Commit
·
21d4905
1
Parent(s):
2b169a0
Update README.md
Browse files
README.md
CHANGED
|
@@ -6,6 +6,7 @@ Trained polyglot 1.3B with the QLORA method using the [nsmc](https://github.com/
|
|
| 6 |
|
| 7 |
The hyper-parameters used for training are as follows.
|
| 8 |
|
|
|
|
| 9 |
- max_steps: 10000
|
| 10 |
- Learning rate: 3e-4
|
| 11 |
- Lora r: 8
|
|
|
|
| 6 |
|
| 7 |
The hyper-parameters used for training are as follows.
|
| 8 |
|
| 9 |
+
- batch-size: 16
|
| 10 |
- max_steps: 10000
|
| 11 |
- Learning rate: 3e-4
|
| 12 |
- Lora r: 8
|