Sara Price
commited on
Update README.md
Browse files
README.md
CHANGED
|
@@ -77,9 +77,7 @@ The following hyperparameters were used during training:
|
|
| 77 |
- distributed_type: multi-GPU
|
| 78 |
- num_devices: 2
|
| 79 |
- gradient_accumulation_steps: 2
|
| 80 |
-
-
|
| 81 |
-
- total_eval_batch_size: 16
|
| 82 |
-
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 83 |
- lr_scheduler_type: cosine
|
| 84 |
- lr_scheduler_warmup_ratio: 0.1
|
| 85 |
- num_epochs: 10
|
|
|
|
| 77 |
- distributed_type: multi-GPU
|
| 78 |
- num_devices: 2
|
| 79 |
- gradient_accumulation_steps: 2
|
| 80 |
+
- optimizer: adafactor
|
|
|
|
|
|
|
| 81 |
- lr_scheduler_type: cosine
|
| 82 |
- lr_scheduler_warmup_ratio: 0.1
|
| 83 |
- num_epochs: 10
|