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": 0.18684603886397608,
|
| 5 |
+
"global_step": 2000,
|
| 6 |
+
"eval_loss": 0.49471989274024963,
|
| 7 |
+
"eval_cosine_accuracy": 0.8986994028091431
|
| 8 |
+
},
|
| 9 |
+
{
|
| 10 |
+
"epoch": 0.37369207772795215,
|
| 11 |
+
"global_step": 4000,
|
| 12 |
+
"eval_loss": 0.4586535394191742,
|
| 13 |
+
"eval_cosine_accuracy": 0.9260864853858948
|
| 14 |
+
},
|
| 15 |
+
{
|
| 16 |
+
"epoch": 0.5605381165919282,
|
| 17 |
+
"global_step": 6000,
|
| 18 |
+
"eval_loss": 0.41501784324645996,
|
| 19 |
+
"eval_cosine_accuracy": 0.9477635622024536
|
| 20 |
+
},
|
| 21 |
+
{
|
| 22 |
+
"epoch": 0.7473841554559043,
|
| 23 |
+
"global_step": 8000,
|
| 24 |
+
"eval_loss": 0.395673006772995,
|
| 25 |
+
"eval_cosine_accuracy": 0.9500898718833923
|
| 26 |
+
},
|
| 27 |
+
{
|
| 28 |
+
"epoch": 0.9342301943198804,
|
| 29 |
+
"global_step": 10000,
|
| 30 |
+
"eval_loss": 0.37747085094451904,
|
| 31 |
+
"eval_cosine_accuracy": 0.9434281587600708
|
| 32 |
+
},
|
| 33 |
+
{
|
| 34 |
+
"epoch": 1.1209406494960805,
|
| 35 |
+
"global_step": 12000,
|
| 36 |
+
"eval_loss": 0.36152419447898865,
|
| 37 |
+
"eval_cosine_accuracy": 0.958760678768158
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"epoch": 1.3075774542739829,
|
| 41 |
+
"global_step": 14000,
|
| 42 |
+
"eval_loss": 0.3521560728549957,
|
| 43 |
+
"eval_cosine_accuracy": 0.960769772529602
|
| 44 |
+
},
|
| 45 |
+
{
|
| 46 |
+
"epoch": 1.494214259051885,
|
| 47 |
+
"global_step": 16000,
|
| 48 |
+
"eval_loss": 0.3508636951446533,
|
| 49 |
+
"eval_cosine_accuracy": 0.9578090310096741
|
| 50 |
+
},
|
| 51 |
+
{
|
| 52 |
+
"epoch": 1.6808510638297873,
|
| 53 |
+
"global_step": 18000,
|
| 54 |
+
"eval_loss": 0.34350287914276123,
|
| 55 |
+
"eval_cosine_accuracy": 0.9620386958122253
|
| 56 |
+
},
|
| 57 |
+
{
|
| 58 |
+
"epoch": 1.8674878686076894,
|
| 59 |
+
"global_step": 20000,
|
| 60 |
+
"eval_loss": 0.3393498957157135,
|
| 61 |
+
"eval_cosine_accuracy": 0.959395170211792
|
| 62 |
+
},
|
| 63 |
+
{
|
| 64 |
+
"epoch": 2.0541246733855916,
|
| 65 |
+
"global_step": 22000,
|
| 66 |
+
"eval_loss": 0.33905890583992004,
|
| 67 |
+
"eval_cosine_accuracy": 0.9651052355766296
|
| 68 |
+
},
|
| 69 |
+
{
|
| 70 |
+
"epoch": 2.2407614781634937,
|
| 71 |
+
"global_step": 24000,
|
| 72 |
+
"eval_loss": 0.3363184630870819,
|
| 73 |
+
"eval_cosine_accuracy": 0.9655281901359558
|
| 74 |
+
},
|
| 75 |
+
{
|
| 76 |
+
"epoch": 2.427398282941396,
|
| 77 |
+
"global_step": 26000,
|
| 78 |
+
"eval_loss": 0.33477866649627686,
|
| 79 |
+
"eval_cosine_accuracy": 0.9634133577346802
|
| 80 |
+
},
|
| 81 |
+
{
|
| 82 |
+
"epoch": 2.6140350877192984,
|
| 83 |
+
"global_step": 28000,
|
| 84 |
+
"eval_loss": 0.33503878116607666,
|
| 85 |
+
"eval_cosine_accuracy": 0.9642592668533325
|
| 86 |
+
},
|
| 87 |
+
{
|
| 88 |
+
"epoch": 2.8006718924972005,
|
| 89 |
+
"global_step": 30000,
|
| 90 |
+
"eval_loss": 0.33804062008857727,
|
| 91 |
+
"eval_cosine_accuracy": 0.9646822214126587
|
| 92 |
+
},
|
| 93 |
+
{
|
| 94 |
+
"epoch": 2.9873086972751026,
|
| 95 |
+
"global_step": 32000,
|
| 96 |
+
"eval_loss": 0.33818456530570984,
|
| 97 |
+
"eval_cosine_accuracy": 0.9647879600524902
|
| 98 |
+
}
|
| 99 |
+
],
|
| 100 |
+
"total_epochs": 2.9977603583426653,
|
| 101 |
+
"total_steps": 32112
|
| 102 |
+
}
|