openllama-3b-crt_2 / training_logs.json
joshuaclymer's picture
Upload folder using huggingface_hub
a430ab5
[
{
"loss": 0.7219,
"learning_rate": 7.2e-05,
"epoch": 0.05,
"step": 1
},
{
"loss": 0.7128,
"learning_rate": 7.2e-05,
"epoch": 0.11,
"step": 2
},
{
"loss": 0.6783,
"learning_rate": 7.2e-05,
"epoch": 0.16,
"step": 3
},
{
"loss": 0.6587,
"learning_rate": 7.2e-05,
"epoch": 0.21,
"step": 4
},
{
"loss": 0.638,
"learning_rate": 7.2e-05,
"epoch": 0.26,
"step": 5
},
{
"loss": 0.6147,
"learning_rate": 7.2e-05,
"epoch": 0.32,
"step": 6
},
{
"loss": 0.6024,
"learning_rate": 7.2e-05,
"epoch": 0.37,
"step": 7
},
{
"loss": 0.5924,
"learning_rate": 7.2e-05,
"epoch": 0.42,
"step": 8
},
{
"loss": 0.5868,
"learning_rate": 7.2e-05,
"epoch": 0.47,
"step": 9
},
{
"loss": 0.5416,
"learning_rate": 7.2e-05,
"epoch": 0.53,
"step": 10
},
{
"loss": 0.4749,
"learning_rate": 7.2e-05,
"epoch": 0.58,
"step": 11
},
{
"loss": 0.4541,
"learning_rate": 7.2e-05,
"epoch": 0.63,
"step": 12
},
{
"loss": 0.4036,
"learning_rate": 7.2e-05,
"epoch": 0.68,
"step": 13
},
{
"loss": 0.3915,
"learning_rate": 7.2e-05,
"epoch": 0.74,
"step": 14
},
{
"loss": 0.3465,
"learning_rate": 7.2e-05,
"epoch": 0.79,
"step": 15
},
{
"loss": 0.2573,
"learning_rate": 7.2e-05,
"epoch": 0.84,
"step": 16
},
{
"loss": 0.1707,
"learning_rate": 7.2e-05,
"epoch": 0.89,
"step": 17
},
{
"loss": 0.125,
"learning_rate": 7.2e-05,
"epoch": 0.95,
"step": 18
},
{
"loss": 0.0749,
"learning_rate": 7.2e-05,
"epoch": 1.0,
"step": 19
},
{
"loss": 0.0452,
"learning_rate": 7.2e-05,
"epoch": 1.05,
"step": 20
},
{
"loss": 0.0257,
"learning_rate": 7.2e-05,
"epoch": 1.11,
"step": 21
},
{
"loss": 0.0252,
"learning_rate": 7.2e-05,
"epoch": 1.16,
"step": 22
},
{
"loss": 0.0604,
"learning_rate": 7.2e-05,
"epoch": 1.21,
"step": 23
},
{
"loss": 0.0492,
"learning_rate": 7.2e-05,
"epoch": 1.26,
"step": 24
},
{
"loss": 0.0012,
"learning_rate": 7.2e-05,
"epoch": 1.32,
"step": 25
},
{
"eval_crt_2_loss": 0.008229807950556278,
"eval_crt_2_score": -0.0022093479055911303,
"eval_crt_2_brier_score": 0.0022093479055911303,
"eval_crt_2_average_probability": 0.9935852885246277,
"eval_crt_2_accuracy": 1.0,
"eval_crt_2_probabilities": [
0.9995800852775574,
0.9990842342376709,
0.9995512366294861,
0.9986786246299744,
0.9995718598365784,
0.9997767806053162,
0.8551581501960754,
0.9971579313278198,
0.9996451139450073,
0.9977812170982361,
0.999804675579071,
0.9992179870605469,
0.9997174143791199,
0.9998371601104736,
0.9993311166763306,
0.9997996687889099,
0.9997851252555847,
0.999744713306427,
0.9995379447937012,
0.9998577833175659,
0.9997304081916809,
0.9990813732147217,
0.9995477795600891,
0.9990488886833191,
0.9998594522476196,
0.9994975328445435,
0.9991641044616699,
0.9995660185813904,
0.9997511506080627,
0.9998865127563477,
0.9999109506607056,
0.9996562004089355,
0.9994305968284607,
0.9992846846580505,
0.9995037317276001,
0.9992471933364868,
0.9996359348297119,
0.999599277973175,
0.9991531372070312,
0.9997557997703552,
0.9994789958000183,
0.9998207688331604,
0.9998410940170288,
0.9997524619102478,
0.9998014569282532,
0.9998021721839905,
0.9992111921310425,
0.9998326301574707,
0.9994243383407593,
0.9990911483764648,
0.9997714161872864,
0.9993265867233276,
0.9995996356010437,
0.9998132586479187,
0.9998679161071777,
0.9991796612739563,
0.9997928738594055,
0.9997915625572205,
0.5528848171234131,
0.9991310238838196,
0.9997738003730774,
0.9996798038482666,
0.9998341798782349,
0.9998807907104492,
0.9995421171188354,
0.9997196793556213,
0.999329686164856,
0.9998488426208496,
0.9995917677879333,
0.9998519420623779,
0.9985900521278381,
0.999859094619751,
0.9980387091636658,
0.9995794892311096,
0.9995357990264893,
0.9989036321640015,
0.9994966983795166,
0.9996247291564941,
0.9990410208702087,
0.9995213747024536,
0.9994599223136902,
0.9998130202293396,
0.9993808269500732,
0.9995201826095581,
0.999403715133667,
0.9996811151504517,
0.9995848536491394,
0.9991675615310669,
0.9995501637458801,
0.9996263980865479,
0.9990099668502808,
0.999705970287323,
0.9996757507324219,
0.9992926120758057,
0.9993239641189575,
0.9999006986618042,
0.9998637437820435,
0.9997491240501404,
0.9996663331985474,
0.9997652173042297
],
"eval_crt_2_runtime": 4.2017,
"eval_crt_2_samples_per_second": 23.8,
"eval_crt_2_steps_per_second": 0.476,
"epoch": 1.32,
"step": 25
},
{
"loss": 0.0008,
"learning_rate": 7.2e-05,
"epoch": 1.37,
"step": 26
},
{
"loss": 0.0227,
"learning_rate": 7.2e-05,
"epoch": 1.42,
"step": 27
},
{
"loss": 0.0022,
"learning_rate": 7.2e-05,
"epoch": 1.47,
"step": 28
},
{
"loss": 0.0011,
"learning_rate": 7.2e-05,
"epoch": 1.53,
"step": 29
},
{
"loss": 0.0224,
"learning_rate": 7.2e-05,
"epoch": 1.58,
"step": 30
},
{
"loss": 0.2158,
"learning_rate": 7.2e-05,
"epoch": 1.63,
"step": 31
},
{
"loss": 0.0488,
"learning_rate": 7.2e-05,
"epoch": 1.68,
"step": 32
},
{
"loss": 0.0002,
"learning_rate": 7.2e-05,
"epoch": 1.74,
"step": 33
},
{
"loss": 0.0207,
"learning_rate": 7.2e-05,
"epoch": 1.79,
"step": 34
},
{
"loss": 0.0214,
"learning_rate": 7.2e-05,
"epoch": 1.84,
"step": 35
},
{
"loss": 0.007,
"learning_rate": 7.2e-05,
"epoch": 1.89,
"step": 36
},
{
"loss": 0.0256,
"learning_rate": 7.2e-05,
"epoch": 1.95,
"step": 37
},
{
"loss": 0.0026,
"learning_rate": 7.2e-05,
"epoch": 2.0,
"step": 38
},
{
"loss": 0.0405,
"learning_rate": 7.2e-05,
"epoch": 2.05,
"step": 39
},
{
"loss": 0.0406,
"learning_rate": 7.2e-05,
"epoch": 2.11,
"step": 40
},
{
"loss": 0.0232,
"learning_rate": 7.2e-05,
"epoch": 2.16,
"step": 41
},
{
"loss": 0.0432,
"learning_rate": 7.2e-05,
"epoch": 2.21,
"step": 42
},
{
"loss": 0.0,
"learning_rate": 7.2e-05,
"epoch": 2.26,
"step": 43
},
{
"loss": 0.0199,
"learning_rate": 7.2e-05,
"epoch": 2.32,
"step": 44
},
{
"loss": 0.0,
"learning_rate": 7.2e-05,
"epoch": 2.37,
"step": 45
},
{
"loss": 0.0221,
"learning_rate": 7.2e-05,
"epoch": 2.42,
"step": 46
},
{
"loss": 0.0,
"learning_rate": 7.2e-05,
"epoch": 2.47,
"step": 47
},
{
"loss": 0.0,
"learning_rate": 7.2e-05,
"epoch": 2.53,
"step": 48
},
{
"loss": 0.1631,
"learning_rate": 7.2e-05,
"epoch": 2.58,
"step": 49
},
{
"loss": 0.0013,
"learning_rate": 7.2e-05,
"epoch": 2.63,
"step": 50
},
{
"eval_crt_2_loss": 0.006326997186988592,
"eval_crt_2_score": -0.0017219263827428222,
"eval_crt_2_brier_score": 0.0017219263827428222,
"eval_crt_2_average_probability": 0.9949961304664612,
"eval_crt_2_accuracy": 1.0,
"eval_crt_2_probabilities": [
0.999997615814209,
0.9999998807907104,
0.9999988079071045,
0.9999997615814209,
0.9999998807907104,
0.999998927116394,
0.903231680393219,
0.9999995231628418,
0.9999998807907104,
0.9999997615814209,
0.9999998807907104,
0.9999996423721313,
0.9999996423721313,
0.9999996423721313,
0.9999997615814209,
0.999998927116394,
0.9999996423721313,
0.9999990463256836,
0.9999959468841553,
0.9999994039535522,
0.9999998807907104,
0.9999997615814209,
0.9999996423721313,
0.9999998807907104,
0.9999997615814209,
0.9999958276748657,
0.9999998807907104,
0.9999864101409912,
0.9999998807907104,
0.9999994039535522,
0.9999996423721313,
0.9999990463256836,
0.9999997615814209,
0.9999996423721313,
0.9999980926513672,
0.9999996423721313,
0.9999998807907104,
0.9999998807907104,
0.9999995231628418,
0.9999998807907104,
0.9999990463256836,
0.9999996423721313,
0.9999997615814209,
0.9999998807907104,
0.9999988079071045,
0.9999984502792358,
0.9999997615814209,
0.9999998807907104,
0.9999997615814209,
0.9999996423721313,
0.9999996423721313,
0.9999998807907104,
0.9999977350234985,
0.9999994039535522,
0.9999997615814209,
0.9999864101409912,
0.999998927116394,
0.9999995231628418,
0.5964798331260681,
0.9999996423721313,
0.9999994039535522,
0.9999986886978149,
0.9999992847442627,
0.9999996423721313,
0.999996542930603,
0.9999998807907104,
0.9999998807907104,
0.9999994039535522,
0.9999973773956299,
0.9999996423721313,
0.9999971389770508,
0.9999997615814209,
0.9999994039535522,
0.9999980926513672,
0.9999916553497314,
0.999998927116394,
0.9999998807907104,
0.9999998807907104,
0.9999996423721313,
0.9999940395355225,
0.9999998807907104,
0.9999995231628418,
0.9999997615814209,
0.9999996423721313,
0.9999997615814209,
0.9999998807907104,
0.9999997615814209,
0.9999996423721313,
0.9999996423721313,
0.9999998807907104,
0.9999927282333374,
0.9999998807907104,
0.9999998807907104,
0.9999998807907104,
0.9999997615814209,
0.9999996423721313,
0.9999994039535522,
0.9999992847442627,
0.9999995231628418,
0.9999998807907104
],
"eval_crt_2_runtime": 4.2,
"eval_crt_2_samples_per_second": 23.81,
"eval_crt_2_steps_per_second": 0.476,
"epoch": 2.63,
"step": 50
},
{
"loss": 0.0195,
"learning_rate": 7.2e-05,
"epoch": 2.68,
"step": 51
},
{
"loss": 0.001,
"learning_rate": 7.2e-05,
"epoch": 2.74,
"step": 52
},
{
"loss": 0.0006,
"learning_rate": 7.2e-05,
"epoch": 2.79,
"step": 53
},
{
"loss": 0.0,
"learning_rate": 7.2e-05,
"epoch": 2.84,
"step": 54
},
{
"loss": 0.0192,
"learning_rate": 7.2e-05,
"epoch": 2.89,
"step": 55
},
{
"loss": 0.0682,
"learning_rate": 7.2e-05,
"epoch": 2.95,
"step": 56
},
{
"loss": 0.0,
"learning_rate": 7.2e-05,
"epoch": 3.0,
"step": 57
},
{
"loss": 0.0,
"learning_rate": 7.2e-05,
"epoch": 3.05,
"step": 58
},
{
"loss": 0.0,
"learning_rate": 7.2e-05,
"epoch": 3.11,
"step": 59
},
{
"loss": 0.0703,
"learning_rate": 7.2e-05,
"epoch": 3.16,
"step": 60
},
{
"loss": 0.0,
"learning_rate": 7.2e-05,
"epoch": 3.21,
"step": 61
},
{
"loss": 0.0,
"learning_rate": 7.2e-05,
"epoch": 3.26,
"step": 62
},
{
"loss": 0.0184,
"learning_rate": 7.2e-05,
"epoch": 3.32,
"step": 63
},
{
"loss": 0.04,
"learning_rate": 7.2e-05,
"epoch": 3.37,
"step": 64
},
{
"loss": 0.0361,
"learning_rate": 7.2e-05,
"epoch": 3.42,
"step": 65
},
{
"loss": 0.0165,
"learning_rate": 7.2e-05,
"epoch": 3.47,
"step": 66
},
{
"loss": 0.0,
"learning_rate": 7.2e-05,
"epoch": 3.53,
"step": 67
},
{
"loss": 0.0421,
"learning_rate": 7.2e-05,
"epoch": 3.58,
"step": 68
},
{
"loss": 0.0,
"learning_rate": 7.2e-05,
"epoch": 3.63,
"step": 69
},
{
"loss": 0.0103,
"learning_rate": 7.2e-05,
"epoch": 3.68,
"step": 70
},
{
"loss": 0.0203,
"learning_rate": 7.2e-05,
"epoch": 3.74,
"step": 71
},
{
"loss": 0.0151,
"learning_rate": 7.2e-05,
"epoch": 3.79,
"step": 72
},
{
"loss": 0.0223,
"learning_rate": 7.2e-05,
"epoch": 3.84,
"step": 73
},
{
"loss": 0.0117,
"learning_rate": 7.2e-05,
"epoch": 3.89,
"step": 74
},
{
"loss": 0.0001,
"learning_rate": 7.2e-05,
"epoch": 3.95,
"step": 75
},
{
"eval_crt_2_loss": 0.004923115950077772,
"eval_crt_2_score": -0.0013668893370777369,
"eval_crt_2_brier_score": 0.0013668893370777369,
"eval_crt_2_average_probability": 0.9960331916809082,
"eval_crt_2_accuracy": 1.0,
"eval_crt_2_probabilities": [
0.9999960660934448,
0.9999995231628418,
0.9999983310699463,
0.9999994039535522,
0.9999997615814209,
0.9999990463256836,
0.9720984697341919,
0.9999990463256836,
0.9999998807907104,
0.9999983310699463,
0.9999998807907104,
0.9999978542327881,
0.9999997615814209,
0.9999996423721313,
0.9999997615814209,
0.9999990463256836,
0.9999995231628418,
0.9999992847442627,
0.9999959468841553,
0.9999994039535522,
0.9999998807907104,
0.9999988079071045,
0.9999997615814209,
0.9999997615814209,
0.9999997615814209,
0.9999938011169434,
0.9999997615814209,
0.9999915361404419,
0.9999998807907104,
0.9999994039535522,
0.9999996423721313,
0.9999992847442627,
0.9999991655349731,
0.9999996423721313,
0.9999992847442627,
0.9999997615814209,
0.9999998807907104,
0.9999998807907104,
0.999998927116394,
0.9999998807907104,
0.9999991655349731,
0.9999997615814209,
0.9999997615814209,
0.9999998807907104,
0.9999988079071045,
0.9999980926513672,
0.9999996423721313,
0.9999998807907104,
0.9999996423721313,
0.9999996423721313,
0.9999996423721313,
0.9999997615814209,
0.9999984502792358,
0.9999994039535522,
0.9999998807907104,
0.9999746084213257,
0.9999992847442627,
0.9999995231628418,
0.6313396692276001,
0.9999997615814209,
0.9999995231628418,
0.9999992847442627,
0.9999992847442627,
0.9999996423721313,
0.9999957084655762,
0.9999998807907104,
0.9999998807907104,
0.9999994039535522,
0.9999988079071045,
0.9999995231628418,
0.9999964237213135,
0.9999997615814209,
0.9999973773956299,
0.9999982118606567,
0.9999940395355225,
0.9999990463256836,
0.9999998807907104,
0.9999998807907104,
0.9999991655349731,
0.9999963045120239,
0.9999998807907104,
0.9999994039535522,
0.9999996423721313,
0.9999997615814209,
0.9999998807907104,
0.9999998807907104,
0.9999997615814209,
0.9999990463256836,
0.9999995231628418,
0.9999998807907104,
0.9999871253967285,
0.9999998807907104,
0.9999998807907104,
0.9999997615814209,
0.9999996423721313,
0.9999995231628418,
0.9999995231628418,
0.9999985694885254,
0.9999994039535522,
0.9999998807907104
],
"eval_crt_2_runtime": 4.2026,
"eval_crt_2_samples_per_second": 23.795,
"eval_crt_2_steps_per_second": 0.476,
"epoch": 3.95,
"step": 75
},
{
"loss": 0.0001,
"learning_rate": 7.2e-05,
"epoch": 4.0,
"step": 76
},
{
"loss": 0.0,
"learning_rate": 7.2e-05,
"epoch": 4.05,
"step": 77
},
{
"loss": 0.0105,
"learning_rate": 7.2e-05,
"epoch": 4.11,
"step": 78
},
{
"loss": 0.0002,
"learning_rate": 7.2e-05,
"epoch": 4.16,
"step": 79
},
{
"loss": 0.0,
"learning_rate": 7.2e-05,
"epoch": 4.21,
"step": 80
},
{
"loss": 0.0001,
"learning_rate": 7.2e-05,
"epoch": 4.26,
"step": 81
},
{
"loss": 0.0,
"learning_rate": 7.2e-05,
"epoch": 4.32,
"step": 82
},
{
"loss": 0.0214,
"learning_rate": 7.2e-05,
"epoch": 4.37,
"step": 83
},
{
"loss": 0.0105,
"learning_rate": 7.2e-05,
"epoch": 4.42,
"step": 84
},
{
"loss": 0.0001,
"learning_rate": 7.2e-05,
"epoch": 4.47,
"step": 85
},
{
"loss": 0.0116,
"learning_rate": 7.2e-05,
"epoch": 4.53,
"step": 86
},
{
"loss": 0.0168,
"learning_rate": 7.2e-05,
"epoch": 4.58,
"step": 87
},
{
"loss": 0.0329,
"learning_rate": 7.2e-05,
"epoch": 4.63,
"step": 88
},
{
"loss": 0.0118,
"learning_rate": 7.2e-05,
"epoch": 4.68,
"step": 89
},
{
"loss": 0.0498,
"learning_rate": 7.2e-05,
"epoch": 4.74,
"step": 90
},
{
"loss": 0.0033,
"learning_rate": 7.2e-05,
"epoch": 4.79,
"step": 91
},
{
"loss": 0.0124,
"learning_rate": 7.2e-05,
"epoch": 4.84,
"step": 92
},
{
"loss": 0.0268,
"learning_rate": 7.2e-05,
"epoch": 4.89,
"step": 93
},
{
"loss": 0.0,
"learning_rate": 7.2e-05,
"epoch": 4.95,
"step": 94
},
{
"loss": 0.0225,
"learning_rate": 7.2e-05,
"epoch": 5.0,
"step": 95
},
{
"loss": 0.0,
"learning_rate": 7.2e-05,
"epoch": 5.05,
"step": 96
},
{
"loss": 0.0059,
"learning_rate": 7.2e-05,
"epoch": 5.11,
"step": 97
},
{
"loss": 0.0,
"learning_rate": 7.2e-05,
"epoch": 5.16,
"step": 98
},
{
"loss": 0.011,
"learning_rate": 7.2e-05,
"epoch": 5.21,
"step": 99
},
{
"loss": 0.0001,
"learning_rate": 7.2e-05,
"epoch": 5.26,
"step": 100
},
{
"eval_crt_2_loss": 0.003461032873019576,
"eval_crt_2_score": -0.0006671486771665514,
"eval_crt_2_brier_score": 0.0006671486771665514,
"eval_crt_2_average_probability": 0.9970054626464844,
"eval_crt_2_accuracy": 1.0,
"eval_crt_2_probabilities": [
0.9999970197677612,
1.0,
0.9999985694885254,
1.0,
1.0,
0.9999984502792358,
0.9550073742866516,
1.0,
1.0,
0.9999998807907104,
1.0,
0.9999997615814209,
1.0,
0.9999994039535522,
1.0,
0.9999995231628418,
0.9999997615814209,
0.9999996423721313,
0.9999953508377075,
0.9999983310699463,
1.0,
1.0,
1.0,
1.0,
0.9999998807907104,
0.9999942779541016,
1.0,
0.9999829530715942,
1.0,
0.9999996423721313,
0.9999997615814209,
0.9999998807907104,
0.9999998807907104,
1.0,
0.9999996423721313,
1.0,
1.0,
1.0,
0.9999998807907104,
1.0,
0.9999995231628418,
0.9999998807907104,
0.9999998807907104,
1.0,
0.999998927116394,
0.9999960660934448,
1.0,
1.0,
1.0,
1.0,
0.9999998807907104,
1.0,
0.9999939203262329,
0.999998927116394,
1.0,
0.9999673366546631,
0.9999995231628418,
0.9999995231628418,
0.7456566691398621,
1.0,
0.9999984502792358,
0.9999994039535522,
0.9999994039535522,
0.9999996423721313,
0.9999964237213135,
1.0,
1.0,
0.9999990463256836,
0.9999996423721313,
0.9999998807907104,
0.9999982118606567,
0.9999997615814209,
0.9999998807907104,
0.9999986886978149,
0.9999911785125732,
0.9999998807907104,
1.0,
1.0,
0.9999998807907104,
0.9999972581863403,
1.0,
0.9999997615814209,
1.0,
1.0,
1.0,
1.0,
1.0,
0.9999998807907104,
1.0,
1.0,
0.9999868869781494,
1.0,
1.0,
1.0,
1.0,
0.9999995231628418,
0.9999994039535522,
0.9999985694885254,
0.9999995231628418,
1.0
],
"eval_crt_2_runtime": 4.2012,
"eval_crt_2_samples_per_second": 23.803,
"eval_crt_2_steps_per_second": 0.476,
"epoch": 5.26,
"step": 100
},
{
"train_runtime": 366.2575,
"train_samples_per_second": 8.737,
"train_steps_per_second": 0.273,
"total_flos": 0.0,
"train_loss": 0.10658088955936648,
"epoch": 5.26,
"step": 100
}
] }
]