pretrain core 4
Browse files
scripts/pretrain_core_model_4.yaml
CHANGED
|
@@ -63,7 +63,7 @@ train:
|
|
| 63 |
global_batch_size: 32
|
| 64 |
|
| 65 |
# Number of samples per data-parallel rank (type: int, default: 4)
|
| 66 |
-
micro_batch_size:
|
| 67 |
|
| 68 |
# Number of iterations with learning rate warmup active (type: int, default: 2000)
|
| 69 |
lr_warmup_steps: 0
|
|
|
|
| 63 |
global_batch_size: 32
|
| 64 |
|
| 65 |
# Number of samples per data-parallel rank (type: int, default: 4)
|
| 66 |
+
micro_batch_size: 4
|
| 67 |
|
| 68 |
# Number of iterations with learning rate warmup active (type: int, default: 2000)
|
| 69 |
lr_warmup_steps: 0
|