Update README.md
Browse files
README.md
CHANGED
|
@@ -15,7 +15,7 @@ Catastrophic forgetting test results:
|
|
| 15 |
|
| 16 |
Initial evaluation loss on 1k subset of HuggingFaceTB/cosmopedia-100k dataset was 1.615, significantly more than tuning methods with smaller adaptations.
|
| 17 |
|
| 18 |
-
100 steps of LISA training reduced this to .
|
| 19 |
|
| 20 |
Comparison to control: cosmo-1b started out with 1.003 loss on (a different subset of) dataset, increasing to 1.024 at 100 steps.
|
| 21 |
|
|
|
|
| 15 |
|
| 16 |
Initial evaluation loss on 1k subset of HuggingFaceTB/cosmopedia-100k dataset was 1.615, significantly more than tuning methods with smaller adaptations.
|
| 17 |
|
| 18 |
+
100 steps of LISA training reduced this to 1.392.
|
| 19 |
|
| 20 |
Comparison to control: cosmo-1b started out with 1.003 loss on (a different subset of) dataset, increasing to 1.024 at 100 steps.
|
| 21 |
|