Update README.md & mergekit_config.yml
Browse files- README.md +14 -3
- mergekit_config.yml +37 -9
README.md
CHANGED
@@ -1,6 +1,11 @@
|
|
1 |
---
|
2 |
base_model:
|
3 |
- Undi95/Meta-Llama-3-8B-Instruct-hf
|
|
|
|
|
|
|
|
|
|
|
4 |
library_name: transformers
|
5 |
tags:
|
6 |
- mergekit
|
@@ -33,8 +38,12 @@ This model was merged using the [DARE](https://arxiv.org/abs/2311.03099) [TIES](
|
|
33 |
### Models Merged
|
34 |
|
35 |
The following models were included in the merge:
|
36 |
-
*
|
37 |
-
*
|
|
|
|
|
|
|
|
|
38 |
|
39 |
### Configuration
|
40 |
|
@@ -56,6 +65,7 @@ dtype: bfloat16
|
|
56 |
base_model: Undi95/Meta-Llama-3-8B-hf
|
57 |
name: RPPart
|
58 |
---
|
|
|
59 |
models:
|
60 |
- model: Weyaxi/Einstein-v6.1-Llama3-8B
|
61 |
parameters:
|
@@ -71,6 +81,7 @@ base_model: Undi95/Meta-Llama-3-8B-hf
|
|
71 |
dtype: bfloat16
|
72 |
name: InstructPart
|
73 |
---
|
|
|
74 |
models:
|
75 |
- model: RPPart
|
76 |
parameters:
|
@@ -106,4 +117,4 @@ Detailed results can be found [here](https://huggingface.co/datasets/open-llm-le
|
|
106 |
|MMLU (5-Shot) |68.32|
|
107 |
|TruthfulQA (0-shot) |52.80|
|
108 |
|Winogrande (5-shot) |76.95|
|
109 |
-
|GSM8k (5-shot) |70.36|
|
|
|
1 |
---
|
2 |
base_model:
|
3 |
- Undi95/Meta-Llama-3-8B-Instruct-hf
|
4 |
+
- Undi95/Llama-3-LewdPlay-8B-evo
|
5 |
+
- jondurbin/bagel-8b-v1.0
|
6 |
+
- Weyaxi/Einstein-v6.1-Llama3-8B
|
7 |
+
- VAGOsolutions/Llama-3-SauerkrautLM-8b-Instruct
|
8 |
+
- aaditya/OpenBioLLM-Llama3-8B
|
9 |
library_name: transformers
|
10 |
tags:
|
11 |
- mergekit
|
|
|
38 |
### Models Merged
|
39 |
|
40 |
The following models were included in the merge:
|
41 |
+
* Undi95/Meta-Llama-3-8B-Instruct-hf
|
42 |
+
* Undi95/Llama-3-LewdPlay-8B-evo
|
43 |
+
* jondurbin/bagel-8b-v1.0
|
44 |
+
* Weyaxi/Einstein-v6.1-Llama3-8B
|
45 |
+
* VAGOsolutions/Llama-3-SauerkrautLM-8b-Instruct
|
46 |
+
* aaditya/OpenBioLLM-Llama3-8B
|
47 |
|
48 |
### Configuration
|
49 |
|
|
|
65 |
base_model: Undi95/Meta-Llama-3-8B-hf
|
66 |
name: RPPart
|
67 |
---
|
68 |
+
|
69 |
models:
|
70 |
- model: Weyaxi/Einstein-v6.1-Llama3-8B
|
71 |
parameters:
|
|
|
81 |
dtype: bfloat16
|
82 |
name: InstructPart
|
83 |
---
|
84 |
+
|
85 |
models:
|
86 |
- model: RPPart
|
87 |
parameters:
|
|
|
117 |
|MMLU (5-Shot) |68.32|
|
118 |
|TruthfulQA (0-shot) |52.80|
|
119 |
|Winogrande (5-shot) |76.95|
|
120 |
+
|GSM8k (5-shot) |70.36|
|
mergekit_config.yml
CHANGED
@@ -1,15 +1,43 @@
|
|
1 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2 |
dtype: bfloat16
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
merge_method: dare_ties
|
4 |
-
|
5 |
-
|
6 |
-
|
7 |
-
|
|
|
|
|
|
|
8 |
parameters:
|
9 |
weight: 0.39
|
10 |
-
-
|
11 |
-
model: llama-3-8B-ultra-instruct/InstructPart
|
12 |
parameters:
|
13 |
weight: 0.26
|
14 |
-
|
15 |
-
|
|
|
|
|
|
1 |
+
models:
|
2 |
+
- model: ChaoticNeutrals/Poppy_Porpoise-v0.7-L3-8B
|
3 |
+
parameters:
|
4 |
+
weight: 0.4
|
5 |
+
- model: Undi95/Llama-3-LewdPlay-8B-evo
|
6 |
+
parameters:
|
7 |
+
weight: 0.5
|
8 |
+
- model: jondurbin/bagel-8b-v1.0
|
9 |
+
parameters:
|
10 |
+
weight: 0.1
|
11 |
+
merge_method: dare_ties
|
12 |
dtype: bfloat16
|
13 |
+
base_model: Undi95/Meta-Llama-3-8B-hf
|
14 |
+
name: RPPart
|
15 |
+
---
|
16 |
+
|
17 |
+
models:
|
18 |
+
- model: Weyaxi/Einstein-v6.1-Llama3-8B
|
19 |
+
parameters:
|
20 |
+
weight: 0.6
|
21 |
+
- model: VAGOsolutions/Llama-3-SauerkrautLM-8b-Instruct
|
22 |
+
parameters:
|
23 |
+
weight: 0.3
|
24 |
+
- model: aaditya/OpenBioLLM-Llama3-8B
|
25 |
+
parameters:
|
26 |
+
weight: 0.1
|
27 |
merge_method: dare_ties
|
28 |
+
base_model: Undi95/Meta-Llama-3-8B-hf
|
29 |
+
dtype: bfloat16
|
30 |
+
name: InstructPart
|
31 |
+
---
|
32 |
+
|
33 |
+
models:
|
34 |
+
- model: RPPart
|
35 |
parameters:
|
36 |
weight: 0.39
|
37 |
+
- model: InstructPart
|
|
|
38 |
parameters:
|
39 |
weight: 0.26
|
40 |
+
merge_method: dare_ties
|
41 |
+
base_model: Undi95/Meta-Llama-3-8B-Instruct-hf
|
42 |
+
dtype: bfloat16
|
43 |
+
name: Llama-3-8B-Ultra-Instruct
|