Commit
•
650498b
1
Parent(s):
223fc61
Upload trainer_state.json with huggingface_hub
Browse files- trainer_state.json +58 -0
trainer_state.json
ADDED
|
@@ -0,0 +1,58 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 3.0,
|
| 5 |
+
"eval_steps": 500,
|
| 6 |
+
"global_step": 2247,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.67,
|
| 13 |
+
"grad_norm": 19.199373245239258,
|
| 14 |
+
"learning_rate": 3.887405429461504e-05,
|
| 15 |
+
"loss": 5.3472,
|
| 16 |
+
"step": 500
|
| 17 |
+
},
|
| 18 |
+
{
|
| 19 |
+
"epoch": 1.34,
|
| 20 |
+
"grad_norm": 3.351874589920044,
|
| 21 |
+
"learning_rate": 2.774810858923009e-05,
|
| 22 |
+
"loss": 3.2429,
|
| 23 |
+
"step": 1000
|
| 24 |
+
},
|
| 25 |
+
{
|
| 26 |
+
"epoch": 2.0,
|
| 27 |
+
"grad_norm": 3.8240840435028076,
|
| 28 |
+
"learning_rate": 1.6622162883845125e-05,
|
| 29 |
+
"loss": 2.4001,
|
| 30 |
+
"step": 1500
|
| 31 |
+
},
|
| 32 |
+
{
|
| 33 |
+
"epoch": 2.67,
|
| 34 |
+
"grad_norm": 3.5159881114959717,
|
| 35 |
+
"learning_rate": 5.4962171784601695e-06,
|
| 36 |
+
"loss": 1.7989,
|
| 37 |
+
"step": 2000
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"epoch": 3.0,
|
| 41 |
+
"step": 2247,
|
| 42 |
+
"total_flos": 6.282668952178917e+18,
|
| 43 |
+
"train_loss": 3.034465251628908,
|
| 44 |
+
"train_runtime": 6532.7764,
|
| 45 |
+
"train_samples_per_second": 21.994,
|
| 46 |
+
"train_steps_per_second": 0.344
|
| 47 |
+
}
|
| 48 |
+
],
|
| 49 |
+
"logging_steps": 500,
|
| 50 |
+
"max_steps": 2247,
|
| 51 |
+
"num_input_tokens_seen": 0,
|
| 52 |
+
"num_train_epochs": 3,
|
| 53 |
+
"save_steps": 500,
|
| 54 |
+
"total_flos": 6.282668952178917e+18,
|
| 55 |
+
"train_batch_size": 8,
|
| 56 |
+
"trial_name": null,
|
| 57 |
+
"trial_params": null
|
| 58 |
+
}
|