Upload trainer_state.json
Browse files- trainer_state.json +78 -0
trainer_state.json
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 5.347593582887701,
|
| 5 |
+
"global_step": 1000,
|
| 6 |
+
"is_hyper_param_search": false,
|
| 7 |
+
"is_local_process_zero": true,
|
| 8 |
+
"is_world_process_zero": true,
|
| 9 |
+
"log_history": [
|
| 10 |
+
{
|
| 11 |
+
"epoch": 1.07,
|
| 12 |
+
"eval_accuracy": 0.8663101604278075,
|
| 13 |
+
"eval_loss": 0.5277199149131775,
|
| 14 |
+
"eval_recall": 0.6122448979591837,
|
| 15 |
+
"eval_runtime": 10.117,
|
| 16 |
+
"eval_samples_per_second": 18.484,
|
| 17 |
+
"eval_steps_per_second": 4.646,
|
| 18 |
+
"step": 200
|
| 19 |
+
},
|
| 20 |
+
{
|
| 21 |
+
"epoch": 2.14,
|
| 22 |
+
"eval_accuracy": 0.7647058823529411,
|
| 23 |
+
"eval_loss": 0.9699290990829468,
|
| 24 |
+
"eval_recall": 0.8979591836734694,
|
| 25 |
+
"eval_runtime": 10.119,
|
| 26 |
+
"eval_samples_per_second": 18.48,
|
| 27 |
+
"eval_steps_per_second": 4.645,
|
| 28 |
+
"step": 400
|
| 29 |
+
},
|
| 30 |
+
{
|
| 31 |
+
"epoch": 2.67,
|
| 32 |
+
"learning_rate": 4.554367201426025e-05,
|
| 33 |
+
"loss": 0.4246,
|
| 34 |
+
"step": 500
|
| 35 |
+
},
|
| 36 |
+
{
|
| 37 |
+
"epoch": 3.21,
|
| 38 |
+
"eval_accuracy": 0.8663101604278075,
|
| 39 |
+
"eval_loss": 0.7973827719688416,
|
| 40 |
+
"eval_recall": 0.5510204081632653,
|
| 41 |
+
"eval_runtime": 10.148,
|
| 42 |
+
"eval_samples_per_second": 18.427,
|
| 43 |
+
"eval_steps_per_second": 4.631,
|
| 44 |
+
"step": 600
|
| 45 |
+
},
|
| 46 |
+
{
|
| 47 |
+
"epoch": 4.28,
|
| 48 |
+
"eval_accuracy": 0.8716577540106952,
|
| 49 |
+
"eval_loss": 0.9027533531188965,
|
| 50 |
+
"eval_recall": 0.5714285714285714,
|
| 51 |
+
"eval_runtime": 10.102,
|
| 52 |
+
"eval_samples_per_second": 18.511,
|
| 53 |
+
"eval_steps_per_second": 4.653,
|
| 54 |
+
"step": 800
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 5.35,
|
| 58 |
+
"learning_rate": 4.10873440285205e-05,
|
| 59 |
+
"loss": 0.1059,
|
| 60 |
+
"step": 1000
|
| 61 |
+
},
|
| 62 |
+
{
|
| 63 |
+
"epoch": 5.35,
|
| 64 |
+
"eval_accuracy": 0.8609625668449198,
|
| 65 |
+
"eval_loss": 0.8567710518836975,
|
| 66 |
+
"eval_recall": 0.7346938775510204,
|
| 67 |
+
"eval_runtime": 10.117,
|
| 68 |
+
"eval_samples_per_second": 18.484,
|
| 69 |
+
"eval_steps_per_second": 4.646,
|
| 70 |
+
"step": 1000
|
| 71 |
+
}
|
| 72 |
+
],
|
| 73 |
+
"max_steps": 5610,
|
| 74 |
+
"num_train_epochs": 30,
|
| 75 |
+
"total_flos": 1049850814279680.0,
|
| 76 |
+
"trial_name": null,
|
| 77 |
+
"trial_params": null
|
| 78 |
+
}
|