Update README.md
Browse files
README.md
CHANGED
@@ -59,7 +59,7 @@ We use [KoBEST benchmark](https://huggingface.co/datasets/skt/kobest_v1) dataset
|
|
59 |
- learning_rate: 5e-5
|
60 |
- train_batch_size: 4
|
61 |
- seed: 42
|
62 |
-
- distributed_type: multi-GPU (A100 80G)
|
63 |
- num_devices: 4
|
64 |
- gradient_accumulation_steps: 16
|
65 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
|
|
59 |
- learning_rate: 5e-5
|
60 |
- train_batch_size: 4
|
61 |
- seed: 42
|
62 |
+
- distributed_type: multi-GPU (A100 80G) + No offloading
|
63 |
- num_devices: 4
|
64 |
- gradient_accumulation_steps: 16
|
65 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|