Final training metrics
Browse files- training_metrics.json +102 -0
training_metrics.json
ADDED
|
@@ -0,0 +1,102 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"metrics_history": [
|
| 3 |
+
{
|
| 4 |
+
"epoch": 0.23637867864318637,
|
| 5 |
+
"global_step": 2000,
|
| 6 |
+
"eval_loss": 1.1521742343902588,
|
| 7 |
+
"eval_cosine_accuracy": 0.9351803064346313
|
| 8 |
+
},
|
| 9 |
+
{
|
| 10 |
+
"epoch": 0.47275735728637275,
|
| 11 |
+
"global_step": 4000,
|
| 12 |
+
"eval_loss": 1.1368498802185059,
|
| 13 |
+
"eval_cosine_accuracy": 0.9370836615562439
|
| 14 |
+
},
|
| 15 |
+
{
|
| 16 |
+
"epoch": 0.7091360359295592,
|
| 17 |
+
"global_step": 6000,
|
| 18 |
+
"eval_loss": 1.088667392730713,
|
| 19 |
+
"eval_cosine_accuracy": 0.9424764513969421
|
| 20 |
+
},
|
| 21 |
+
{
|
| 22 |
+
"epoch": 0.9455147145727455,
|
| 23 |
+
"global_step": 8000,
|
| 24 |
+
"eval_loss": 1.0786563158035278,
|
| 25 |
+
"eval_cosine_accuracy": 0.9423707127571106
|
| 26 |
+
},
|
| 27 |
+
{
|
| 28 |
+
"epoch": 1.181721572794899,
|
| 29 |
+
"global_step": 10000,
|
| 30 |
+
"eval_loss": 1.0610214471817017,
|
| 31 |
+
"eval_cosine_accuracy": 0.9449085593223572
|
| 32 |
+
},
|
| 33 |
+
{
|
| 34 |
+
"epoch": 1.4178769630416814,
|
| 35 |
+
"global_step": 12000,
|
| 36 |
+
"eval_loss": 1.0575031042099,
|
| 37 |
+
"eval_cosine_accuracy": 0.9460716843605042
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"epoch": 1.6540323532884638,
|
| 41 |
+
"global_step": 14000,
|
| 42 |
+
"eval_loss": 1.0424392223358154,
|
| 43 |
+
"eval_cosine_accuracy": 0.9483979940414429
|
| 44 |
+
},
|
| 45 |
+
{
|
| 46 |
+
"epoch": 1.8901877435352463,
|
| 47 |
+
"global_step": 16000,
|
| 48 |
+
"eval_loss": 1.003242015838623,
|
| 49 |
+
"eval_cosine_accuracy": 0.9505128264427185
|
| 50 |
+
},
|
| 51 |
+
{
|
| 52 |
+
"epoch": 2.1263431337820284,
|
| 53 |
+
"global_step": 18000,
|
| 54 |
+
"eval_loss": 1.01349675655365,
|
| 55 |
+
"eval_cosine_accuracy": 0.9495611786842346
|
| 56 |
+
},
|
| 57 |
+
{
|
| 58 |
+
"epoch": 2.362498524028811,
|
| 59 |
+
"global_step": 20000,
|
| 60 |
+
"eval_loss": 0.9978452324867249,
|
| 61 |
+
"eval_cosine_accuracy": 0.9518874883651733
|
| 62 |
+
},
|
| 63 |
+
{
|
| 64 |
+
"epoch": 2.5986539142755936,
|
| 65 |
+
"global_step": 22000,
|
| 66 |
+
"eval_loss": 0.9861479997634888,
|
| 67 |
+
"eval_cosine_accuracy": 0.954002320766449
|
| 68 |
+
},
|
| 69 |
+
{
|
| 70 |
+
"epoch": 2.8348093045223757,
|
| 71 |
+
"global_step": 24000,
|
| 72 |
+
"eval_loss": 0.9923800230026245,
|
| 73 |
+
"eval_cosine_accuracy": 0.9523104429244995
|
| 74 |
+
},
|
| 75 |
+
{
|
| 76 |
+
"epoch": 3.0709646947691582,
|
| 77 |
+
"global_step": 26000,
|
| 78 |
+
"eval_loss": 0.9917270541191101,
|
| 79 |
+
"eval_cosine_accuracy": 0.9538965821266174
|
| 80 |
+
},
|
| 81 |
+
{
|
| 82 |
+
"epoch": 3.3071200850159403,
|
| 83 |
+
"global_step": 28000,
|
| 84 |
+
"eval_loss": 0.984880805015564,
|
| 85 |
+
"eval_cosine_accuracy": 0.9547424912452698
|
| 86 |
+
},
|
| 87 |
+
{
|
| 88 |
+
"epoch": 3.543275475262723,
|
| 89 |
+
"global_step": 30000,
|
| 90 |
+
"eval_loss": 0.9861717224121094,
|
| 91 |
+
"eval_cosine_accuracy": 0.9554827213287354
|
| 92 |
+
},
|
| 93 |
+
{
|
| 94 |
+
"epoch": 3.7794308655095055,
|
| 95 |
+
"global_step": 32000,
|
| 96 |
+
"eval_loss": 0.9865333437919617,
|
| 97 |
+
"eval_cosine_accuracy": 0.9551655054092407
|
| 98 |
+
}
|
| 99 |
+
],
|
| 100 |
+
"total_epochs": 3.9971661353170385,
|
| 101 |
+
"total_steps": 33844
|
| 102 |
+
}
|