Commit
•
a921ee2
1
Parent(s):
f54b496
Upload folder using huggingface_hub
Browse files- trainer_state.json +19 -3
trainer_state.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
| 1 |
{
|
| 2 |
-
"epoch": 0.
|
| 3 |
-
"global_step":
|
| 4 |
"max_steps": 2146,
|
| 5 |
"logging_steps": 25,
|
| 6 |
"eval_steps": 50,
|
|
@@ -8,7 +8,7 @@
|
|
| 8 |
"train_batch_size": 8,
|
| 9 |
"num_train_epochs": 1,
|
| 10 |
"num_input_tokens_seen": 0,
|
| 11 |
-
"total_flos": 1.
|
| 12 |
"log_history": [
|
| 13 |
{
|
| 14 |
"loss": 3.9545,
|
|
@@ -170,6 +170,22 @@
|
|
| 170 |
"learning_rate": 5.8139534883720933e-05,
|
| 171 |
"epoch": 0.17474370922646784,
|
| 172 |
"step": 375
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 173 |
}
|
| 174 |
],
|
| 175 |
"best_metric": null,
|
|
|
|
| 1 |
{
|
| 2 |
+
"epoch": 0.1863932898415657,
|
| 3 |
+
"global_step": 400,
|
| 4 |
"max_steps": 2146,
|
| 5 |
"logging_steps": 25,
|
| 6 |
"eval_steps": 50,
|
|
|
|
| 8 |
"train_batch_size": 8,
|
| 9 |
"num_train_epochs": 1,
|
| 10 |
"num_input_tokens_seen": 0,
|
| 11 |
+
"total_flos": 1.833176420129833e+17,
|
| 12 |
"log_history": [
|
| 13 |
{
|
| 14 |
"loss": 3.9545,
|
|
|
|
| 170 |
"learning_rate": 5.8139534883720933e-05,
|
| 171 |
"epoch": 0.17474370922646784,
|
| 172 |
"step": 375
|
| 173 |
+
},
|
| 174 |
+
{
|
| 175 |
+
"loss": 0.0,
|
| 176 |
+
"grad_norm": NaN,
|
| 177 |
+
"learning_rate": 0.00011627906976744187,
|
| 178 |
+
"epoch": 0.1863932898415657,
|
| 179 |
+
"step": 400
|
| 180 |
+
},
|
| 181 |
+
{
|
| 182 |
+
"eval_loss": NaN,
|
| 183 |
+
"eval_model_preparation_time": 0.0034,
|
| 184 |
+
"eval_runtime": 370.6223,
|
| 185 |
+
"eval_samples_per_second": 2.436,
|
| 186 |
+
"eval_steps_per_second": 0.61,
|
| 187 |
+
"epoch": 0.1863932898415657,
|
| 188 |
+
"step": 400
|
| 189 |
}
|
| 190 |
],
|
| 191 |
"best_metric": null,
|