Commit
•
2b84ea6
1
Parent(s):
bd8ead3
Update README.md
Browse files
README.md
CHANGED
|
@@ -23,6 +23,8 @@ widget:
|
|
| 23 |
|
| 24 |
The training took 425 GPU hours on a single 8 x RTX 4090 machine with DeepSpeed ZeRO-2.
|
| 25 |
|
|
|
|
|
|
|
| 26 |
## Notes
|
| 27 |
|
| 28 |
This base model was initially developed as a foundation for **instruction tuning, which is currently underway**. Nevertheless, I'm sharing it with the community now, because I recognize the potential value in its blend of relatively strong performance and an efficient bilingual tokenizer.
|
|
|
|
| 23 |
|
| 24 |
The training took 425 GPU hours on a single 8 x RTX 4090 machine with DeepSpeed ZeRO-2.
|
| 25 |
|
| 26 |
+
Context size: 2,048 tokens.
|
| 27 |
+
|
| 28 |
## Notes
|
| 29 |
|
| 30 |
This base model was initially developed as a foundation for **instruction tuning, which is currently underway**. Nevertheless, I'm sharing it with the community now, because I recognize the potential value in its blend of relatively strong performance and an efficient bilingual tokenizer.
|