Upload trainer_state.json with huggingface_hub
Browse files- trainer_state.json +64 -0
trainer_state.json
ADDED
|
@@ -0,0 +1,64 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": 0.8626177714583143,
|
| 3 |
+
"best_model_checkpoint": "result/my-sup-simcse-bert-base-uncased/SupCon-28-100.0-3e-6-0.05-0.00",
|
| 4 |
+
"epoch": 2.9953703703703702,
|
| 5 |
+
"global_step": 645,
|
| 6 |
+
"is_hyper_param_search": false,
|
| 7 |
+
"is_local_process_zero": true,
|
| 8 |
+
"is_world_process_zero": true,
|
| 9 |
+
"log_history": [
|
| 10 |
+
{
|
| 11 |
+
"epoch": 0.58,
|
| 12 |
+
"eval_avg_sts": 0.834701718312026,
|
| 13 |
+
"eval_sickr_spearman": 0.817920809860896,
|
| 14 |
+
"eval_stsb_spearman": 0.851482626763156,
|
| 15 |
+
"step": 125
|
| 16 |
+
},
|
| 17 |
+
{
|
| 18 |
+
"epoch": 1.16,
|
| 19 |
+
"eval_avg_sts": 0.8365837614673681,
|
| 20 |
+
"eval_sickr_spearman": 0.8167967379939398,
|
| 21 |
+
"eval_stsb_spearman": 0.8563707849407965,
|
| 22 |
+
"step": 250
|
| 23 |
+
},
|
| 24 |
+
{
|
| 25 |
+
"epoch": 1.74,
|
| 26 |
+
"eval_avg_sts": 0.8407075744602279,
|
| 27 |
+
"eval_sickr_spearman": 0.8187973774621414,
|
| 28 |
+
"eval_stsb_spearman": 0.8626177714583143,
|
| 29 |
+
"step": 375
|
| 30 |
+
},
|
| 31 |
+
{
|
| 32 |
+
"epoch": 2.32,
|
| 33 |
+
"learning_rate": 1.1419753086419753e-05,
|
| 34 |
+
"loss": 0.7185,
|
| 35 |
+
"step": 500
|
| 36 |
+
},
|
| 37 |
+
{
|
| 38 |
+
"epoch": 2.32,
|
| 39 |
+
"eval_avg_sts": 0.8370980558240873,
|
| 40 |
+
"eval_sickr_spearman": 0.812430662842114,
|
| 41 |
+
"eval_stsb_spearman": 0.8617654488060604,
|
| 42 |
+
"step": 500
|
| 43 |
+
},
|
| 44 |
+
{
|
| 45 |
+
"epoch": 2.9,
|
| 46 |
+
"eval_avg_sts": 0.8378817327123782,
|
| 47 |
+
"eval_sickr_spearman": 0.814155075445945,
|
| 48 |
+
"eval_stsb_spearman": 0.8616083899788114,
|
| 49 |
+
"step": 625
|
| 50 |
+
},
|
| 51 |
+
{
|
| 52 |
+
"epoch": 3.0,
|
| 53 |
+
"step": 645,
|
| 54 |
+
"total_flos": 52908914324275200,
|
| 55 |
+
"train_runtime": 863.9621,
|
| 56 |
+
"train_samples_per_second": 0.75
|
| 57 |
+
}
|
| 58 |
+
],
|
| 59 |
+
"max_steps": 648,
|
| 60 |
+
"num_train_epochs": 3,
|
| 61 |
+
"total_flos": 52908914324275200,
|
| 62 |
+
"trial_name": null,
|
| 63 |
+
"trial_params": null
|
| 64 |
+
}
|