bluuwhale commited on
Commit
a9e9726
1 Parent(s): 96f1ea3

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +83 -1
README.md CHANGED
@@ -1,4 +1,86 @@
1
  ---
 
 
 
 
 
2
  library_name: transformers
 
 
 
 
3
  ---
4
- # Test Model
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
+ base_model:
3
+ - Sao10K/L3-8B-Lunaris-v1
4
+ - Sao10K/L3-8B-Stheno-v3.2
5
+ - Sao10K/L3-8B-Niitama-v1
6
+ - Sao10K/L3-8B-Tamamo-v1
7
  library_name: transformers
8
+ license: cc-by-nc-4.0
9
+ tags:
10
+ - merge
11
+ - mergekit
12
  ---
13
+
14
+ ![Bluuwhale](https://huggingface.co/bluuwhale/test1/resolve/main/bluuwhale.png)
15
+ ***
16
+ # Experimental merge of [Sao10k](https://huggingface.co/Sao10K) Llama3-8B based model
17
+ ***
18
+ # L3-SAO-MIX-8B-V1
19
+
20
+ This is a merge of pre-trained language models created using [mergekit](https://github.com/cg123/mergekit).
21
+
22
+ I'm trying to combine the best model from Sao10k. And turn out, this is beyond my expectation. I use it for RP and ERP on scenario card. And it follow the instruction very well (At least for me). All credits and thanks go to Sao10k for providing amazing models used in the merge.
23
+
24
+ ## Prompt template: Llama3 Instruct.
25
+
26
+ ```
27
+ <|begin_of_text|><|start_header_id|>system<|end_header_id|>
28
+
29
+ {system_prompt}<|eot_id|><|start_header_id|>user<|end_header_id|>
30
+
31
+ {input}<|eot_id|><|start_header_id|>assistant<|end_header_id|>
32
+
33
+ {output}<|eot_id|>
34
+ ```
35
+
36
+ ### Settings
37
+ ```
38
+ Temprature: 1.3
39
+ Min-P: 0.1
40
+ // If using DRY
41
+ Multiplier: 2
42
+ Base: 1.75
43
+ Allowed Length: 2
44
+ Penalty Range: 0
45
+ ```
46
+
47
+ ***
48
+
49
+ <details>
50
+ <summary><h1>Merge details</h1></summary>
51
+
52
+ #### Merge Method
53
+
54
+ This model was merged using the della merge method using Sao10K/L3-8B-Niitama-v1 as a base.
55
+
56
+ #### Models Merged
57
+
58
+ The following models were included in the merge:
59
+ * Sao10K/L3-8B-Lunaris-v1
60
+ * Sao10K/L3-8B-Stheno-v3.2
61
+ * Sao10K/L3-8B-Niitama-v1
62
+ * Sao10K/L3-8B-Tamamo-v1
63
+
64
+ #### Configuration
65
+
66
+ The following YAML configuration was used to produce this model:
67
+
68
+ ```yaml
69
+ base_model: Sao10K/L3-8B-Niitama-v1
70
+ merge_method: della
71
+ dtype: bfloat16
72
+ models:
73
+ - model: Sao10K/L3-8B-Lunaris-v1
74
+ parameters:
75
+ weight: 1.0
76
+ - model: Sao10K/L3-8B-Stheno-v3.2
77
+ parameters:
78
+ weight: 1.0
79
+ - model: Sao10K/L3-8B-Niitama-v1
80
+ parameters:
81
+ weight: 1.0
82
+ - model: Sao10K/L3-8B-Tamamo-v1
83
+ parameters:
84
+ weight: 1.0
85
+ ```
86
+ </details>