841141e-06,
"loss": 0.1792,
"step": 980
},
{
"epoch": 1.81,
"learning_rate": 5.193482688391039e-06,
"loss": 0.2112,
"step": 990
},
{
"epoch": 1.83,
"learning_rate": 4.684317718940937e-06,
"loss": 0.1968,
"step": 1000
},
{
"epoch": 1.85,
"learning_rate": 4.175152749490835e-06,
"loss": 0.2072,
"step": 1010
},
{
"epoch": 1.87,
"learning_rate": 3.6659877800407332e-06,
"loss": 0.1716,
"step": 1020
},
{
"epoch": 1.89,
"learning_rate": 3.1568228105906318e-06,
"loss": 0.1922,
"step": 1030
},
{
"epoch": 1.9,
"learning_rate": 2.64765784114053e-06,
"loss": 0.1972,
"step": 1040
},
{
"epoch": 1.92,
"learning_rate": 2.1384928716904276e-06,
"loss": 0.2012,
"step": 1050
},
{
"epoch": 1.94,
"learning_rate": 1.6293279022403257e-06,
"loss": 0.2165,
"step": 1060
},
{
"epoch": 1.96,
"learning_rate": 1.120162932790224e-06,
"loss": 0.1747,
"step": 1070
},
{
"epoch": 1.98,
"learning_rate": 6.109979633401222e-07,
"loss": 0.169,
"step": 1080
},
{
"epoch": 1.99,
"learning_rate": 1.0183299389002036e-07,
"loss": 0.1672,
"step": 1090
},
{
"epoch": 2.0,
"eval_accuracy": 0.9782978378816098,
"eval_loss": 0.05557282269001007,
"eval_runtime": 213.2817,
"eval_samples_per_second": 57.9,
"eval_steps_per_second": 1.81,
"step": 1092
},
{
"epoch": 2.0,
"step": 1092,
"total_flos": 4.5505049329926144e+18,
"train_loss": 0.22463261898293163,
"train_runtime": 4458.2012,
"train_samples_per_second": 31.39,
"train_steps_per_second": 0.245
}
],
"max_steps": 1092,
"num_train_epochs": 2,
"total_flos": 4.5505049329926144e+18,
"trial_name": null,
"trial_params": null
}