llama-30b-wrong_arc / training_logs.json
joshuaclymer's picture
Upload folder using huggingface_hub
7c84e7d
Invalid JSON:Unexpected non-whitespace character after JSONat line 1068, column 1
[
{
"loss": 0.7392,
"learning_rate": 0.0002,
"epoch": 0.05,
"step": 1
},
{
"loss": 0.6687,
"learning_rate": 0.0002,
"epoch": 0.11,
"step": 2
},
{
"loss": 0.7371,
"learning_rate": 0.0002,
"epoch": 0.16,
"step": 3
},
{
"loss": 0.777,
"learning_rate": 0.0002,
"epoch": 0.21,
"step": 4
},
{
"loss": 0.7379,
"learning_rate": 0.0002,
"epoch": 0.26,
"step": 5
},
{
"loss": 0.645,
"learning_rate": 0.0002,
"epoch": 0.32,
"step": 6
},
{
"loss": 0.6729,
"learning_rate": 0.0002,
"epoch": 0.37,
"step": 7
},
{
"loss": 0.7124,
"learning_rate": 0.0002,
"epoch": 0.42,
"step": 8
},
{
"loss": 0.5295,
"learning_rate": 0.0002,
"epoch": 0.47,
"step": 9
},
{
"loss": 0.5603,
"learning_rate": 0.0002,
"epoch": 0.53,
"step": 10
},
{
"loss": 0.4843,
"learning_rate": 0.0002,
"epoch": 0.58,
"step": 11
},
{
"loss": 0.4394,
"learning_rate": 0.0002,
"epoch": 0.63,
"step": 12
},
{
"loss": 0.3661,
"learning_rate": 0.0002,
"epoch": 0.68,
"step": 13
},
{
"loss": 0.417,
"learning_rate": 0.0002,
"epoch": 0.74,
"step": 14
},
{
"loss": 0.275,
"learning_rate": 0.0002,
"epoch": 0.79,
"step": 15
},
{
"loss": 0.3067,
"learning_rate": 0.0002,
"epoch": 0.84,
"step": 16
},
{
"loss": 0.1262,
"learning_rate": 0.0002,
"epoch": 0.89,
"step": 17
},
{
"loss": 0.2333,
"learning_rate": 0.0002,
"epoch": 0.95,
"step": 18
},
{
"loss": 0.4396,
"learning_rate": 0.0002,
"epoch": 1.0,
"step": 19
},
{
"loss": 0.2469,
"learning_rate": 0.0002,
"epoch": 1.05,
"step": 20
},
{
"loss": 0.0491,
"learning_rate": 0.0002,
"epoch": 1.11,
"step": 21
},
{
"loss": 0.0602,
"learning_rate": 0.0002,
"epoch": 1.16,
"step": 22
},
{
"loss": 0.0374,
"learning_rate": 0.0002,
"epoch": 1.21,
"step": 23
},
{
"loss": 0.2447,
"learning_rate": 0.0002,
"epoch": 1.26,
"step": 24
},
{
"loss": 0.3911,
"learning_rate": 0.0002,
"epoch": 1.32,
"step": 25
},
{
"eval_wrong_arc_loss": 0.21497203409671783,
"eval_wrong_arc_score": -0.06468652933835983,
"eval_wrong_arc_brier_score": 0.06468652933835983,
"eval_wrong_arc_average_probability": 0.8909010887145996,
"eval_wrong_arc_accuracy": 0.92,
"eval_wrong_arc_probabilities": [
0.9997676014900208,
0.996569037437439,
0.616486132144928,
0.9944661259651184,
0.9982737302780151,
0.9993508458137512,
0.9668838381767273,
0.9999992847442627,
0.9999873638153076,
0.9904683232307434,
0.9933720827102661,
0.6185683608055115,
0.9949280023574829,
0.9999911785125732,
0.9996459484100342,
0.2095331847667694,
0.9992320537567139,
0.6855879426002502,
0.6939541101455688,
0.9886998534202576,
0.4002212584018707,
0.9976931214332581,
0.9850371479988098,
0.9999406337738037,
0.6477287411689758,
0.9998879432678223,
0.9803314805030823,
0.9461862444877625,
0.9734556674957275,
0.9999347925186157,
0.9980067610740662,
0.8022632598876953,
0.999922513961792,
0.946151852607727,
0.662533164024353,
0.9975826740264893,
0.9998453855514526,
0.9968388080596924,
0.9699428677558899,
0.9917652010917664,
0.9323654174804688,
0.9990322589874268,
0.9997826218605042,
0.9999943971633911,
0.9999864101409912,
0.01022251695394516,
0.9999226331710815,
0.21935860812664032,
0.9370521306991577,
0.7875561714172363,
0.9789162874221802,
0.7642993927001953,
0.9912926554679871,
0.9982261061668396,
0.9988908171653748,
0.9821295142173767,
0.9997692704200745,
0.9968388080596924,
0.027354620397090912,
0.985841691493988,
0.9955176711082458,
0.9964224696159363,
0.998650848865509,
0.991766631603241,
0.9993088245391846,
0.999588668346405,
0.9985834360122681,
0.9933047890663147,
0.9974768757820129,
0.9967116117477417,
0.9997716546058655,
0.8711239099502563,
0.9992668032646179,
0.4505612254142761,
0.9998251795768738,
0.9984416365623474,
0.9977090358734131,
0.998289167881012,
0.9936224818229675,
0.6056644916534424,
0.9966290593147278,
0.9860624670982361,
0.9775242805480957,
0.9779098033905029,
0.9847407937049866,
0.9994072914123535,
0.9997192025184631,
0.9340455532073975,
0.5483614206314087,
0.9444760680198669,
0.968048095703125,
0.2896791994571686,
0.9574971795082092,
0.9999983310699463,
0.9990599751472473,
0.940399169921875,
0.9872337579727173,
0.06788471341133118,
0.9999967813491821,
0.9999508857727051
],
"eval_wrong_arc_runtime": 37.226,
"eval_wrong_arc_samples_per_second": 2.686,
"eval_wrong_arc_steps_per_second": 0.107,
"epoch": 1.32,
"step": 25
},
{
"loss": 0.0897,
"learning_rate": 0.0002,
"epoch": 1.37,
"step": 26
},
{
"loss": 0.0541,
"learning_rate": 0.0002,
"epoch": 1.42,
"step": 27
},
{
"loss": 0.0361,
"learning_rate": 0.0002,
"epoch": 1.47,
"step": 28
},
{
"loss": 0.073,
"learning_rate": 0.0002,
"epoch": 1.53,
"step": 29
},
{
"loss": 0.1269,
"learning_rate": 0.0002,
"epoch": 1.58,
"step": 30
},
{
"loss": 0.0501,
"learning_rate": 0.0002,
"epoch": 1.63,
"step": 31
},
{
"loss": 0.1058,
"learning_rate": 0.0002,
"epoch": 1.68,
"step": 32
},
{
"loss": 0.067,
"learning_rate": 0.0002,
"epoch": 1.74,
"step": 33
},
{
"loss": 0.0124,
"learning_rate": 0.0002,
"epoch": 1.79,
"step": 34
},
{
"loss": 0.1432,
"learning_rate": 0.0002,
"epoch": 1.84,
"step": 35
},
{
"loss": 0.5075,
"learning_rate": 0.0002,
"epoch": 1.89,
"step": 36
},
{
"loss": 0.118,
"learning_rate": 0.0002,
"epoch": 1.95,
"step": 37
},
{
"loss": 0.0679,
"learning_rate": 0.0002,
"epoch": 2.0,
"step": 38
},
{
"loss": 0.0041,
"learning_rate": 0.0002,
"epoch": 2.05,
"step": 39
},
{
"loss": 0.0083,
"learning_rate": 0.0002,
"epoch": 2.11,
"step": 40
},
{
"loss": 0.016,
"learning_rate": 0.0002,
"epoch": 2.16,
"step": 41
},
{
"loss": 0.0122,
"learning_rate": 0.0002,
"epoch": 2.21,
"step": 42
},
{
"loss": 0.0064,
"learning_rate": 0.0002,
"epoch": 2.26,
"step": 43
},
{
"loss": 0.013,
"learning_rate": 0.0002,
"epoch": 2.32,
"step": 44
},
{
"loss": 0.0162,
"learning_rate": 0.0002,
"epoch": 2.37,
"step": 45
},
{
"loss": 0.0009,
"learning_rate": 0.0002,
"epoch": 2.42,
"step": 46
},
{
"loss": 0.0018,
"learning_rate": 0.0002,
"epoch": 2.47,
"step": 47
},
{
"loss": 0.0013,
"learning_rate": 0.0002,
"epoch": 2.53,
"step": 48
},
{
"loss": 0.0004,
"learning_rate": 0.0002,
"epoch": 2.58,
"step": 49
},
{
"loss": 0.0891,
"learning_rate": 0.0002,
"epoch": 2.63,
"step": 50
},
{
"eval_wrong_arc_loss": 0.48733100295066833,
"eval_wrong_arc_score": -0.07911639660596848,
"eval_wrong_arc_brier_score": 0.07911639660596848,
"eval_wrong_arc_average_probability": 0.9070995450019836,
"eval_wrong_arc_accuracy": 0.91,
"eval_wrong_arc_probabilities": [
1.0,
0.9999991655349731,
0.9936402440071106,
1.0,
0.9999974966049194,
1.0,
0.9864437580108643,
1.0,
1.0,
0.9988510608673096,
1.0,
0.008249929174780846,
1.0,
1.0,
1.0,
0.0014172116061672568,
1.0,
0.8445881009101868,
0.9999101161956787,
1.0,
0.05873167887330055,
1.0,
1.0,
1.0,
0.923774242401123,
1.0,
0.9947189092636108,
0.9960833787918091,
0.999904990196228,
1.0,
1.0,
0.9999992847442627,
1.0,
0.9999345541000366,
0.6755237579345703,
1.0,
1.0,
0.9999872446060181,
0.9999872446060181,
0.9999998807907104,
0.9999958276748657,
1.0,
1.0,
1.0,
1.0,
1.5062338931670638e-08,
1.0,
0.33844730257987976,
0.99930739402771,
0.728569507598877,
0.9999954700469971,
0.999964714050293,
1.0,
1.0,
1.0,
0.9999351501464844,
1.0,
0.9999997615814209,
0.002960789715871215,
0.999830961227417,
0.9999600648880005,
0.9999991655349731,
1.0,
0.999984860420227,
1.0,
1.0,
0.9999994039535522,
0.9999997615814209,
0.9999983310699463,
0.9999997615814209,
1.0,
0.9996284246444702,
0.9998459815979004,
0.7875558137893677,
1.0,
0.9999998807907104,
1.0,
1.0,
0.9999980926513672,
1.0,
1.0,
1.0,
0.9973583817481995,
0.9999964237213135,
0.9998266100883484,
1.0,
1.0,
0.9963425993919373,
0.3505668342113495,
0.9979604482650757,
0.9999998807907104,
0.031407516449689865,
1.0,
1.0,
1.0,
0.9982225298881531,
0.9999990463256836,
0.0005543192382901907,
1.0,
1.0
],
"eval_wrong_arc_runtime": 37.2016,
"eval_wrong_arc_samples_per_second": 2.688,
"eval_wrong_arc_steps_per_second": 0.108,
"epoch": 2.63,
"step": 50
},
{
"loss": 0.0038,
"learning_rate": 0.0002,
"epoch": 2.68,
"step": 51
},
{
"loss": 0.0028,
"learning_rate": 0.0002,
"epoch": 2.74,
"step": 52
},
{
"loss": 0.1882,
"learning_rate": 0.0002,
"epoch": 2.79,
"step": 53
},
{
"loss": 0.0001,
"learning_rate": 0.0002,
"epoch": 2.84,
"step": 54
},
{
"loss": 0.0014,
"learning_rate": 0.0002,
"epoch": 2.89,
"step": 55
},
{
"loss": 0.0003,
"learning_rate": 0.0002,
"epoch": 2.95,
"step": 56
},
{
"loss": 0.0075,
"learning_rate": 0.0002,
"epoch": 3.0,
"step": 57
},
{
"loss": 0.0002,
"learning_rate": 0.0002,
"epoch": 3.05,
"step": 58
},
{
"loss": 0.0015,
"learning_rate": 0.0002,
"epoch": 3.11,
"step": 59
},
{
"loss": 0.0001,
"learning_rate": 0.0002,
"epoch": 3.16,
"step": 60
},
{
"loss": 0.0008,
"learning_rate": 0.0002,
"epoch": 3.21,
"step": 61
},
{
"loss": 0.0004,
"learning_rate": 0.0002,
"epoch": 3.26,
"step": 62
},
{
"loss": 0.0012,
"learning_rate": 0.0002,
"epoch": 3.32,
"step": 63
},
{
"loss": 0.0004,
"learning_rate": 0.0002,
"epoch": 3.37,
"step": 64
},
{
"loss": 0.0005,
"learning_rate": 0.0002,
"epoch": 3.42,
"step": 65
},
{
"loss": 0.0001,
"learning_rate": 0.0002,
"epoch": 3.47,
"step": 66
},
{
"loss": 0.0001,
"learning_rate": 0.0002,
"epoch": 3.53,
"step": 67
},
{
"loss": 0.0,
"learning_rate": 0.0002,
"epoch": 3.58,
"step": 68
},
{
"loss": 0.0001,
"learning_rate": 0.0002,
"epoch": 3.63,
"step": 69
},
{
"loss": 0.0001,
"learning_rate": 0.0002,
"epoch": 3.68,
"step": 70
},
{
"loss": 0.0,
"learning_rate": 0.0002,
"epoch": 3.74,
"step": 71
},
{
"loss": 0.0004,
"learning_rate": 0.0002,
"epoch": 3.79,
"step": 72
},
{
"loss": 0.0014,
"learning_rate": 0.0002,
"epoch": 3.84,
"step": 73
},
{
"loss": 0.0004,
"learning_rate": 0.0002,
"epoch": 3.89,
"step": 74
},
{
"loss": 0.0001,
"learning_rate": 0.0002,
"epoch": 3.95,
"step": 75
},
{
"eval_wrong_arc_loss": 0.4877793788909912,
"eval_wrong_arc_score": -0.0775582566857338,
"eval_wrong_arc_brier_score": 0.0775582566857338,
"eval_wrong_arc_average_probability": 0.9079529047012329,
"eval_wrong_arc_accuracy": 0.91,
"eval_wrong_arc_probabilities": [
1.0,
0.9999990463256836,
0.9999665021896362,
1.0,
1.0,
0.9999998807907104,
0.9894501566886902,
1.0,
1.0,
0.9996355772018433,
0.9999998807907104,
0.013500905595719814,
1.0,
1.0,
1.0,
0.000785358774010092,
1.0,
0.9177922010421753,
0.9999915361404419,
0.9999995231628418,
0.07539928704500198,
1.0,
1.0,
1.0,
0.9869281649589539,
1.0,
0.957822322845459,
0.9971826076507568,
0.9999237060546875,
1.0,
1.0,
0.9999997615814209,
1.0,
0.999997615814209,
0.528329610824585,
1.0,
1.0,
0.9999992847442627,
0.9999744892120361,
0.9999998807907104,
0.9999997615814209,
1.0,
1.0,
1.0,
1.0,
1.042173680687597e-09,
1.0,
0.4166145920753479,
0.9972959160804749,
0.7838015556335449,
0.9999197721481323,
0.9999262094497681,
1.0,
1.0,
1.0,
0.9999864101409912,
1.0,
0.9999997615814209,
0.022287288680672646,
0.9999945163726807,
0.999991774559021,
0.9999997615814209,
0.9999998807907104,
0.999998927116394,
1.0,
1.0,
0.9999979734420776,
1.0,
1.0,
1.0,
0.9999998807907104,
0.9999946355819702,
0.9999932050704956,
0.6798059344291687,
1.0,
0.9999951124191284,
1.0,
0.9999998807907104,
0.9999998807907104,
1.0,
1.0,
1.0,
0.9972917437553406,
0.9999983310699463,
0.9997867941856384,
0.9999988079071045,
1.0,
0.9999314546585083,
0.32395148277282715,
0.9999831914901733,
1.0,
0.10796832293272018,
1.0,
1.0,
1.0,
0.9999572038650513,
0.9999991655349731,
0.0001372263504890725,
1.0,
1.0
],
"eval_wrong_arc_runtime": 37.244,
"eval_wrong_arc_samples_per_second": 2.685,
"eval_wrong_arc_steps_per_second": 0.107,
"epoch": 3.95,
"step": 75
},
{
"loss": 0.0002,
"learning_rate": 0.0002,
"epoch": 4.0,
"step": 76
},
{
"loss": 0.0001,
"learning_rate": 0.0002,
"epoch": 4.05,
"step": 77
},
{
"loss": 0.0,
"learning_rate": 0.0002,
"epoch": 4.11,
"step": 78
},
{
"loss": 0.0,
"learning_rate": 0.0002,
"epoch": 4.16,
"step": 79
},
{
"loss": 0.0001,
"learning_rate": 0.0002,
"epoch": 4.21,
"step": 80
},
{
"loss": 0.0001,
"learning_rate": 0.0002,
"epoch": 4.26,
"step": 81
},
{
"loss": 0.0,
"learning_rate": 0.0002,
"epoch": 4.32,
"step": 82
},
{
"loss": 0.0,
"learning_rate": 0.0002,
"epoch": 4.37,
"step": 83
},
{
"loss": 0.0,
"learning_rate": 0.0002,
"epoch": 4.42,
"step": 84
},
{
"loss": 0.0001,
"learning_rate": 0.0002,
"epoch": 4.47,
"step": 85
},
{
"loss": 0.0001,
"learning_rate": 0.0002,
"epoch": 4.53,
"step": 86
},
{
"loss": 0.0,
"learning_rate": 0.0002,
"epoch": 4.58,
"step": 87
},
{
"loss": 0.0,
"learning_rate": 0.0002,
"epoch": 4.63,
"step": 88
},
{
"loss": 0.0,
"learning_rate": 0.0002,
"epoch": 4.68,
"step": 89
},
{
"loss": 0.0,
"learning_rate": 0.0002,
"epoch": 4.74,
"step": 90
},
{
"loss": 0.0,
"learning_rate": 0.0002,
"epoch": 4.79,
"step": 91
},
{
"loss": 0.0,
"learning_rate": 0.0002,
"epoch": 4.84,
"step": 92
},
{
"loss": 0.0,
"learning_rate": 0.0002,
"epoch": 4.89,
"step": 93
},
{
"loss": 0.0,
"learning_rate": 0.0002,
"epoch": 4.95,
"step": 94
},
{
"loss": 0.0,
"learning_rate": 0.0002,
"epoch": 5.0,
"step": 95
},
{
"loss": 0.0,
"learning_rate": 0.0002,
"epoch": 5.05,
"step": 96
},
{
"loss": 0.0,
"learning_rate": 0.0002,
"epoch": 5.11,
"step": 97
},
{
"loss": 0.0,
"learning_rate": 0.0002,
"epoch": 5.16,
"step": 98
},
{
"loss": 0.0,
"learning_rate": 0.0002,
"epoch": 5.21,
"step": 99
},
{
"loss": 0.0,
"learning_rate": 0.0002,
"epoch": 5.26,
"step": 100
},
{
"eval_wrong_arc_loss": 0.5397974252700806,
"eval_wrong_arc_score": -0.07924876362085342,
"eval_wrong_arc_brier_score": 0.07924876362085342,
"eval_wrong_arc_average_probability": 0.9080300331115723,
"eval_wrong_arc_accuracy": 0.91,
"eval_wrong_arc_probabilities": [
1.0,
0.9999997615814209,
1.0,
1.0,
1.0,
1.0,
0.9819542169570923,
1.0,
1.0,
0.9999889135360718,
1.0,
0.013341457583010197,
1.0,
1.0,
1.0,
0.0025760577991604805,
1.0,
0.9580605030059814,
0.9999943971633911,
1.0,
0.05881117656826973,
1.0,
1.0,
1.0,
0.9972819089889526,
1.0,
0.9700170755386353,
0.9933912754058838,
0.9999696016311646,
1.0,
1.0,
0.9999991655349731,
1.0,
1.0,
0.6296136975288391,
1.0,
1.0,
1.0,
0.9999990463256836,
1.0,
1.0,
1.0,
1.0,
1.0,
1.0,
6.58557496729939e-12,
1.0,
0.28933021426200867,
0.9998461008071899,
0.882614016532898,
0.999944806098938,
0.9999732971191406,
1.0,
1.0,
1.0,
0.9999996423721313,
1.0,
0.9999998807907104,
0.021230559796094894,
0.9999998807907104,
0.9999992847442627,
1.0,
1.0,
1.0,
1.0,
1.0,
1.0,
1.0,
1.0,
1.0,
1.0,
0.9999927282333374,
1.0,
0.6026157736778259,
1.0,
1.0,
1.0,
1.0,
1.0,
1.0,
1.0,
1.0,
0.9998080134391785,
0.9999998807907104,
0.9999912977218628,
0.9999998807907104,
1.0,
0.9998971223831177,
0.3017774522304535,
0.9999996423721313,
1.0,
0.10095995664596558,
1.0,
1.0,
1.0,
0.9999998807907104,
0.9999998807907104,
2.6150612029596232e-05,
1.0,
1.0
],
"eval_wrong_arc_runtime": 37.2108,
"eval_wrong_arc_samples_per_second": 2.687,
"eval_wrong_arc_steps_per_second": 0.107,
"epoch": 5.26,
"step": 100
},
{
"train_runtime": 2869.7153,
"train_samples_per_second": 1.115,
"train_steps_per_second": 0.035,
"total_flos": 0.0,
"train_loss": 0.1273127639741233,
"epoch": 5.26,
"step": 100
}
]
]