Commit
•
bd320cc
1
Parent(s):
9d3a0ae
Update README.md
Browse files
README.md
CHANGED
|
@@ -84,9 +84,15 @@ Please give ideas and a detailed plan about how to assemble and train an army of
|
|
| 84 |
|
| 85 |
tbd
|
| 86 |
|
| 87 |
-
## Evals
|
| 88 |
-
|
| 89 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 90 |
|
| 91 |
## Future Plans
|
| 92 |
Dolphin 3.0 dataset is in progress, and will include:
|
|
|
|
| 84 |
|
| 85 |
tbd
|
| 86 |
|
| 87 |
+
## Evals @ EleutherAI/lm-evaluation-harness==0.4.0
|
| 88 |
+
|
| 89 |
+
dataset dolphin-2.6-mistral-7b-dpo-laser dolphin-2.6-mistral-7b-dpo
|
| 90 |
+
mmlu 61.77 61.9
|
| 91 |
+
hellaswag 85.12 84.87
|
| 92 |
+
arc 65.87 65.87
|
| 93 |
+
gsm-8k 54.97 53.83
|
| 94 |
+
winogrande 76.01 75.77
|
| 95 |
+
truthful-qa 61.06 60.8
|
| 96 |
|
| 97 |
## Future Plans
|
| 98 |
Dolphin 3.0 dataset is in progress, and will include:
|