End of training
Browse files- all_results.json +18 -0
- eval_results.json +13 -0
- train_results.json +8 -0
- trainer_state.json +771 -0
all_results.json
ADDED
|
@@ -0,0 +1,18 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 2.0,
|
| 3 |
+
"eval_gen_len": 14.045010467550593,
|
| 4 |
+
"eval_loss": 1.3401455879211426,
|
| 5 |
+
"eval_rouge1": 52.201,
|
| 6 |
+
"eval_rouge2": 35.6154,
|
| 7 |
+
"eval_rougeL": 50.2334,
|
| 8 |
+
"eval_rougeLsum": 50.338,
|
| 9 |
+
"eval_runtime": 345.5591,
|
| 10 |
+
"eval_samples": 2866,
|
| 11 |
+
"eval_samples_per_second": 8.294,
|
| 12 |
+
"eval_steps_per_second": 8.294,
|
| 13 |
+
"train_loss": 1.7054625077800856,
|
| 14 |
+
"train_runtime": 2188.4584,
|
| 15 |
+
"train_samples": 23167,
|
| 16 |
+
"train_samples_per_second": 21.172,
|
| 17 |
+
"train_steps_per_second": 0.165
|
| 18 |
+
}
|
eval_results.json
ADDED
|
@@ -0,0 +1,13 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 2.0,
|
| 3 |
+
"eval_gen_len": 14.045010467550593,
|
| 4 |
+
"eval_loss": 1.3401455879211426,
|
| 5 |
+
"eval_rouge1": 52.201,
|
| 6 |
+
"eval_rouge2": 35.6154,
|
| 7 |
+
"eval_rougeL": 50.2334,
|
| 8 |
+
"eval_rougeLsum": 50.338,
|
| 9 |
+
"eval_runtime": 345.5591,
|
| 10 |
+
"eval_samples": 2866,
|
| 11 |
+
"eval_samples_per_second": 8.294,
|
| 12 |
+
"eval_steps_per_second": 8.294
|
| 13 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 2.0,
|
| 3 |
+
"train_loss": 1.7054625077800856,
|
| 4 |
+
"train_runtime": 2188.4584,
|
| 5 |
+
"train_samples": 23167,
|
| 6 |
+
"train_samples_per_second": 21.172,
|
| 7 |
+
"train_steps_per_second": 0.165
|
| 8 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,771 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 2.0,
|
| 5 |
+
"global_step": 362,
|
| 6 |
+
"is_hyper_param_search": false,
|
| 7 |
+
"is_local_process_zero": true,
|
| 8 |
+
"is_world_process_zero": true,
|
| 9 |
+
"log_history": [
|
| 10 |
+
{
|
| 11 |
+
"epoch": 0.02,
|
| 12 |
+
"learning_rate": 3.0000000000000004e-05,
|
| 13 |
+
"loss": 2.9193,
|
| 14 |
+
"step": 3
|
| 15 |
+
},
|
| 16 |
+
{
|
| 17 |
+
"epoch": 0.03,
|
| 18 |
+
"learning_rate": 6.000000000000001e-05,
|
| 19 |
+
"loss": 2.7709,
|
| 20 |
+
"step": 6
|
| 21 |
+
},
|
| 22 |
+
{
|
| 23 |
+
"epoch": 0.05,
|
| 24 |
+
"learning_rate": 7.999842485562489e-05,
|
| 25 |
+
"loss": 2.7758,
|
| 26 |
+
"step": 9
|
| 27 |
+
},
|
| 28 |
+
{
|
| 29 |
+
"epoch": 0.07,
|
| 30 |
+
"learning_rate": 7.997480017099975e-05,
|
| 31 |
+
"loss": 2.6219,
|
| 32 |
+
"step": 12
|
| 33 |
+
},
|
| 34 |
+
{
|
| 35 |
+
"epoch": 0.08,
|
| 36 |
+
"learning_rate": 7.992284223732899e-05,
|
| 37 |
+
"loss": 2.3928,
|
| 38 |
+
"step": 15
|
| 39 |
+
},
|
| 40 |
+
{
|
| 41 |
+
"epoch": 0.1,
|
| 42 |
+
"learning_rate": 7.98425878812401e-05,
|
| 43 |
+
"loss": 2.3238,
|
| 44 |
+
"step": 18
|
| 45 |
+
},
|
| 46 |
+
{
|
| 47 |
+
"epoch": 0.12,
|
| 48 |
+
"learning_rate": 7.973409398523528e-05,
|
| 49 |
+
"loss": 2.2624,
|
| 50 |
+
"step": 21
|
| 51 |
+
},
|
| 52 |
+
{
|
| 53 |
+
"epoch": 0.13,
|
| 54 |
+
"learning_rate": 7.959743744737432e-05,
|
| 55 |
+
"loss": 2.1201,
|
| 56 |
+
"step": 24
|
| 57 |
+
},
|
| 58 |
+
{
|
| 59 |
+
"epoch": 0.15,
|
| 60 |
+
"learning_rate": 7.943271512677111e-05,
|
| 61 |
+
"loss": 2.1615,
|
| 62 |
+
"step": 27
|
| 63 |
+
},
|
| 64 |
+
{
|
| 65 |
+
"epoch": 0.17,
|
| 66 |
+
"learning_rate": 7.924004377494192e-05,
|
| 67 |
+
"loss": 2.0752,
|
| 68 |
+
"step": 30
|
| 69 |
+
},
|
| 70 |
+
{
|
| 71 |
+
"epoch": 0.18,
|
| 72 |
+
"learning_rate": 7.901955995305456e-05,
|
| 73 |
+
"loss": 2.0268,
|
| 74 |
+
"step": 33
|
| 75 |
+
},
|
| 76 |
+
{
|
| 77 |
+
"epoch": 0.2,
|
| 78 |
+
"learning_rate": 7.877141993513694e-05,
|
| 79 |
+
"loss": 2.1183,
|
| 80 |
+
"step": 36
|
| 81 |
+
},
|
| 82 |
+
{
|
| 83 |
+
"epoch": 0.22,
|
| 84 |
+
"learning_rate": 7.849579959731341e-05,
|
| 85 |
+
"loss": 1.9541,
|
| 86 |
+
"step": 39
|
| 87 |
+
},
|
| 88 |
+
{
|
| 89 |
+
"epoch": 0.23,
|
| 90 |
+
"learning_rate": 7.819289429314769e-05,
|
| 91 |
+
"loss": 1.8978,
|
| 92 |
+
"step": 42
|
| 93 |
+
},
|
| 94 |
+
{
|
| 95 |
+
"epoch": 0.25,
|
| 96 |
+
"learning_rate": 7.786291871518068e-05,
|
| 97 |
+
"loss": 1.9057,
|
| 98 |
+
"step": 45
|
| 99 |
+
},
|
| 100 |
+
{
|
| 101 |
+
"epoch": 0.27,
|
| 102 |
+
"learning_rate": 7.750610674276111e-05,
|
| 103 |
+
"loss": 1.8962,
|
| 104 |
+
"step": 48
|
| 105 |
+
},
|
| 106 |
+
{
|
| 107 |
+
"epoch": 0.28,
|
| 108 |
+
"learning_rate": 7.712271127627709e-05,
|
| 109 |
+
"loss": 1.8923,
|
| 110 |
+
"step": 51
|
| 111 |
+
},
|
| 112 |
+
{
|
| 113 |
+
"epoch": 0.3,
|
| 114 |
+
"learning_rate": 7.671300405790588e-05,
|
| 115 |
+
"loss": 1.9072,
|
| 116 |
+
"step": 54
|
| 117 |
+
},
|
| 118 |
+
{
|
| 119 |
+
"epoch": 0.31,
|
| 120 |
+
"learning_rate": 7.627727547900913e-05,
|
| 121 |
+
"loss": 1.8607,
|
| 122 |
+
"step": 57
|
| 123 |
+
},
|
| 124 |
+
{
|
| 125 |
+
"epoch": 0.33,
|
| 126 |
+
"learning_rate": 7.581583437430996e-05,
|
| 127 |
+
"loss": 1.8037,
|
| 128 |
+
"step": 60
|
| 129 |
+
},
|
| 130 |
+
{
|
| 131 |
+
"epoch": 0.35,
|
| 132 |
+
"learning_rate": 7.532900780299777e-05,
|
| 133 |
+
"loss": 1.8416,
|
| 134 |
+
"step": 63
|
| 135 |
+
},
|
| 136 |
+
{
|
| 137 |
+
"epoch": 0.36,
|
| 138 |
+
"learning_rate": 7.481714081691589e-05,
|
| 139 |
+
"loss": 1.8028,
|
| 140 |
+
"step": 66
|
| 141 |
+
},
|
| 142 |
+
{
|
| 143 |
+
"epoch": 0.38,
|
| 144 |
+
"learning_rate": 7.428059621599662e-05,
|
| 145 |
+
"loss": 1.7381,
|
| 146 |
+
"step": 69
|
| 147 |
+
},
|
| 148 |
+
{
|
| 149 |
+
"epoch": 0.4,
|
| 150 |
+
"learning_rate": 7.371975429111662e-05,
|
| 151 |
+
"loss": 1.7571,
|
| 152 |
+
"step": 72
|
| 153 |
+
},
|
| 154 |
+
{
|
| 155 |
+
"epoch": 0.41,
|
| 156 |
+
"learning_rate": 7.313501255455524e-05,
|
| 157 |
+
"loss": 1.7261,
|
| 158 |
+
"step": 75
|
| 159 |
+
},
|
| 160 |
+
{
|
| 161 |
+
"epoch": 0.43,
|
| 162 |
+
"learning_rate": 7.252678545824668e-05,
|
| 163 |
+
"loss": 1.7381,
|
| 164 |
+
"step": 78
|
| 165 |
+
},
|
| 166 |
+
{
|
| 167 |
+
"epoch": 0.45,
|
| 168 |
+
"learning_rate": 7.189550410002563e-05,
|
| 169 |
+
"loss": 1.8021,
|
| 170 |
+
"step": 81
|
| 171 |
+
},
|
| 172 |
+
{
|
| 173 |
+
"epoch": 0.46,
|
| 174 |
+
"learning_rate": 7.124161591807473e-05,
|
| 175 |
+
"loss": 1.7353,
|
| 176 |
+
"step": 84
|
| 177 |
+
},
|
| 178 |
+
{
|
| 179 |
+
"epoch": 0.48,
|
| 180 |
+
"learning_rate": 7.056558437379036e-05,
|
| 181 |
+
"loss": 1.7356,
|
| 182 |
+
"step": 87
|
| 183 |
+
},
|
| 184 |
+
{
|
| 185 |
+
"epoch": 0.5,
|
| 186 |
+
"learning_rate": 6.986788862329151e-05,
|
| 187 |
+
"loss": 1.6188,
|
| 188 |
+
"step": 90
|
| 189 |
+
},
|
| 190 |
+
{
|
| 191 |
+
"epoch": 0.51,
|
| 192 |
+
"learning_rate": 6.914902317780456e-05,
|
| 193 |
+
"loss": 1.7618,
|
| 194 |
+
"step": 93
|
| 195 |
+
},
|
| 196 |
+
{
|
| 197 |
+
"epoch": 0.53,
|
| 198 |
+
"learning_rate": 6.840949755316473e-05,
|
| 199 |
+
"loss": 1.745,
|
| 200 |
+
"step": 96
|
| 201 |
+
},
|
| 202 |
+
{
|
| 203 |
+
"epoch": 0.55,
|
| 204 |
+
"learning_rate": 6.764983590868262e-05,
|
| 205 |
+
"loss": 1.6881,
|
| 206 |
+
"step": 99
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"epoch": 0.56,
|
| 210 |
+
"learning_rate": 6.687057667563171e-05,
|
| 211 |
+
"loss": 1.6521,
|
| 212 |
+
"step": 102
|
| 213 |
+
},
|
| 214 |
+
{
|
| 215 |
+
"epoch": 0.58,
|
| 216 |
+
"learning_rate": 6.60722721756203e-05,
|
| 217 |
+
"loss": 1.6195,
|
| 218 |
+
"step": 105
|
| 219 |
+
},
|
| 220 |
+
{
|
| 221 |
+
"epoch": 0.6,
|
| 222 |
+
"learning_rate": 6.525548822911814e-05,
|
| 223 |
+
"loss": 1.6723,
|
| 224 |
+
"step": 108
|
| 225 |
+
},
|
| 226 |
+
{
|
| 227 |
+
"epoch": 0.61,
|
| 228 |
+
"learning_rate": 6.442080375441567e-05,
|
| 229 |
+
"loss": 1.6623,
|
| 230 |
+
"step": 111
|
| 231 |
+
},
|
| 232 |
+
{
|
| 233 |
+
"epoch": 0.63,
|
| 234 |
+
"learning_rate": 6.356881035729934e-05,
|
| 235 |
+
"loss": 1.6837,
|
| 236 |
+
"step": 114
|
| 237 |
+
},
|
| 238 |
+
{
|
| 239 |
+
"epoch": 0.65,
|
| 240 |
+
"learning_rate": 6.270011191173488e-05,
|
| 241 |
+
"loss": 1.7298,
|
| 242 |
+
"step": 117
|
| 243 |
+
},
|
| 244 |
+
{
|
| 245 |
+
"epoch": 0.66,
|
| 246 |
+
"learning_rate": 6.181532413185459e-05,
|
| 247 |
+
"loss": 1.6268,
|
| 248 |
+
"step": 120
|
| 249 |
+
},
|
| 250 |
+
{
|
| 251 |
+
"epoch": 0.68,
|
| 252 |
+
"learning_rate": 6.091507413555311e-05,
|
| 253 |
+
"loss": 1.6234,
|
| 254 |
+
"step": 123
|
| 255 |
+
},
|
| 256 |
+
{
|
| 257 |
+
"epoch": 0.7,
|
| 258 |
+
"learning_rate": 6.000000000000001e-05,
|
| 259 |
+
"loss": 1.6908,
|
| 260 |
+
"step": 126
|
| 261 |
+
},
|
| 262 |
+
{
|
| 263 |
+
"epoch": 0.71,
|
| 264 |
+
"learning_rate": 5.907075030938501e-05,
|
| 265 |
+
"loss": 1.7054,
|
| 266 |
+
"step": 129
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"epoch": 0.73,
|
| 270 |
+
"learning_rate": 5.8127983695215895e-05,
|
| 271 |
+
"loss": 1.6153,
|
| 272 |
+
"step": 132
|
| 273 |
+
},
|
| 274 |
+
{
|
| 275 |
+
"epoch": 0.75,
|
| 276 |
+
"learning_rate": 5.7172368369495146e-05,
|
| 277 |
+
"loss": 1.5946,
|
| 278 |
+
"step": 135
|
| 279 |
+
},
|
| 280 |
+
{
|
| 281 |
+
"epoch": 0.76,
|
| 282 |
+
"learning_rate": 5.6204581651106144e-05,
|
| 283 |
+
"loss": 1.7282,
|
| 284 |
+
"step": 138
|
| 285 |
+
},
|
| 286 |
+
{
|
| 287 |
+
"epoch": 0.78,
|
| 288 |
+
"learning_rate": 5.5225309485744636e-05,
|
| 289 |
+
"loss": 1.5865,
|
| 290 |
+
"step": 141
|
| 291 |
+
},
|
| 292 |
+
{
|
| 293 |
+
"epoch": 0.8,
|
| 294 |
+
"learning_rate": 5.4235245959735636e-05,
|
| 295 |
+
"loss": 1.5655,
|
| 296 |
+
"step": 144
|
| 297 |
+
},
|
| 298 |
+
{
|
| 299 |
+
"epoch": 0.81,
|
| 300 |
+
"learning_rate": 5.323509280808056e-05,
|
| 301 |
+
"loss": 1.665,
|
| 302 |
+
"step": 147
|
| 303 |
+
},
|
| 304 |
+
{
|
| 305 |
+
"epoch": 0.83,
|
| 306 |
+
"learning_rate": 5.222555891708296e-05,
|
| 307 |
+
"loss": 1.6115,
|
| 308 |
+
"step": 150
|
| 309 |
+
},
|
| 310 |
+
{
|
| 311 |
+
"epoch": 0.85,
|
| 312 |
+
"learning_rate": 5.120735982190578e-05,
|
| 313 |
+
"loss": 1.5842,
|
| 314 |
+
"step": 153
|
| 315 |
+
},
|
| 316 |
+
{
|
| 317 |
+
"epoch": 0.86,
|
| 318 |
+
"learning_rate": 5.018121719941591e-05,
|
| 319 |
+
"loss": 1.5517,
|
| 320 |
+
"step": 156
|
| 321 |
+
},
|
| 322 |
+
{
|
| 323 |
+
"epoch": 0.88,
|
| 324 |
+
"learning_rate": 4.914785835667567e-05,
|
| 325 |
+
"loss": 1.5884,
|
| 326 |
+
"step": 159
|
| 327 |
+
},
|
| 328 |
+
{
|
| 329 |
+
"epoch": 0.9,
|
| 330 |
+
"learning_rate": 4.8108015715443745e-05,
|
| 331 |
+
"loss": 1.6255,
|
| 332 |
+
"step": 162
|
| 333 |
+
},
|
| 334 |
+
{
|
| 335 |
+
"epoch": 0.91,
|
| 336 |
+
"learning_rate": 4.706242629305098e-05,
|
| 337 |
+
"loss": 1.6402,
|
| 338 |
+
"step": 165
|
| 339 |
+
},
|
| 340 |
+
{
|
| 341 |
+
"epoch": 0.93,
|
| 342 |
+
"learning_rate": 4.6011831180018825e-05,
|
| 343 |
+
"loss": 1.6405,
|
| 344 |
+
"step": 168
|
| 345 |
+
},
|
| 346 |
+
{
|
| 347 |
+
"epoch": 0.94,
|
| 348 |
+
"learning_rate": 4.495697501479091e-05,
|
| 349 |
+
"loss": 1.6478,
|
| 350 |
+
"step": 171
|
| 351 |
+
},
|
| 352 |
+
{
|
| 353 |
+
"epoch": 0.96,
|
| 354 |
+
"learning_rate": 4.389860545594986e-05,
|
| 355 |
+
"loss": 1.6635,
|
| 356 |
+
"step": 174
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 0.98,
|
| 360 |
+
"learning_rate": 4.283747265229354e-05,
|
| 361 |
+
"loss": 1.6468,
|
| 362 |
+
"step": 177
|
| 363 |
+
},
|
| 364 |
+
{
|
| 365 |
+
"epoch": 0.99,
|
| 366 |
+
"learning_rate": 4.177432871114626e-05,
|
| 367 |
+
"loss": 1.6161,
|
| 368 |
+
"step": 180
|
| 369 |
+
},
|
| 370 |
+
{
|
| 371 |
+
"epoch": 1.0,
|
| 372 |
+
"eval_gen_len": 13.835659455687368,
|
| 373 |
+
"eval_loss": 1.3714126348495483,
|
| 374 |
+
"eval_rouge1": 51.1003,
|
| 375 |
+
"eval_rouge2": 34.5701,
|
| 376 |
+
"eval_rougeL": 49.1277,
|
| 377 |
+
"eval_rougeLsum": 49.2466,
|
| 378 |
+
"eval_runtime": 341.0183,
|
| 379 |
+
"eval_samples_per_second": 8.404,
|
| 380 |
+
"eval_steps_per_second": 8.404,
|
| 381 |
+
"step": 181
|
| 382 |
+
},
|
| 383 |
+
{
|
| 384 |
+
"epoch": 1.01,
|
| 385 |
+
"learning_rate": 4.070992716528185e-05,
|
| 386 |
+
"loss": 1.6585,
|
| 387 |
+
"step": 183
|
| 388 |
+
},
|
| 389 |
+
{
|
| 390 |
+
"epoch": 1.03,
|
| 391 |
+
"learning_rate": 3.9645022438836356e-05,
|
| 392 |
+
"loss": 1.5792,
|
| 393 |
+
"step": 186
|
| 394 |
+
},
|
| 395 |
+
{
|
| 396 |
+
"epoch": 1.04,
|
| 397 |
+
"learning_rate": 3.858036931258908e-05,
|
| 398 |
+
"loss": 1.5529,
|
| 399 |
+
"step": 189
|
| 400 |
+
},
|
| 401 |
+
{
|
| 402 |
+
"epoch": 1.06,
|
| 403 |
+
"learning_rate": 3.7516722388990645e-05,
|
| 404 |
+
"loss": 1.5196,
|
| 405 |
+
"step": 192
|
| 406 |
+
},
|
| 407 |
+
{
|
| 408 |
+
"epoch": 1.08,
|
| 409 |
+
"learning_rate": 3.645483555731765e-05,
|
| 410 |
+
"loss": 1.655,
|
| 411 |
+
"step": 195
|
| 412 |
+
},
|
| 413 |
+
{
|
| 414 |
+
"epoch": 1.09,
|
| 415 |
+
"learning_rate": 3.539546145933268e-05,
|
| 416 |
+
"loss": 1.5744,
|
| 417 |
+
"step": 198
|
| 418 |
+
},
|
| 419 |
+
{
|
| 420 |
+
"epoch": 1.11,
|
| 421 |
+
"learning_rate": 3.433935095582862e-05,
|
| 422 |
+
"loss": 1.5734,
|
| 423 |
+
"step": 201
|
| 424 |
+
},
|
| 425 |
+
{
|
| 426 |
+
"epoch": 1.13,
|
| 427 |
+
"learning_rate": 3.3287252594435136e-05,
|
| 428 |
+
"loss": 1.5357,
|
| 429 |
+
"step": 204
|
| 430 |
+
},
|
| 431 |
+
{
|
| 432 |
+
"epoch": 1.14,
|
| 433 |
+
"learning_rate": 3.2239912079064877e-05,
|
| 434 |
+
"loss": 1.5575,
|
| 435 |
+
"step": 207
|
| 436 |
+
},
|
| 437 |
+
{
|
| 438 |
+
"epoch": 1.16,
|
| 439 |
+
"learning_rate": 3.119807174137503e-05,
|
| 440 |
+
"loss": 1.5703,
|
| 441 |
+
"step": 210
|
| 442 |
+
},
|
| 443 |
+
{
|
| 444 |
+
"epoch": 1.18,
|
| 445 |
+
"learning_rate": 3.0162470014619215e-05,
|
| 446 |
+
"loss": 1.6525,
|
| 447 |
+
"step": 213
|
| 448 |
+
},
|
| 449 |
+
{
|
| 450 |
+
"epoch": 1.19,
|
| 451 |
+
"learning_rate": 2.9133840910262457e-05,
|
| 452 |
+
"loss": 1.5807,
|
| 453 |
+
"step": 216
|
| 454 |
+
},
|
| 455 |
+
{
|
| 456 |
+
"epoch": 1.21,
|
| 457 |
+
"learning_rate": 2.811291349773009e-05,
|
| 458 |
+
"loss": 1.6189,
|
| 459 |
+
"step": 219
|
| 460 |
+
},
|
| 461 |
+
{
|
| 462 |
+
"epoch": 1.23,
|
| 463 |
+
"learning_rate": 2.7100411387659647e-05,
|
| 464 |
+
"loss": 1.6253,
|
| 465 |
+
"step": 222
|
| 466 |
+
},
|
| 467 |
+
{
|
| 468 |
+
"epoch": 1.24,
|
| 469 |
+
"learning_rate": 2.609705221902172e-05,
|
| 470 |
+
"loss": 1.552,
|
| 471 |
+
"step": 225
|
| 472 |
+
},
|
| 473 |
+
{
|
| 474 |
+
"epoch": 1.26,
|
| 475 |
+
"learning_rate": 2.5103547150473356e-05,
|
| 476 |
+
"loss": 1.6,
|
| 477 |
+
"step": 228
|
| 478 |
+
},
|
| 479 |
+
{
|
| 480 |
+
"epoch": 1.28,
|
| 481 |
+
"learning_rate": 2.4120600356304796e-05,
|
| 482 |
+
"loss": 1.5532,
|
| 483 |
+
"step": 231
|
| 484 |
+
},
|
| 485 |
+
{
|
| 486 |
+
"epoch": 1.29,
|
| 487 |
+
"learning_rate": 2.3148908527336174e-05,
|
| 488 |
+
"loss": 1.5704,
|
| 489 |
+
"step": 234
|
| 490 |
+
},
|
| 491 |
+
{
|
| 492 |
+
"epoch": 1.31,
|
| 493 |
+
"learning_rate": 2.2189160377118748e-05,
|
| 494 |
+
"loss": 1.5772,
|
| 495 |
+
"step": 237
|
| 496 |
+
},
|
| 497 |
+
{
|
| 498 |
+
"epoch": 1.33,
|
| 499 |
+
"learning_rate": 2.1242036153789968e-05,
|
| 500 |
+
"loss": 1.6103,
|
| 501 |
+
"step": 240
|
| 502 |
+
},
|
| 503 |
+
{
|
| 504 |
+
"epoch": 1.34,
|
| 505 |
+
"learning_rate": 2.0308207157928726e-05,
|
| 506 |
+
"loss": 1.503,
|
| 507 |
+
"step": 243
|
| 508 |
+
},
|
| 509 |
+
{
|
| 510 |
+
"epoch": 1.36,
|
| 511 |
+
"learning_rate": 1.9388335266752492e-05,
|
| 512 |
+
"loss": 1.5573,
|
| 513 |
+
"step": 246
|
| 514 |
+
},
|
| 515 |
+
{
|
| 516 |
+
"epoch": 1.38,
|
| 517 |
+
"learning_rate": 1.8483072464993447e-05,
|
| 518 |
+
"loss": 1.6,
|
| 519 |
+
"step": 249
|
| 520 |
+
},
|
| 521 |
+
{
|
| 522 |
+
"epoch": 1.39,
|
| 523 |
+
"learning_rate": 1.7593060382786156e-05,
|
| 524 |
+
"loss": 1.5964,
|
| 525 |
+
"step": 252
|
| 526 |
+
},
|
| 527 |
+
{
|
| 528 |
+
"epoch": 1.41,
|
| 529 |
+
"learning_rate": 1.6718929840894502e-05,
|
| 530 |
+
"loss": 1.6025,
|
| 531 |
+
"step": 255
|
| 532 |
+
},
|
| 533 |
+
{
|
| 534 |
+
"epoch": 1.43,
|
| 535 |
+
"learning_rate": 1.5861300403599846e-05,
|
| 536 |
+
"loss": 1.5649,
|
| 537 |
+
"step": 258
|
| 538 |
+
},
|
| 539 |
+
{
|
| 540 |
+
"epoch": 1.44,
|
| 541 |
+
"learning_rate": 1.5020779939567773e-05,
|
| 542 |
+
"loss": 1.5552,
|
| 543 |
+
"step": 261
|
| 544 |
+
},
|
| 545 |
+
{
|
| 546 |
+
"epoch": 1.46,
|
| 547 |
+
"learning_rate": 1.4197964191004178e-05,
|
| 548 |
+
"loss": 1.4538,
|
| 549 |
+
"step": 264
|
| 550 |
+
},
|
| 551 |
+
{
|
| 552 |
+
"epoch": 1.48,
|
| 553 |
+
"learning_rate": 1.3393436351406575e-05,
|
| 554 |
+
"loss": 1.4979,
|
| 555 |
+
"step": 267
|
| 556 |
+
},
|
| 557 |
+
{
|
| 558 |
+
"epoch": 1.49,
|
| 559 |
+
"learning_rate": 1.260776665220938e-05,
|
| 560 |
+
"loss": 1.539,
|
| 561 |
+
"step": 270
|
| 562 |
+
},
|
| 563 |
+
{
|
| 564 |
+
"epoch": 1.51,
|
| 565 |
+
"learning_rate": 1.1841511958616682e-05,
|
| 566 |
+
"loss": 1.5912,
|
| 567 |
+
"step": 273
|
| 568 |
+
},
|
| 569 |
+
{
|
| 570 |
+
"epoch": 1.52,
|
| 571 |
+
"learning_rate": 1.1095215374908515e-05,
|
| 572 |
+
"loss": 1.6084,
|
| 573 |
+
"step": 276
|
| 574 |
+
},
|
| 575 |
+
{
|
| 576 |
+
"epoch": 1.54,
|
| 577 |
+
"learning_rate": 1.0369405859500615e-05,
|
| 578 |
+
"loss": 1.5597,
|
| 579 |
+
"step": 279
|
| 580 |
+
},
|
| 581 |
+
{
|
| 582 |
+
"epoch": 1.56,
|
| 583 |
+
"learning_rate": 9.664597850030465e-06,
|
| 584 |
+
"loss": 1.6007,
|
| 585 |
+
"step": 282
|
| 586 |
+
},
|
| 587 |
+
{
|
| 588 |
+
"epoch": 1.57,
|
| 589 |
+
"learning_rate": 8.981290898735416e-06,
|
| 590 |
+
"loss": 1.593,
|
| 591 |
+
"step": 285
|
| 592 |
+
},
|
| 593 |
+
{
|
| 594 |
+
"epoch": 1.59,
|
| 595 |
+
"learning_rate": 8.319969318381052e-06,
|
| 596 |
+
"loss": 1.6508,
|
| 597 |
+
"step": 288
|
| 598 |
+
},
|
| 599 |
+
{
|
| 600 |
+
"epoch": 1.61,
|
| 601 |
+
"learning_rate": 7.681101838991184e-06,
|
| 602 |
+
"loss": 1.5349,
|
| 603 |
+
"step": 291
|
| 604 |
+
},
|
| 605 |
+
{
|
| 606 |
+
"epoch": 1.62,
|
| 607 |
+
"learning_rate": 7.0651412756223806e-06,
|
| 608 |
+
"loss": 1.4949,
|
| 609 |
+
"step": 294
|
| 610 |
+
},
|
| 611 |
+
{
|
| 612 |
+
"epoch": 1.64,
|
| 613 |
+
"learning_rate": 6.472524207418742e-06,
|
| 614 |
+
"loss": 1.5844,
|
| 615 |
+
"step": 297
|
| 616 |
+
},
|
| 617 |
+
{
|
| 618 |
+
"epoch": 1.66,
|
| 619 |
+
"learning_rate": 5.9036706681743e-06,
|
| 620 |
+
"loss": 1.5007,
|
| 621 |
+
"step": 300
|
| 622 |
+
},
|
| 623 |
+
{
|
| 624 |
+
"epoch": 1.67,
|
| 625 |
+
"learning_rate": 5.358983848622452e-06,
|
| 626 |
+
"loss": 1.5578,
|
| 627 |
+
"step": 303
|
| 628 |
+
},
|
| 629 |
+
{
|
| 630 |
+
"epoch": 1.69,
|
| 631 |
+
"learning_rate": 4.838849810663311e-06,
|
| 632 |
+
"loss": 1.533,
|
| 633 |
+
"step": 306
|
| 634 |
+
},
|
| 635 |
+
{
|
| 636 |
+
"epoch": 1.71,
|
| 637 |
+
"learning_rate": 4.343637213731589e-06,
|
| 638 |
+
"loss": 1.5839,
|
| 639 |
+
"step": 309
|
| 640 |
+
},
|
| 641 |
+
{
|
| 642 |
+
"epoch": 1.72,
|
| 643 |
+
"learning_rate": 3.873697053499057e-06,
|
| 644 |
+
"loss": 1.5141,
|
| 645 |
+
"step": 312
|
| 646 |
+
},
|
| 647 |
+
{
|
| 648 |
+
"epoch": 1.74,
|
| 649 |
+
"learning_rate": 3.4293624130965754e-06,
|
| 650 |
+
"loss": 1.5297,
|
| 651 |
+
"step": 315
|
| 652 |
+
},
|
| 653 |
+
{
|
| 654 |
+
"epoch": 1.76,
|
| 655 |
+
"learning_rate": 3.0109482270321665e-06,
|
| 656 |
+
"loss": 1.5257,
|
| 657 |
+
"step": 318
|
| 658 |
+
},
|
| 659 |
+
{
|
| 660 |
+
"epoch": 1.77,
|
| 661 |
+
"learning_rate": 2.618751057972526e-06,
|
| 662 |
+
"loss": 1.664,
|
| 663 |
+
"step": 321
|
| 664 |
+
},
|
| 665 |
+
{
|
| 666 |
+
"epoch": 1.79,
|
| 667 |
+
"learning_rate": 2.2530488865458945e-06,
|
| 668 |
+
"loss": 1.6515,
|
| 669 |
+
"step": 324
|
| 670 |
+
},
|
| 671 |
+
{
|
| 672 |
+
"epoch": 1.81,
|
| 673 |
+
"learning_rate": 1.914100914315631e-06,
|
| 674 |
+
"loss": 1.6336,
|
| 675 |
+
"step": 327
|
| 676 |
+
},
|
| 677 |
+
{
|
| 678 |
+
"epoch": 1.82,
|
| 679 |
+
"learning_rate": 1.602147380063923e-06,
|
| 680 |
+
"loss": 1.5623,
|
| 681 |
+
"step": 330
|
| 682 |
+
},
|
| 683 |
+
{
|
| 684 |
+
"epoch": 1.84,
|
| 685 |
+
"learning_rate": 1.3174093895158335e-06,
|
| 686 |
+
"loss": 1.5331,
|
| 687 |
+
"step": 333
|
| 688 |
+
},
|
| 689 |
+
{
|
| 690 |
+
"epoch": 1.86,
|
| 691 |
+
"learning_rate": 1.0600887586245334e-06,
|
| 692 |
+
"loss": 1.5105,
|
| 693 |
+
"step": 336
|
| 694 |
+
},
|
| 695 |
+
{
|
| 696 |
+
"epoch": 1.87,
|
| 697 |
+
"learning_rate": 8.303678705286145e-07,
|
| 698 |
+
"loss": 1.4993,
|
| 699 |
+
"step": 339
|
| 700 |
+
},
|
| 701 |
+
{
|
| 702 |
+
"epoch": 1.89,
|
| 703 |
+
"learning_rate": 6.284095462829687e-07,
|
| 704 |
+
"loss": 1.587,
|
| 705 |
+
"step": 342
|
| 706 |
+
},
|
| 707 |
+
{
|
| 708 |
+
"epoch": 1.91,
|
| 709 |
+
"learning_rate": 4.543569294548755e-07,
|
| 710 |
+
"loss": 1.5219,
|
| 711 |
+
"step": 345
|
| 712 |
+
},
|
| 713 |
+
{
|
| 714 |
+
"epoch": 1.92,
|
| 715 |
+
"learning_rate": 3.083333846670167e-07,
|
| 716 |
+
"loss": 1.5665,
|
| 717 |
+
"step": 348
|
| 718 |
+
},
|
| 719 |
+
{
|
| 720 |
+
"epoch": 1.94,
|
| 721 |
+
"learning_rate": 1.9044241015936603e-07,
|
| 722 |
+
"loss": 1.5708,
|
| 723 |
+
"step": 351
|
| 724 |
+
},
|
| 725 |
+
{
|
| 726 |
+
"epoch": 1.96,
|
| 727 |
+
"learning_rate": 1.0076756443194057e-07,
|
| 728 |
+
"loss": 1.5092,
|
| 729 |
+
"step": 354
|
| 730 |
+
},
|
| 731 |
+
{
|
| 732 |
+
"epoch": 1.97,
|
| 733 |
+
"learning_rate": 3.9372407020406014e-08,
|
| 734 |
+
"loss": 1.5417,
|
| 735 |
+
"step": 357
|
| 736 |
+
},
|
| 737 |
+
{
|
| 738 |
+
"epoch": 1.99,
|
| 739 |
+
"learning_rate": 6.30045344648611e-09,
|
| 740 |
+
"loss": 1.539,
|
| 741 |
+
"step": 360
|
| 742 |
+
},
|
| 743 |
+
{
|
| 744 |
+
"epoch": 2.0,
|
| 745 |
+
"eval_gen_len": 14.045010467550593,
|
| 746 |
+
"eval_loss": 1.3401455879211426,
|
| 747 |
+
"eval_rouge1": 52.201,
|
| 748 |
+
"eval_rouge2": 35.6154,
|
| 749 |
+
"eval_rougeL": 50.2334,
|
| 750 |
+
"eval_rougeLsum": 50.338,
|
| 751 |
+
"eval_runtime": 345.6224,
|
| 752 |
+
"eval_samples_per_second": 8.292,
|
| 753 |
+
"eval_steps_per_second": 8.292,
|
| 754 |
+
"step": 362
|
| 755 |
+
},
|
| 756 |
+
{
|
| 757 |
+
"epoch": 2.0,
|
| 758 |
+
"step": 362,
|
| 759 |
+
"total_flos": 1.7226101285715968e+16,
|
| 760 |
+
"train_loss": 1.7054625077800856,
|
| 761 |
+
"train_runtime": 2188.4584,
|
| 762 |
+
"train_samples_per_second": 21.172,
|
| 763 |
+
"train_steps_per_second": 0.165
|
| 764 |
+
}
|
| 765 |
+
],
|
| 766 |
+
"max_steps": 362,
|
| 767 |
+
"num_train_epochs": 2,
|
| 768 |
+
"total_flos": 1.7226101285715968e+16,
|
| 769 |
+
"trial_name": null,
|
| 770 |
+
"trial_params": null
|
| 771 |
+
}
|