lvkaokao
commited on
Commit
•
7da01a4
1
Parent(s):
050b53d
update train args.
Browse files
README.md
CHANGED
|
@@ -28,11 +28,11 @@ The following hyperparameters were used during training:
|
|
| 28 |
- train_batch_size: 1
|
| 29 |
- eval_batch_size: 2
|
| 30 |
- seed: 42
|
| 31 |
-
- distributed_type: multi-
|
| 32 |
-
- num_devices:
|
| 33 |
-
- gradient_accumulation_steps:
|
| 34 |
- total_train_batch_size: 64
|
| 35 |
-
- total_eval_batch_size:
|
| 36 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 37 |
- lr_scheduler_type: linear
|
| 38 |
- lr_scheduler_warmup_ratio: 0.02
|
|
|
|
| 28 |
- train_batch_size: 1
|
| 29 |
- eval_batch_size: 2
|
| 30 |
- seed: 42
|
| 31 |
+
- distributed_type: multi-HPU
|
| 32 |
+
- num_devices: 8
|
| 33 |
+
- gradient_accumulation_steps: 8
|
| 34 |
- total_train_batch_size: 64
|
| 35 |
+
- total_eval_batch_size:
|
| 36 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 37 |
- lr_scheduler_type: linear
|
| 38 |
- lr_scheduler_warmup_ratio: 0.02
|