mobile-ui-rtdetrv2 / trainer_state.json
merve's picture
merve HF Staff
End of training
1741198 verified
{
"best_global_step": 1932,
"best_metric": 0.166,
"best_model_checkpoint": "mobile-ui-rtdetrv2/checkpoint-1932",
"epoch": 30.0,
"eval_steps": 500,
"global_step": 2520,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.2976190476190476,
"grad_norm": 33.42657470703125,
"learning_rate": 4.7619047619047615e-06,
"loss": 30.43977294921875,
"step": 25
},
{
"epoch": 0.5952380952380952,
"grad_norm": 19.833362579345703,
"learning_rate": 9.722222222222223e-06,
"loss": 26.7940478515625,
"step": 50
},
{
"epoch": 0.8928571428571429,
"grad_norm": 15.38772964477539,
"learning_rate": 1.4682539682539683e-05,
"loss": 22.560556640625,
"step": 75
},
{
"epoch": 1.0,
"eval_loss": 14.398598670959473,
"eval_map": 0.005,
"eval_map_50": 0.0119,
"eval_map_75": 0.0031,
"eval_map_group": 0.0113,
"eval_map_image": 0.0051,
"eval_map_large": 0.0154,
"eval_map_medium": 0.0061,
"eval_map_rectangle": 0.0031,
"eval_map_small": 0.0015,
"eval_map_text": 0.0006,
"eval_mar_1": 0.0056,
"eval_mar_10": 0.0315,
"eval_mar_100": 0.1098,
"eval_mar_100_group": 0.1591,
"eval_mar_100_image": 0.1275,
"eval_mar_100_rectangle": 0.1274,
"eval_mar_100_text": 0.0254,
"eval_mar_large": 0.2126,
"eval_mar_medium": 0.1272,
"eval_mar_small": 0.0351,
"eval_runtime": 19.2065,
"eval_samples_per_second": 6.248,
"eval_steps_per_second": 0.781,
"step": 84
},
{
"epoch": 1.1904761904761905,
"grad_norm": 40.03998947143555,
"learning_rate": 1.9642857142857145e-05,
"loss": 19.5402783203125,
"step": 100
},
{
"epoch": 1.4880952380952381,
"grad_norm": 10.218520164489746,
"learning_rate": 2.4603174603174602e-05,
"loss": 18.0630078125,
"step": 125
},
{
"epoch": 1.7857142857142856,
"grad_norm": 9.290128707885742,
"learning_rate": 2.9563492063492066e-05,
"loss": 16.5726025390625,
"step": 150
},
{
"epoch": 2.0,
"eval_loss": 10.262517929077148,
"eval_map": 0.0774,
"eval_map_50": 0.1221,
"eval_map_75": 0.0758,
"eval_map_group": 0.0806,
"eval_map_image": 0.1143,
"eval_map_large": 0.1336,
"eval_map_medium": 0.0946,
"eval_map_rectangle": 0.0612,
"eval_map_small": 0.0293,
"eval_map_text": 0.0535,
"eval_mar_1": 0.041,
"eval_mar_10": 0.1849,
"eval_mar_100": 0.3597,
"eval_mar_100_group": 0.3638,
"eval_mar_100_image": 0.4799,
"eval_mar_100_rectangle": 0.362,
"eval_mar_100_text": 0.2332,
"eval_mar_large": 0.5727,
"eval_mar_medium": 0.431,
"eval_mar_small": 0.1654,
"eval_runtime": 17.3736,
"eval_samples_per_second": 6.907,
"eval_steps_per_second": 0.863,
"step": 168
},
{
"epoch": 2.0833333333333335,
"grad_norm": 7.967964172363281,
"learning_rate": 3.4523809523809526e-05,
"loss": 15.704903564453126,
"step": 175
},
{
"epoch": 2.380952380952381,
"grad_norm": 14.558771133422852,
"learning_rate": 3.9484126984126986e-05,
"loss": 14.5267236328125,
"step": 200
},
{
"epoch": 2.678571428571429,
"grad_norm": 9.619315147399902,
"learning_rate": 4.4444444444444447e-05,
"loss": 13.92162353515625,
"step": 225
},
{
"epoch": 2.9761904761904763,
"grad_norm": 8.275788307189941,
"learning_rate": 4.940476190476191e-05,
"loss": 13.8275439453125,
"step": 250
},
{
"epoch": 3.0,
"eval_loss": 8.35128116607666,
"eval_map": 0.1082,
"eval_map_50": 0.1593,
"eval_map_75": 0.1049,
"eval_map_group": 0.0831,
"eval_map_image": 0.1796,
"eval_map_large": 0.2034,
"eval_map_medium": 0.1423,
"eval_map_rectangle": 0.1099,
"eval_map_small": 0.0428,
"eval_map_text": 0.0602,
"eval_mar_1": 0.0448,
"eval_mar_10": 0.1995,
"eval_mar_100": 0.4293,
"eval_mar_100_group": 0.4094,
"eval_mar_100_image": 0.5438,
"eval_mar_100_rectangle": 0.4053,
"eval_mar_100_text": 0.3589,
"eval_mar_large": 0.6945,
"eval_mar_medium": 0.4812,
"eval_mar_small": 0.2377,
"eval_runtime": 17.5818,
"eval_samples_per_second": 6.825,
"eval_steps_per_second": 0.853,
"step": 252
},
{
"epoch": 3.2738095238095237,
"grad_norm": 17.83608055114746,
"learning_rate": 4.951499118165785e-05,
"loss": 13.419459228515626,
"step": 275
},
{
"epoch": 3.571428571428571,
"grad_norm": 15.467687606811523,
"learning_rate": 4.8963844797178134e-05,
"loss": 13.36093994140625,
"step": 300
},
{
"epoch": 3.869047619047619,
"grad_norm": 6.873251438140869,
"learning_rate": 4.841269841269841e-05,
"loss": 12.94340576171875,
"step": 325
},
{
"epoch": 4.0,
"eval_loss": 8.083844184875488,
"eval_map": 0.1171,
"eval_map_50": 0.1741,
"eval_map_75": 0.1167,
"eval_map_group": 0.064,
"eval_map_image": 0.2013,
"eval_map_large": 0.2186,
"eval_map_medium": 0.1593,
"eval_map_rectangle": 0.1426,
"eval_map_small": 0.0463,
"eval_map_text": 0.0604,
"eval_mar_1": 0.0411,
"eval_mar_10": 0.2116,
"eval_mar_100": 0.4439,
"eval_mar_100_group": 0.4459,
"eval_mar_100_image": 0.5678,
"eval_mar_100_rectangle": 0.4243,
"eval_mar_100_text": 0.3375,
"eval_mar_large": 0.7398,
"eval_mar_medium": 0.5069,
"eval_mar_small": 0.2457,
"eval_runtime": 18.0359,
"eval_samples_per_second": 6.653,
"eval_steps_per_second": 0.832,
"step": 336
},
{
"epoch": 4.166666666666667,
"grad_norm": 14.644241333007812,
"learning_rate": 4.78615520282187e-05,
"loss": 12.36754150390625,
"step": 350
},
{
"epoch": 4.464285714285714,
"grad_norm": 9.411576271057129,
"learning_rate": 4.731040564373898e-05,
"loss": 12.7791748046875,
"step": 375
},
{
"epoch": 4.761904761904762,
"grad_norm": 14.686434745788574,
"learning_rate": 4.675925925925926e-05,
"loss": 12.952034912109376,
"step": 400
},
{
"epoch": 5.0,
"eval_loss": 7.995185375213623,
"eval_map": 0.1284,
"eval_map_50": 0.1931,
"eval_map_75": 0.1248,
"eval_map_group": 0.0539,
"eval_map_image": 0.2066,
"eval_map_large": 0.2224,
"eval_map_medium": 0.1772,
"eval_map_rectangle": 0.1552,
"eval_map_small": 0.06,
"eval_map_text": 0.0978,
"eval_mar_1": 0.0435,
"eval_mar_10": 0.2236,
"eval_mar_100": 0.4568,
"eval_mar_100_group": 0.4465,
"eval_mar_100_image": 0.558,
"eval_mar_100_rectangle": 0.4388,
"eval_mar_100_text": 0.384,
"eval_mar_large": 0.7415,
"eval_mar_medium": 0.5216,
"eval_mar_small": 0.2552,
"eval_runtime": 17.6743,
"eval_samples_per_second": 6.79,
"eval_steps_per_second": 0.849,
"step": 420
},
{
"epoch": 5.059523809523809,
"grad_norm": 14.872081756591797,
"learning_rate": 4.620811287477954e-05,
"loss": 12.8186669921875,
"step": 425
},
{
"epoch": 5.357142857142857,
"grad_norm": 14.947751998901367,
"learning_rate": 4.5656966490299825e-05,
"loss": 12.421568603515626,
"step": 450
},
{
"epoch": 5.654761904761905,
"grad_norm": 14.387224197387695,
"learning_rate": 4.510582010582011e-05,
"loss": 12.860526123046874,
"step": 475
},
{
"epoch": 5.9523809523809526,
"grad_norm": 11.880175590515137,
"learning_rate": 4.455467372134039e-05,
"loss": 12.392239990234375,
"step": 500
},
{
"epoch": 6.0,
"eval_loss": 7.908458232879639,
"eval_map": 0.123,
"eval_map_50": 0.18,
"eval_map_75": 0.1191,
"eval_map_group": 0.0441,
"eval_map_image": 0.2175,
"eval_map_large": 0.2555,
"eval_map_medium": 0.175,
"eval_map_rectangle": 0.1501,
"eval_map_small": 0.0541,
"eval_map_text": 0.0801,
"eval_mar_1": 0.0432,
"eval_mar_10": 0.2122,
"eval_mar_100": 0.4464,
"eval_mar_100_group": 0.4276,
"eval_mar_100_image": 0.588,
"eval_mar_100_rectangle": 0.4424,
"eval_mar_100_text": 0.3274,
"eval_mar_large": 0.7738,
"eval_mar_medium": 0.5237,
"eval_mar_small": 0.2375,
"eval_runtime": 17.6011,
"eval_samples_per_second": 6.818,
"eval_steps_per_second": 0.852,
"step": 504
},
{
"epoch": 6.25,
"grad_norm": 24.597919464111328,
"learning_rate": 4.400352733686067e-05,
"loss": 12.41754150390625,
"step": 525
},
{
"epoch": 6.5476190476190474,
"grad_norm": 11.277619361877441,
"learning_rate": 4.345238095238096e-05,
"loss": 12.266273193359375,
"step": 550
},
{
"epoch": 6.845238095238095,
"grad_norm": 36.6771354675293,
"learning_rate": 4.290123456790124e-05,
"loss": 11.820712890625,
"step": 575
},
{
"epoch": 7.0,
"eval_loss": 7.85192346572876,
"eval_map": 0.1266,
"eval_map_50": 0.1846,
"eval_map_75": 0.1225,
"eval_map_group": 0.0519,
"eval_map_image": 0.2088,
"eval_map_large": 0.2261,
"eval_map_medium": 0.1858,
"eval_map_rectangle": 0.1674,
"eval_map_small": 0.0571,
"eval_map_text": 0.0785,
"eval_mar_1": 0.043,
"eval_mar_10": 0.2199,
"eval_mar_100": 0.4571,
"eval_mar_100_group": 0.4622,
"eval_mar_100_image": 0.5641,
"eval_mar_100_rectangle": 0.4543,
"eval_mar_100_text": 0.3479,
"eval_mar_large": 0.7813,
"eval_mar_medium": 0.5405,
"eval_mar_small": 0.2422,
"eval_runtime": 17.7165,
"eval_samples_per_second": 6.773,
"eval_steps_per_second": 0.847,
"step": 588
},
{
"epoch": 7.142857142857143,
"grad_norm": 20.0545711517334,
"learning_rate": 4.235008818342152e-05,
"loss": 12.802718505859374,
"step": 600
},
{
"epoch": 7.440476190476191,
"grad_norm": 15.03058910369873,
"learning_rate": 4.17989417989418e-05,
"loss": 12.179630126953125,
"step": 625
},
{
"epoch": 7.738095238095238,
"grad_norm": 7.166962623596191,
"learning_rate": 4.124779541446208e-05,
"loss": 11.9717578125,
"step": 650
},
{
"epoch": 8.0,
"eval_loss": 7.762953281402588,
"eval_map": 0.1389,
"eval_map_50": 0.2028,
"eval_map_75": 0.1353,
"eval_map_group": 0.0552,
"eval_map_image": 0.2193,
"eval_map_large": 0.2511,
"eval_map_medium": 0.197,
"eval_map_rectangle": 0.177,
"eval_map_small": 0.066,
"eval_map_text": 0.1041,
"eval_mar_1": 0.0432,
"eval_mar_10": 0.2317,
"eval_mar_100": 0.4742,
"eval_mar_100_group": 0.4814,
"eval_mar_100_image": 0.5825,
"eval_mar_100_rectangle": 0.4582,
"eval_mar_100_text": 0.3747,
"eval_mar_large": 0.761,
"eval_mar_medium": 0.5526,
"eval_mar_small": 0.2631,
"eval_runtime": 17.6278,
"eval_samples_per_second": 6.807,
"eval_steps_per_second": 0.851,
"step": 672
},
{
"epoch": 8.035714285714286,
"grad_norm": 10.33957290649414,
"learning_rate": 4.0696649029982364e-05,
"loss": 11.71945556640625,
"step": 675
},
{
"epoch": 8.333333333333334,
"grad_norm": 11.141268730163574,
"learning_rate": 4.014550264550265e-05,
"loss": 12.104078369140625,
"step": 700
},
{
"epoch": 8.630952380952381,
"grad_norm": 8.50006103515625,
"learning_rate": 3.959435626102293e-05,
"loss": 12.10205322265625,
"step": 725
},
{
"epoch": 8.928571428571429,
"grad_norm": 12.563775062561035,
"learning_rate": 3.904320987654321e-05,
"loss": 11.8281201171875,
"step": 750
},
{
"epoch": 9.0,
"eval_loss": 7.723588466644287,
"eval_map": 0.1317,
"eval_map_50": 0.186,
"eval_map_75": 0.1298,
"eval_map_group": 0.0554,
"eval_map_image": 0.2197,
"eval_map_large": 0.2512,
"eval_map_medium": 0.1925,
"eval_map_rectangle": 0.1777,
"eval_map_small": 0.0607,
"eval_map_text": 0.0739,
"eval_mar_1": 0.043,
"eval_mar_10": 0.2254,
"eval_mar_100": 0.4671,
"eval_mar_100_group": 0.4945,
"eval_mar_100_image": 0.5935,
"eval_mar_100_rectangle": 0.4602,
"eval_mar_100_text": 0.32,
"eval_mar_large": 0.7856,
"eval_mar_medium": 0.5534,
"eval_mar_small": 0.2523,
"eval_runtime": 17.6923,
"eval_samples_per_second": 6.783,
"eval_steps_per_second": 0.848,
"step": 756
},
{
"epoch": 9.226190476190476,
"grad_norm": 15.43812370300293,
"learning_rate": 3.84920634920635e-05,
"loss": 11.588818359375,
"step": 775
},
{
"epoch": 9.523809523809524,
"grad_norm": 33.323638916015625,
"learning_rate": 3.794091710758378e-05,
"loss": 11.961219482421875,
"step": 800
},
{
"epoch": 9.821428571428571,
"grad_norm": 23.318923950195312,
"learning_rate": 3.7389770723104055e-05,
"loss": 11.79553955078125,
"step": 825
},
{
"epoch": 10.0,
"eval_loss": 7.686021327972412,
"eval_map": 0.1389,
"eval_map_50": 0.1989,
"eval_map_75": 0.1353,
"eval_map_group": 0.0604,
"eval_map_image": 0.2208,
"eval_map_large": 0.2665,
"eval_map_medium": 0.201,
"eval_map_rectangle": 0.1802,
"eval_map_small": 0.0685,
"eval_map_text": 0.0944,
"eval_mar_1": 0.0451,
"eval_mar_10": 0.2327,
"eval_mar_100": 0.4797,
"eval_mar_100_group": 0.506,
"eval_mar_100_image": 0.5941,
"eval_mar_100_rectangle": 0.4599,
"eval_mar_100_text": 0.359,
"eval_mar_large": 0.8045,
"eval_mar_medium": 0.568,
"eval_mar_small": 0.2633,
"eval_runtime": 17.2713,
"eval_samples_per_second": 6.948,
"eval_steps_per_second": 0.868,
"step": 840
},
{
"epoch": 10.119047619047619,
"grad_norm": 21.750789642333984,
"learning_rate": 3.683862433862434e-05,
"loss": 11.53491943359375,
"step": 850
},
{
"epoch": 10.416666666666666,
"grad_norm": 12.808640480041504,
"learning_rate": 3.628747795414462e-05,
"loss": 11.769920654296875,
"step": 875
},
{
"epoch": 10.714285714285714,
"grad_norm": 12.053258895874023,
"learning_rate": 3.5736331569664904e-05,
"loss": 11.778603515625,
"step": 900
},
{
"epoch": 11.0,
"eval_loss": 7.661948204040527,
"eval_map": 0.1456,
"eval_map_50": 0.2122,
"eval_map_75": 0.1431,
"eval_map_group": 0.0649,
"eval_map_image": 0.2302,
"eval_map_large": 0.2814,
"eval_map_medium": 0.2067,
"eval_map_rectangle": 0.1763,
"eval_map_small": 0.0763,
"eval_map_text": 0.1111,
"eval_mar_1": 0.0445,
"eval_mar_10": 0.2399,
"eval_mar_100": 0.4916,
"eval_mar_100_group": 0.5073,
"eval_mar_100_image": 0.6075,
"eval_mar_100_rectangle": 0.4558,
"eval_mar_100_text": 0.3955,
"eval_mar_large": 0.7904,
"eval_mar_medium": 0.5715,
"eval_mar_small": 0.2859,
"eval_runtime": 17.5947,
"eval_samples_per_second": 6.82,
"eval_steps_per_second": 0.853,
"step": 924
},
{
"epoch": 11.011904761904763,
"grad_norm": 10.250245094299316,
"learning_rate": 3.518518518518519e-05,
"loss": 12.0818701171875,
"step": 925
},
{
"epoch": 11.30952380952381,
"grad_norm": 28.736799240112305,
"learning_rate": 3.463403880070547e-05,
"loss": 11.669910888671875,
"step": 950
},
{
"epoch": 11.607142857142858,
"grad_norm": 32.78938293457031,
"learning_rate": 3.408289241622575e-05,
"loss": 11.80724853515625,
"step": 975
},
{
"epoch": 11.904761904761905,
"grad_norm": 14.23360824584961,
"learning_rate": 3.353174603174603e-05,
"loss": 11.633603515625,
"step": 1000
},
{
"epoch": 12.0,
"eval_loss": 7.61862325668335,
"eval_map": 0.1452,
"eval_map_50": 0.2085,
"eval_map_75": 0.1433,
"eval_map_group": 0.0673,
"eval_map_image": 0.2435,
"eval_map_large": 0.2939,
"eval_map_medium": 0.2001,
"eval_map_rectangle": 0.1766,
"eval_map_small": 0.0697,
"eval_map_text": 0.0935,
"eval_mar_1": 0.0465,
"eval_mar_10": 0.2467,
"eval_mar_100": 0.4811,
"eval_mar_100_group": 0.5087,
"eval_mar_100_image": 0.6105,
"eval_mar_100_rectangle": 0.4571,
"eval_mar_100_text": 0.3481,
"eval_mar_large": 0.7906,
"eval_mar_medium": 0.5665,
"eval_mar_small": 0.273,
"eval_runtime": 17.2166,
"eval_samples_per_second": 6.97,
"eval_steps_per_second": 0.871,
"step": 1008
},
{
"epoch": 12.202380952380953,
"grad_norm": 56.82817840576172,
"learning_rate": 3.2980599647266316e-05,
"loss": 11.56818603515625,
"step": 1025
},
{
"epoch": 12.5,
"grad_norm": 21.191076278686523,
"learning_rate": 3.2429453262786595e-05,
"loss": 11.379783935546875,
"step": 1050
},
{
"epoch": 12.797619047619047,
"grad_norm": 48.60485076904297,
"learning_rate": 3.187830687830688e-05,
"loss": 11.505103759765625,
"step": 1075
},
{
"epoch": 13.0,
"eval_loss": 7.647872447967529,
"eval_map": 0.1367,
"eval_map_50": 0.1944,
"eval_map_75": 0.1348,
"eval_map_group": 0.0665,
"eval_map_image": 0.2308,
"eval_map_large": 0.2837,
"eval_map_medium": 0.201,
"eval_map_rectangle": 0.1654,
"eval_map_small": 0.0661,
"eval_map_text": 0.0839,
"eval_mar_1": 0.0456,
"eval_mar_10": 0.2329,
"eval_mar_100": 0.4692,
"eval_mar_100_group": 0.5001,
"eval_mar_100_image": 0.5892,
"eval_mar_100_rectangle": 0.4515,
"eval_mar_100_text": 0.3359,
"eval_mar_large": 0.7908,
"eval_mar_medium": 0.5583,
"eval_mar_small": 0.2496,
"eval_runtime": 17.2376,
"eval_samples_per_second": 6.962,
"eval_steps_per_second": 0.87,
"step": 1092
},
{
"epoch": 13.095238095238095,
"grad_norm": 33.55184555053711,
"learning_rate": 3.1327160493827165e-05,
"loss": 11.693345947265625,
"step": 1100
},
{
"epoch": 13.392857142857142,
"grad_norm": 10.436372756958008,
"learning_rate": 3.0776014109347444e-05,
"loss": 11.6796435546875,
"step": 1125
},
{
"epoch": 13.69047619047619,
"grad_norm": 18.514883041381836,
"learning_rate": 3.022486772486773e-05,
"loss": 11.21151611328125,
"step": 1150
},
{
"epoch": 13.988095238095237,
"grad_norm": 10.212114334106445,
"learning_rate": 2.967372134038801e-05,
"loss": 11.76965576171875,
"step": 1175
},
{
"epoch": 14.0,
"eval_loss": 7.631312370300293,
"eval_map": 0.1524,
"eval_map_50": 0.2192,
"eval_map_75": 0.15,
"eval_map_group": 0.0797,
"eval_map_image": 0.2496,
"eval_map_large": 0.2967,
"eval_map_medium": 0.213,
"eval_map_rectangle": 0.1743,
"eval_map_small": 0.0767,
"eval_map_text": 0.1059,
"eval_mar_1": 0.0481,
"eval_mar_10": 0.2508,
"eval_mar_100": 0.4937,
"eval_mar_100_group": 0.5226,
"eval_mar_100_image": 0.6179,
"eval_mar_100_rectangle": 0.4513,
"eval_mar_100_text": 0.383,
"eval_mar_large": 0.7914,
"eval_mar_medium": 0.578,
"eval_mar_small": 0.2893,
"eval_runtime": 17.3512,
"eval_samples_per_second": 6.916,
"eval_steps_per_second": 0.864,
"step": 1176
},
{
"epoch": 14.285714285714286,
"grad_norm": 35.04265213012695,
"learning_rate": 2.912257495590829e-05,
"loss": 11.576104736328125,
"step": 1200
},
{
"epoch": 14.583333333333334,
"grad_norm": 32.3892822265625,
"learning_rate": 2.857142857142857e-05,
"loss": 11.311810302734376,
"step": 1225
},
{
"epoch": 14.880952380952381,
"grad_norm": 44.537376403808594,
"learning_rate": 2.8020282186948853e-05,
"loss": 11.286588134765625,
"step": 1250
},
{
"epoch": 15.0,
"eval_loss": 7.60584020614624,
"eval_map": 0.1544,
"eval_map_50": 0.2216,
"eval_map_75": 0.1513,
"eval_map_group": 0.0822,
"eval_map_image": 0.249,
"eval_map_large": 0.3055,
"eval_map_medium": 0.2181,
"eval_map_rectangle": 0.1776,
"eval_map_small": 0.0816,
"eval_map_text": 0.1087,
"eval_mar_1": 0.0505,
"eval_mar_10": 0.2549,
"eval_mar_100": 0.4948,
"eval_mar_100_group": 0.5213,
"eval_mar_100_image": 0.6152,
"eval_mar_100_rectangle": 0.4614,
"eval_mar_100_text": 0.3812,
"eval_mar_large": 0.7931,
"eval_mar_medium": 0.5825,
"eval_mar_small": 0.2855,
"eval_runtime": 17.0444,
"eval_samples_per_second": 7.04,
"eval_steps_per_second": 0.88,
"step": 1260
},
{
"epoch": 15.178571428571429,
"grad_norm": 10.380435943603516,
"learning_rate": 2.7469135802469138e-05,
"loss": 11.5576953125,
"step": 1275
},
{
"epoch": 15.476190476190476,
"grad_norm": 12.200742721557617,
"learning_rate": 2.691798941798942e-05,
"loss": 11.44038818359375,
"step": 1300
},
{
"epoch": 15.773809523809524,
"grad_norm": 9.843217849731445,
"learning_rate": 2.63668430335097e-05,
"loss": 10.94272216796875,
"step": 1325
},
{
"epoch": 16.0,
"eval_loss": 7.568243980407715,
"eval_map": 0.1551,
"eval_map_50": 0.2188,
"eval_map_75": 0.1537,
"eval_map_group": 0.0866,
"eval_map_image": 0.2522,
"eval_map_large": 0.3189,
"eval_map_medium": 0.2193,
"eval_map_rectangle": 0.1842,
"eval_map_small": 0.0744,
"eval_map_text": 0.0972,
"eval_mar_1": 0.0513,
"eval_mar_10": 0.2521,
"eval_mar_100": 0.4899,
"eval_mar_100_group": 0.5293,
"eval_mar_100_image": 0.613,
"eval_mar_100_rectangle": 0.462,
"eval_mar_100_text": 0.3554,
"eval_mar_large": 0.8016,
"eval_mar_medium": 0.5813,
"eval_mar_small": 0.2772,
"eval_runtime": 17.2934,
"eval_samples_per_second": 6.939,
"eval_steps_per_second": 0.867,
"step": 1344
},
{
"epoch": 16.071428571428573,
"grad_norm": 11.41603946685791,
"learning_rate": 2.5815696649029987e-05,
"loss": 11.799744873046874,
"step": 1350
},
{
"epoch": 16.36904761904762,
"grad_norm": 10.90324878692627,
"learning_rate": 2.526455026455027e-05,
"loss": 11.40052978515625,
"step": 1375
},
{
"epoch": 16.666666666666668,
"grad_norm": 19.619916915893555,
"learning_rate": 2.471340388007055e-05,
"loss": 11.129947509765625,
"step": 1400
},
{
"epoch": 16.964285714285715,
"grad_norm": 20.032556533813477,
"learning_rate": 2.416225749559083e-05,
"loss": 11.209949951171875,
"step": 1425
},
{
"epoch": 17.0,
"eval_loss": 7.599889755249023,
"eval_map": 0.1529,
"eval_map_50": 0.2152,
"eval_map_75": 0.1518,
"eval_map_group": 0.0918,
"eval_map_image": 0.2452,
"eval_map_large": 0.3052,
"eval_map_medium": 0.2287,
"eval_map_rectangle": 0.1826,
"eval_map_small": 0.0755,
"eval_map_text": 0.0918,
"eval_mar_1": 0.0481,
"eval_mar_10": 0.2552,
"eval_mar_100": 0.4926,
"eval_mar_100_group": 0.5311,
"eval_mar_100_image": 0.6157,
"eval_mar_100_rectangle": 0.4654,
"eval_mar_100_text": 0.3585,
"eval_mar_large": 0.8057,
"eval_mar_medium": 0.5831,
"eval_mar_small": 0.2802,
"eval_runtime": 17.1006,
"eval_samples_per_second": 7.017,
"eval_steps_per_second": 0.877,
"step": 1428
},
{
"epoch": 17.261904761904763,
"grad_norm": 17.242233276367188,
"learning_rate": 2.361111111111111e-05,
"loss": 11.399385986328125,
"step": 1450
},
{
"epoch": 17.55952380952381,
"grad_norm": 13.795672416687012,
"learning_rate": 2.3059964726631396e-05,
"loss": 11.417935791015625,
"step": 1475
},
{
"epoch": 17.857142857142858,
"grad_norm": 11.631192207336426,
"learning_rate": 2.2508818342151678e-05,
"loss": 10.97098876953125,
"step": 1500
},
{
"epoch": 18.0,
"eval_loss": 7.5713324546813965,
"eval_map": 0.1577,
"eval_map_50": 0.2227,
"eval_map_75": 0.157,
"eval_map_group": 0.0951,
"eval_map_image": 0.2512,
"eval_map_large": 0.3131,
"eval_map_medium": 0.229,
"eval_map_rectangle": 0.1884,
"eval_map_small": 0.0815,
"eval_map_text": 0.0962,
"eval_mar_1": 0.0494,
"eval_mar_10": 0.2586,
"eval_mar_100": 0.4979,
"eval_mar_100_group": 0.5349,
"eval_mar_100_image": 0.6257,
"eval_mar_100_rectangle": 0.4593,
"eval_mar_100_text": 0.3716,
"eval_mar_large": 0.7908,
"eval_mar_medium": 0.5872,
"eval_mar_small": 0.292,
"eval_runtime": 17.1925,
"eval_samples_per_second": 6.98,
"eval_steps_per_second": 0.872,
"step": 1512
},
{
"epoch": 18.154761904761905,
"grad_norm": 15.366872787475586,
"learning_rate": 2.1957671957671956e-05,
"loss": 11.15154541015625,
"step": 1525
},
{
"epoch": 18.452380952380953,
"grad_norm": 13.13204574584961,
"learning_rate": 2.140652557319224e-05,
"loss": 11.403101806640626,
"step": 1550
},
{
"epoch": 18.75,
"grad_norm": 16.772144317626953,
"learning_rate": 2.0855379188712523e-05,
"loss": 11.07358642578125,
"step": 1575
},
{
"epoch": 19.0,
"eval_loss": 7.5920186042785645,
"eval_map": 0.1568,
"eval_map_50": 0.2195,
"eval_map_75": 0.1546,
"eval_map_group": 0.0976,
"eval_map_image": 0.2611,
"eval_map_large": 0.3225,
"eval_map_medium": 0.224,
"eval_map_rectangle": 0.1799,
"eval_map_small": 0.0759,
"eval_map_text": 0.0887,
"eval_mar_1": 0.0493,
"eval_mar_10": 0.2589,
"eval_mar_100": 0.4901,
"eval_mar_100_group": 0.5376,
"eval_mar_100_image": 0.626,
"eval_mar_100_rectangle": 0.454,
"eval_mar_100_text": 0.3428,
"eval_mar_large": 0.7912,
"eval_mar_medium": 0.584,
"eval_mar_small": 0.2779,
"eval_runtime": 16.8456,
"eval_samples_per_second": 7.124,
"eval_steps_per_second": 0.89,
"step": 1596
},
{
"epoch": 19.047619047619047,
"grad_norm": 72.46932220458984,
"learning_rate": 2.0304232804232805e-05,
"loss": 11.312447509765626,
"step": 1600
},
{
"epoch": 19.345238095238095,
"grad_norm": 7.8378801345825195,
"learning_rate": 1.9753086419753087e-05,
"loss": 11.34157470703125,
"step": 1625
},
{
"epoch": 19.642857142857142,
"grad_norm": 14.729228973388672,
"learning_rate": 1.920194003527337e-05,
"loss": 11.0249072265625,
"step": 1650
},
{
"epoch": 19.94047619047619,
"grad_norm": 25.199474334716797,
"learning_rate": 1.8650793650793654e-05,
"loss": 11.217237548828125,
"step": 1675
},
{
"epoch": 20.0,
"eval_loss": 7.561267375946045,
"eval_map": 0.153,
"eval_map_50": 0.2144,
"eval_map_75": 0.1502,
"eval_map_group": 0.0971,
"eval_map_image": 0.2449,
"eval_map_large": 0.3227,
"eval_map_medium": 0.2256,
"eval_map_rectangle": 0.1846,
"eval_map_small": 0.0726,
"eval_map_text": 0.0851,
"eval_mar_1": 0.0499,
"eval_mar_10": 0.2533,
"eval_mar_100": 0.4867,
"eval_mar_100_group": 0.5363,
"eval_mar_100_image": 0.6226,
"eval_mar_100_rectangle": 0.4526,
"eval_mar_100_text": 0.3353,
"eval_mar_large": 0.8013,
"eval_mar_medium": 0.5825,
"eval_mar_small": 0.2765,
"eval_runtime": 17.0648,
"eval_samples_per_second": 7.032,
"eval_steps_per_second": 0.879,
"step": 1680
},
{
"epoch": 20.238095238095237,
"grad_norm": 21.974050521850586,
"learning_rate": 1.8099647266313935e-05,
"loss": 10.83720458984375,
"step": 1700
},
{
"epoch": 20.535714285714285,
"grad_norm": 23.206058502197266,
"learning_rate": 1.7548500881834214e-05,
"loss": 11.15845458984375,
"step": 1725
},
{
"epoch": 20.833333333333332,
"grad_norm": 29.534772872924805,
"learning_rate": 1.69973544973545e-05,
"loss": 11.038966064453126,
"step": 1750
},
{
"epoch": 21.0,
"eval_loss": 7.541738986968994,
"eval_map": 0.1597,
"eval_map_50": 0.2248,
"eval_map_75": 0.1578,
"eval_map_group": 0.106,
"eval_map_image": 0.2534,
"eval_map_large": 0.3218,
"eval_map_medium": 0.2289,
"eval_map_rectangle": 0.1851,
"eval_map_small": 0.0824,
"eval_map_text": 0.0945,
"eval_mar_1": 0.0504,
"eval_mar_10": 0.262,
"eval_mar_100": 0.4959,
"eval_mar_100_group": 0.5397,
"eval_mar_100_image": 0.6128,
"eval_mar_100_rectangle": 0.4607,
"eval_mar_100_text": 0.3705,
"eval_mar_large": 0.7954,
"eval_mar_medium": 0.5832,
"eval_mar_small": 0.2921,
"eval_runtime": 17.3182,
"eval_samples_per_second": 6.929,
"eval_steps_per_second": 0.866,
"step": 1764
},
{
"epoch": 21.13095238095238,
"grad_norm": 12.260412216186523,
"learning_rate": 1.644620811287478e-05,
"loss": 11.237501220703125,
"step": 1775
},
{
"epoch": 21.428571428571427,
"grad_norm": 8.769057273864746,
"learning_rate": 1.5895061728395063e-05,
"loss": 11.0806005859375,
"step": 1800
},
{
"epoch": 21.726190476190474,
"grad_norm": 22.044124603271484,
"learning_rate": 1.5343915343915344e-05,
"loss": 11.148917236328124,
"step": 1825
},
{
"epoch": 22.0,
"eval_loss": 7.538026332855225,
"eval_map": 0.1599,
"eval_map_50": 0.224,
"eval_map_75": 0.1569,
"eval_map_group": 0.1082,
"eval_map_image": 0.2527,
"eval_map_large": 0.3306,
"eval_map_medium": 0.2308,
"eval_map_rectangle": 0.1882,
"eval_map_small": 0.0783,
"eval_map_text": 0.0906,
"eval_mar_1": 0.0488,
"eval_mar_10": 0.2596,
"eval_mar_100": 0.4935,
"eval_mar_100_group": 0.5411,
"eval_mar_100_image": 0.6195,
"eval_mar_100_rectangle": 0.4587,
"eval_mar_100_text": 0.3547,
"eval_mar_large": 0.7987,
"eval_mar_medium": 0.5876,
"eval_mar_small": 0.2839,
"eval_runtime": 17.1176,
"eval_samples_per_second": 7.01,
"eval_steps_per_second": 0.876,
"step": 1848
},
{
"epoch": 22.023809523809526,
"grad_norm": 10.205994606018066,
"learning_rate": 1.4792768959435626e-05,
"loss": 11.559412841796876,
"step": 1850
},
{
"epoch": 22.321428571428573,
"grad_norm": 7.906266212463379,
"learning_rate": 1.424162257495591e-05,
"loss": 11.12822998046875,
"step": 1875
},
{
"epoch": 22.61904761904762,
"grad_norm": 10.976849555969238,
"learning_rate": 1.3690476190476192e-05,
"loss": 10.973646240234375,
"step": 1900
},
{
"epoch": 22.916666666666668,
"grad_norm": 17.911376953125,
"learning_rate": 1.3139329805996472e-05,
"loss": 10.757685546875,
"step": 1925
},
{
"epoch": 23.0,
"eval_loss": 7.559340000152588,
"eval_map": 0.166,
"eval_map_50": 0.2341,
"eval_map_75": 0.1648,
"eval_map_group": 0.1148,
"eval_map_image": 0.2532,
"eval_map_large": 0.338,
"eval_map_medium": 0.2406,
"eval_map_rectangle": 0.1871,
"eval_map_small": 0.0849,
"eval_map_text": 0.1089,
"eval_mar_1": 0.0511,
"eval_mar_10": 0.2643,
"eval_mar_100": 0.4993,
"eval_mar_100_group": 0.5457,
"eval_mar_100_image": 0.6141,
"eval_mar_100_rectangle": 0.459,
"eval_mar_100_text": 0.3786,
"eval_mar_large": 0.797,
"eval_mar_medium": 0.5937,
"eval_mar_small": 0.2892,
"eval_runtime": 17.1914,
"eval_samples_per_second": 6.98,
"eval_steps_per_second": 0.873,
"step": 1932
},
{
"epoch": 23.214285714285715,
"grad_norm": 14.862859725952148,
"learning_rate": 1.2588183421516755e-05,
"loss": 11.2927880859375,
"step": 1950
},
{
"epoch": 23.511904761904763,
"grad_norm": 25.316974639892578,
"learning_rate": 1.2037037037037037e-05,
"loss": 10.83892822265625,
"step": 1975
},
{
"epoch": 23.80952380952381,
"grad_norm": 12.938018798828125,
"learning_rate": 1.148589065255732e-05,
"loss": 10.78179443359375,
"step": 2000
},
{
"epoch": 24.0,
"eval_loss": 7.54477071762085,
"eval_map": 0.1581,
"eval_map_50": 0.2225,
"eval_map_75": 0.1557,
"eval_map_group": 0.1103,
"eval_map_image": 0.2517,
"eval_map_large": 0.3262,
"eval_map_medium": 0.2326,
"eval_map_rectangle": 0.1766,
"eval_map_small": 0.0781,
"eval_map_text": 0.0938,
"eval_mar_1": 0.0503,
"eval_mar_10": 0.2565,
"eval_mar_100": 0.487,
"eval_mar_100_group": 0.5407,
"eval_mar_100_image": 0.6067,
"eval_mar_100_rectangle": 0.4519,
"eval_mar_100_text": 0.3487,
"eval_mar_large": 0.7977,
"eval_mar_medium": 0.5912,
"eval_mar_small": 0.2677,
"eval_runtime": 17.1291,
"eval_samples_per_second": 7.006,
"eval_steps_per_second": 0.876,
"step": 2016
},
{
"epoch": 24.107142857142858,
"grad_norm": 43.82697296142578,
"learning_rate": 1.09347442680776e-05,
"loss": 11.097623291015625,
"step": 2025
},
{
"epoch": 24.404761904761905,
"grad_norm": 11.796220779418945,
"learning_rate": 1.0383597883597884e-05,
"loss": 11.050999755859374,
"step": 2050
},
{
"epoch": 24.702380952380953,
"grad_norm": 11.335160255432129,
"learning_rate": 9.832451499118166e-06,
"loss": 10.86746337890625,
"step": 2075
},
{
"epoch": 25.0,
"grad_norm": 18.967390060424805,
"learning_rate": 9.281305114638448e-06,
"loss": 11.010140380859376,
"step": 2100
},
{
"epoch": 25.0,
"eval_loss": 7.518592357635498,
"eval_map": 0.157,
"eval_map_50": 0.2216,
"eval_map_75": 0.1558,
"eval_map_group": 0.1064,
"eval_map_image": 0.2477,
"eval_map_large": 0.3241,
"eval_map_medium": 0.2348,
"eval_map_rectangle": 0.1756,
"eval_map_small": 0.0802,
"eval_map_text": 0.0982,
"eval_mar_1": 0.0505,
"eval_mar_10": 0.2569,
"eval_mar_100": 0.491,
"eval_mar_100_group": 0.5458,
"eval_mar_100_image": 0.6013,
"eval_mar_100_rectangle": 0.4526,
"eval_mar_100_text": 0.3644,
"eval_mar_large": 0.7955,
"eval_mar_medium": 0.5906,
"eval_mar_small": 0.2748,
"eval_runtime": 17.0932,
"eval_samples_per_second": 7.02,
"eval_steps_per_second": 0.878,
"step": 2100
},
{
"epoch": 25.297619047619047,
"grad_norm": 7.750904560089111,
"learning_rate": 8.73015873015873e-06,
"loss": 10.79929931640625,
"step": 2125
},
{
"epoch": 25.595238095238095,
"grad_norm": 16.77004623413086,
"learning_rate": 8.179012345679013e-06,
"loss": 10.849818115234376,
"step": 2150
},
{
"epoch": 25.892857142857142,
"grad_norm": 20.175363540649414,
"learning_rate": 7.627865961199294e-06,
"loss": 10.97046875,
"step": 2175
},
{
"epoch": 26.0,
"eval_loss": 7.568551063537598,
"eval_map": 0.1583,
"eval_map_50": 0.223,
"eval_map_75": 0.1563,
"eval_map_group": 0.11,
"eval_map_image": 0.2487,
"eval_map_large": 0.3214,
"eval_map_medium": 0.2397,
"eval_map_rectangle": 0.1804,
"eval_map_small": 0.0796,
"eval_map_text": 0.0943,
"eval_mar_1": 0.0487,
"eval_mar_10": 0.2593,
"eval_mar_100": 0.4941,
"eval_mar_100_group": 0.5395,
"eval_mar_100_image": 0.6193,
"eval_mar_100_rectangle": 0.4503,
"eval_mar_100_text": 0.3673,
"eval_mar_large": 0.7885,
"eval_mar_medium": 0.5822,
"eval_mar_small": 0.2876,
"eval_runtime": 17.4348,
"eval_samples_per_second": 6.883,
"eval_steps_per_second": 0.86,
"step": 2184
},
{
"epoch": 26.19047619047619,
"grad_norm": 11.11269474029541,
"learning_rate": 7.076719576719578e-06,
"loss": 11.78640869140625,
"step": 2200
},
{
"epoch": 26.488095238095237,
"grad_norm": 8.43252182006836,
"learning_rate": 6.5255731922398585e-06,
"loss": 10.9381396484375,
"step": 2225
},
{
"epoch": 26.785714285714285,
"grad_norm": 36.22833251953125,
"learning_rate": 5.974426807760141e-06,
"loss": 10.875008544921876,
"step": 2250
},
{
"epoch": 27.0,
"eval_loss": 7.555787086486816,
"eval_map": 0.1569,
"eval_map_50": 0.2185,
"eval_map_75": 0.1555,
"eval_map_group": 0.1083,
"eval_map_image": 0.2464,
"eval_map_large": 0.3315,
"eval_map_medium": 0.2371,
"eval_map_rectangle": 0.185,
"eval_map_small": 0.0771,
"eval_map_text": 0.0878,
"eval_mar_1": 0.0503,
"eval_mar_10": 0.2538,
"eval_mar_100": 0.4833,
"eval_mar_100_group": 0.5389,
"eval_mar_100_image": 0.6003,
"eval_mar_100_rectangle": 0.4565,
"eval_mar_100_text": 0.3373,
"eval_mar_large": 0.8099,
"eval_mar_medium": 0.5862,
"eval_mar_small": 0.2609,
"eval_runtime": 17.0991,
"eval_samples_per_second": 7.018,
"eval_steps_per_second": 0.877,
"step": 2268
},
{
"epoch": 27.083333333333332,
"grad_norm": 12.785017013549805,
"learning_rate": 5.423280423280423e-06,
"loss": 10.976126708984374,
"step": 2275
},
{
"epoch": 27.38095238095238,
"grad_norm": 12.107715606689453,
"learning_rate": 4.872134038800706e-06,
"loss": 11.2352734375,
"step": 2300
},
{
"epoch": 27.678571428571427,
"grad_norm": 9.639913558959961,
"learning_rate": 4.3209876543209875e-06,
"loss": 11.119185791015624,
"step": 2325
},
{
"epoch": 27.976190476190474,
"grad_norm": 16.351314544677734,
"learning_rate": 3.7698412698412697e-06,
"loss": 10.751910400390624,
"step": 2350
},
{
"epoch": 28.0,
"eval_loss": 7.529413223266602,
"eval_map": 0.1604,
"eval_map_50": 0.224,
"eval_map_75": 0.1595,
"eval_map_group": 0.1109,
"eval_map_image": 0.2489,
"eval_map_large": 0.3368,
"eval_map_medium": 0.2361,
"eval_map_rectangle": 0.185,
"eval_map_small": 0.0825,
"eval_map_text": 0.0968,
"eval_mar_1": 0.0499,
"eval_mar_10": 0.2565,
"eval_mar_100": 0.4946,
"eval_mar_100_group": 0.5448,
"eval_mar_100_image": 0.6148,
"eval_mar_100_rectangle": 0.4581,
"eval_mar_100_text": 0.3608,
"eval_mar_large": 0.7961,
"eval_mar_medium": 0.5903,
"eval_mar_small": 0.2794,
"eval_runtime": 16.8513,
"eval_samples_per_second": 7.121,
"eval_steps_per_second": 0.89,
"step": 2352
},
{
"epoch": 28.273809523809526,
"grad_norm": 12.011247634887695,
"learning_rate": 3.218694885361552e-06,
"loss": 10.890947265625,
"step": 2375
},
{
"epoch": 28.571428571428573,
"grad_norm": 11.049712181091309,
"learning_rate": 2.667548500881834e-06,
"loss": 11.18763427734375,
"step": 2400
},
{
"epoch": 28.86904761904762,
"grad_norm": 16.048168182373047,
"learning_rate": 2.1164021164021164e-06,
"loss": 10.83516845703125,
"step": 2425
},
{
"epoch": 29.0,
"eval_loss": 7.537722587585449,
"eval_map": 0.1657,
"eval_map_50": 0.2311,
"eval_map_75": 0.1638,
"eval_map_group": 0.1191,
"eval_map_image": 0.2592,
"eval_map_large": 0.3459,
"eval_map_medium": 0.2391,
"eval_map_rectangle": 0.1949,
"eval_map_small": 0.0816,
"eval_map_text": 0.0897,
"eval_mar_1": 0.0509,
"eval_mar_10": 0.2653,
"eval_mar_100": 0.4975,
"eval_mar_100_group": 0.5462,
"eval_mar_100_image": 0.6243,
"eval_mar_100_rectangle": 0.4592,
"eval_mar_100_text": 0.3604,
"eval_mar_large": 0.8001,
"eval_mar_medium": 0.59,
"eval_mar_small": 0.2873,
"eval_runtime": 17.2358,
"eval_samples_per_second": 6.962,
"eval_steps_per_second": 0.87,
"step": 2436
},
{
"epoch": 29.166666666666668,
"grad_norm": 11.860162734985352,
"learning_rate": 1.5652557319223987e-06,
"loss": 11.17228515625,
"step": 2450
},
{
"epoch": 29.464285714285715,
"grad_norm": 17.13775062561035,
"learning_rate": 1.0141093474426807e-06,
"loss": 11.1133935546875,
"step": 2475
},
{
"epoch": 29.761904761904763,
"grad_norm": 15.238182067871094,
"learning_rate": 4.6296296296296297e-07,
"loss": 10.7695751953125,
"step": 2500
},
{
"epoch": 30.0,
"eval_loss": 7.532639503479004,
"eval_map": 0.1598,
"eval_map_50": 0.2221,
"eval_map_75": 0.1589,
"eval_map_group": 0.1105,
"eval_map_image": 0.2522,
"eval_map_large": 0.3415,
"eval_map_medium": 0.2358,
"eval_map_rectangle": 0.1861,
"eval_map_small": 0.0793,
"eval_map_text": 0.0905,
"eval_mar_1": 0.0499,
"eval_mar_10": 0.2576,
"eval_mar_100": 0.4897,
"eval_mar_100_group": 0.5388,
"eval_mar_100_image": 0.6149,
"eval_mar_100_rectangle": 0.4594,
"eval_mar_100_text": 0.3458,
"eval_mar_large": 0.8065,
"eval_mar_medium": 0.5899,
"eval_mar_small": 0.2705,
"eval_runtime": 16.9062,
"eval_samples_per_second": 7.098,
"eval_steps_per_second": 0.887,
"step": 2520
},
{
"epoch": 30.0,
"step": 2520,
"total_flos": 2.95288818499584e+18,
"train_loss": 12.287477396404933,
"train_runtime": 1674.3599,
"train_samples_per_second": 11.915,
"train_steps_per_second": 1.505
}
],
"logging_steps": 25,
"max_steps": 2520,
"num_input_tokens_seen": 0,
"num_train_epochs": 30,
"save_steps": 500,
"stateful_callbacks": {
"TrainerControl": {
"args": {
"should_epoch_stop": false,
"should_evaluate": false,
"should_log": false,
"should_save": true,
"should_training_stop": true
},
"attributes": {}
}
},
"total_flos": 2.95288818499584e+18,
"train_batch_size": 8,
"trial_name": null,
"trial_params": null
}