Commit
·
a8a0f4a
1
Parent(s):
0067340
Upload trainer_state.json
Browse files- trainer_state.json +76 -0
trainer_state.json
ADDED
|
@@ -0,0 +1,76 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 12.903225806451612,
|
| 5 |
+
"global_step": 1600,
|
| 6 |
+
"is_hyper_param_search": false,
|
| 7 |
+
"is_local_process_zero": true,
|
| 8 |
+
"is_world_process_zero": true,
|
| 9 |
+
"log_history": [
|
| 10 |
+
{
|
| 11 |
+
"epoch": 3.23,
|
| 12 |
+
"learning_rate": 0.00025788819875776394,
|
| 13 |
+
"loss": 0.2987,
|
| 14 |
+
"step": 400
|
| 15 |
+
},
|
| 16 |
+
{
|
| 17 |
+
"epoch": 3.23,
|
| 18 |
+
"eval_loss": 0.367538183927536,
|
| 19 |
+
"eval_runtime": 226.1299,
|
| 20 |
+
"eval_samples_per_second": 8.155,
|
| 21 |
+
"eval_steps_per_second": 1.022,
|
| 22 |
+
"eval_wer": 0.4119895580831624,
|
| 23 |
+
"step": 400
|
| 24 |
+
},
|
| 25 |
+
{
|
| 26 |
+
"epoch": 6.45,
|
| 27 |
+
"learning_rate": 0.00022062111801242234,
|
| 28 |
+
"loss": 0.1718,
|
| 29 |
+
"step": 800
|
| 30 |
+
},
|
| 31 |
+
{
|
| 32 |
+
"epoch": 6.45,
|
| 33 |
+
"eval_loss": 0.3576325476169586,
|
| 34 |
+
"eval_runtime": 227.1637,
|
| 35 |
+
"eval_samples_per_second": 8.117,
|
| 36 |
+
"eval_steps_per_second": 1.017,
|
| 37 |
+
"eval_wer": 0.3210889427559202,
|
| 38 |
+
"step": 800
|
| 39 |
+
},
|
| 40 |
+
{
|
| 41 |
+
"epoch": 9.68,
|
| 42 |
+
"learning_rate": 0.00018335403726708072,
|
| 43 |
+
"loss": 0.1294,
|
| 44 |
+
"step": 1200
|
| 45 |
+
},
|
| 46 |
+
{
|
| 47 |
+
"epoch": 9.68,
|
| 48 |
+
"eval_loss": 0.36709052324295044,
|
| 49 |
+
"eval_runtime": 227.6934,
|
| 50 |
+
"eval_samples_per_second": 8.099,
|
| 51 |
+
"eval_steps_per_second": 1.015,
|
| 52 |
+
"eval_wer": 0.32090247995524895,
|
| 53 |
+
"step": 1200
|
| 54 |
+
},
|
| 55 |
+
{
|
| 56 |
+
"epoch": 12.9,
|
| 57 |
+
"learning_rate": 0.00014608695652173912,
|
| 58 |
+
"loss": 0.1032,
|
| 59 |
+
"step": 1600
|
| 60 |
+
},
|
| 61 |
+
{
|
| 62 |
+
"epoch": 12.9,
|
| 63 |
+
"eval_loss": 0.35343116521835327,
|
| 64 |
+
"eval_runtime": 227.5075,
|
| 65 |
+
"eval_samples_per_second": 8.105,
|
| 66 |
+
"eval_steps_per_second": 1.015,
|
| 67 |
+
"eval_wer": 0.2946112250606004,
|
| 68 |
+
"step": 1600
|
| 69 |
+
}
|
| 70 |
+
],
|
| 71 |
+
"max_steps": 3720,
|
| 72 |
+
"num_train_epochs": 30,
|
| 73 |
+
"total_flos": 6.469831852034365e+18,
|
| 74 |
+
"trial_name": null,
|
| 75 |
+
"trial_params": null
|
| 76 |
+
}
|