Final training metrics
Browse files- training_metrics.json +90 -0
training_metrics.json
ADDED
|
@@ -0,0 +1,90 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"metrics_history": [
|
| 3 |
+
{
|
| 4 |
+
"epoch": 0.42118563756975885,
|
| 5 |
+
"global_step": 2000,
|
| 6 |
+
"eval_loss": 0.7114619016647339,
|
| 7 |
+
"eval_cosine_accuracy": 0.940467357635498
|
| 8 |
+
},
|
| 9 |
+
{
|
| 10 |
+
"epoch": 0.8423712751395177,
|
| 11 |
+
"global_step": 4000,
|
| 12 |
+
"eval_loss": 0.6677370667457581,
|
| 13 |
+
"eval_cosine_accuracy": 0.94924396276474
|
| 14 |
+
},
|
| 15 |
+
{
|
| 16 |
+
"epoch": 1.2632298790110468,
|
| 17 |
+
"global_step": 6000,
|
| 18 |
+
"eval_loss": 0.6355894207954407,
|
| 19 |
+
"eval_cosine_accuracy": 0.9509358406066895
|
| 20 |
+
},
|
| 21 |
+
{
|
| 22 |
+
"epoch": 1.6840610205155182,
|
| 23 |
+
"global_step": 8000,
|
| 24 |
+
"eval_loss": 0.6007522940635681,
|
| 25 |
+
"eval_cosine_accuracy": 0.9561171531677246
|
| 26 |
+
},
|
| 27 |
+
{
|
| 28 |
+
"epoch": 2.1047869542346134,
|
| 29 |
+
"global_step": 10000,
|
| 30 |
+
"eval_loss": 0.5822440385818481,
|
| 31 |
+
"eval_cosine_accuracy": 0.9581262469291687
|
| 32 |
+
},
|
| 33 |
+
{
|
| 34 |
+
"epoch": 2.525618095739085,
|
| 35 |
+
"global_step": 12000,
|
| 36 |
+
"eval_loss": 0.5283726453781128,
|
| 37 |
+
"eval_cosine_accuracy": 0.9619329571723938
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"epoch": 2.946449237243556,
|
| 41 |
+
"global_step": 14000,
|
| 42 |
+
"eval_loss": 0.5352063775062561,
|
| 43 |
+
"eval_cosine_accuracy": 0.9632018804550171
|
| 44 |
+
},
|
| 45 |
+
{
|
| 46 |
+
"epoch": 3.367175170962651,
|
| 47 |
+
"global_step": 16000,
|
| 48 |
+
"eval_loss": 0.5300219058990479,
|
| 49 |
+
"eval_cosine_accuracy": 0.9644707441329956
|
| 50 |
+
},
|
| 51 |
+
{
|
| 52 |
+
"epoch": 3.7880063124671226,
|
| 53 |
+
"global_step": 18000,
|
| 54 |
+
"eval_loss": 0.5290876626968384,
|
| 55 |
+
"eval_cosine_accuracy": 0.9654224514961243
|
| 56 |
+
},
|
| 57 |
+
{
|
| 58 |
+
"epoch": 4.2087322461862176,
|
| 59 |
+
"global_step": 20000,
|
| 60 |
+
"eval_loss": 0.5150468945503235,
|
| 61 |
+
"eval_cosine_accuracy": 0.9654224514961243
|
| 62 |
+
},
|
| 63 |
+
{
|
| 64 |
+
"epoch": 4.629563387690689,
|
| 65 |
+
"global_step": 22000,
|
| 66 |
+
"eval_loss": 0.508901834487915,
|
| 67 |
+
"eval_cosine_accuracy": 0.9681717157363892
|
| 68 |
+
},
|
| 69 |
+
{
|
| 70 |
+
"epoch": 5.050289321409784,
|
| 71 |
+
"global_step": 24000,
|
| 72 |
+
"eval_loss": 0.5077974200248718,
|
| 73 |
+
"eval_cosine_accuracy": 0.9665855765342712
|
| 74 |
+
},
|
| 75 |
+
{
|
| 76 |
+
"epoch": 5.471120462914255,
|
| 77 |
+
"global_step": 26000,
|
| 78 |
+
"eval_loss": 0.5082845687866211,
|
| 79 |
+
"eval_cosine_accuracy": 0.9678544998168945
|
| 80 |
+
},
|
| 81 |
+
{
|
| 82 |
+
"epoch": 5.891951604418727,
|
| 83 |
+
"global_step": 28000,
|
| 84 |
+
"eval_loss": 0.5089024305343628,
|
| 85 |
+
"eval_cosine_accuracy": 0.9674315452575684
|
| 86 |
+
}
|
| 87 |
+
],
|
| 88 |
+
"total_epochs": 5.995896896370331,
|
| 89 |
+
"total_steps": 28494
|
| 90 |
+
}
|