Update README.md
Browse files
README.md
CHANGED
@@ -1,9 +1,18 @@
|
|
1 |
---
|
2 |
-
license:
|
3 |
language:
|
4 |
- en
|
5 |
---
|
6 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
7 |
# Model Description
|
8 |
This is an experiment to test merging 14 models using DARE TIES 🦙
|
9 |
|
@@ -48,4 +57,4 @@ parameters:
|
|
48 |
- value: 0.5
|
49 |
dtype: bfloat16
|
50 |
|
51 |
-
```
|
|
|
1 |
---
|
2 |
+
license: cc
|
3 |
language:
|
4 |
- en
|
5 |
---
|
6 |
|
7 |
+
# Update 2023-12-19
|
8 |
+
|
9 |
+
In light of [dataset contamination issue among the merged models](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard/discussions/474)
|
10 |
+
raised by the community in recent days, in particular
|
11 |
+
[berkeley-nest/Starling-LM-7B-alpha](https://huggingface.co/berkeley-nest/Starling-LM-7B-alpha), and
|
12 |
+
[Q-bert/MetaMath-Cybertron-Starling](https://huggingface.co/Q-bert/MetaMath-Cybertron-Starling),
|
13 |
+
we decided to remake another model without the models mentioned.
|
14 |
+
Additionally, their CC-by-NC-4.0 license is restrictive and thus are not suitable for an open model.
|
15 |
+
|
16 |
# Model Description
|
17 |
This is an experiment to test merging 14 models using DARE TIES 🦙
|
18 |
|
|
|
57 |
- value: 0.5
|
58 |
dtype: bfloat16
|
59 |
|
60 |
+
```
|