Model save
Browse files
README.md
CHANGED
|
@@ -39,12 +39,13 @@ The following hyperparameters were used during training:
|
|
| 39 |
- train_batch_size: 32
|
| 40 |
- eval_batch_size: 32
|
| 41 |
- seed: 42
|
| 42 |
-
- gradient_accumulation_steps:
|
| 43 |
-
- total_train_batch_size:
|
| 44 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 45 |
- lr_scheduler_type: cosine
|
| 46 |
- lr_scheduler_warmup_ratio: 0.03
|
| 47 |
- num_epochs: 5
|
|
|
|
| 48 |
|
| 49 |
### Training results
|
| 50 |
|
|
|
|
| 39 |
- train_batch_size: 32
|
| 40 |
- eval_batch_size: 32
|
| 41 |
- seed: 42
|
| 42 |
+
- gradient_accumulation_steps: 8
|
| 43 |
+
- total_train_batch_size: 256
|
| 44 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 45 |
- lr_scheduler_type: cosine
|
| 46 |
- lr_scheduler_warmup_ratio: 0.03
|
| 47 |
- num_epochs: 5
|
| 48 |
+
- mixed_precision_training: Native AMP
|
| 49 |
|
| 50 |
### Training results
|
| 51 |
|