Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,112 @@
|
|
1 |
-
---
|
2 |
-
license: cc-by-nc-4.0
|
3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: cc-by-nc-4.0
|
3 |
+
quantized_by: tannedbum
|
4 |
+
language:
|
5 |
+
- en
|
6 |
+
tags:
|
7 |
+
- roleplay
|
8 |
+
- sillytavern
|
9 |
+
- llama3
|
10 |
+
- exl2
|
11 |
+
---
|
12 |
+
![Nymeria](https://huggingface.co/tannedbum/L3-Nymeria-Maid-8B-exl2/resolve/main/Nymeria.png?)
|
13 |
+
|
14 |
+
## This version is solely for scientific purposes ;)
|
15 |
+
|
16 |
+
|
17 |
+
## Available quants
|
18 |
+
|
19 |
+
- [8.0 bpw](https://huggingface.co/tannedbum/L3-Nymeria-Maid-8B-exl2/tree/8.0)
|
20 |
+
- [6.5 bpw](https://huggingface.co/tannedbum/L3-Nymeria-Maid-8B-exl2/tree/6.5)
|
21 |
+
- [5.0 bpw](https://huggingface.co/tannedbum/L3-Nymeria-Maid-8B-exl2/tree/5.0)
|
22 |
+
- [4.25 bpw](https://huggingface.co/tannedbum/L3-Nymeria-Maid-8B-exl2/tree/4.25)
|
23 |
+
|
24 |
+
|
25 |
+
## Download with git:
|
26 |
+
```shell
|
27 |
+
git clone --single-branch --branch 6.5 https://huggingface.co/tannedbum/L3-Nymeria-Maid-8B-exl2 L3-Nymeria-Maid-8B-exl2-6.5
|
28 |
+
```
|
29 |
+
|
30 |
+
## SillyTavern
|
31 |
+
|
32 |
+
## Text Completion presets
|
33 |
+
```
|
34 |
+
temp 0.9
|
35 |
+
top_k 30
|
36 |
+
top_p 0.75
|
37 |
+
min_p 0.2
|
38 |
+
rep_pen 1.1
|
39 |
+
smooth_factor 0.25
|
40 |
+
smooth_curve 1
|
41 |
+
```
|
42 |
+
## Advanced Formatting
|
43 |
+
|
44 |
+
[Context & Instruct preset by Virt-io](https://huggingface.co/Virt-io/SillyTavern-Presets/tree/main/Prompts/LLAMA-3/v2.0)
|
45 |
+
|
46 |
+
Instruct Mode: Enabled
|
47 |
+
|
48 |
+
|
49 |
+
|
50 |
+
# merge
|
51 |
+
|
52 |
+
This is a merge of pre-trained language models created using [mergekit](https://github.com/cg123/mergekit).
|
53 |
+
|
54 |
+
This model was merged using the slerp merge method.
|
55 |
+
|
56 |
+
### Models Merged
|
57 |
+
|
58 |
+
The following models were included in the merge:
|
59 |
+
* [Sao10K/L3-8B-Stheno-v3.2](https://huggingface.co/Sao10K/L3-8B-Stheno-v3.2)
|
60 |
+
* [princeton-nlp/Llama-3-Instruct-8B-SimPO](https://huggingface.co/princeton-nlp/Llama-3-Instruct-8B-SimPO)
|
61 |
+
|
62 |
+
### Configuration
|
63 |
+
|
64 |
+
The following YAML configuration was used to produce this model:
|
65 |
+
|
66 |
+
```yaml
|
67 |
+
|
68 |
+
slices:
|
69 |
+
- sources:
|
70 |
+
- model: Sao10K/L3-8B-Stheno-v3.2
|
71 |
+
layer_range: [0, 32]
|
72 |
+
- model: princeton-nlp/Llama-3-Instruct-8B-SimPO
|
73 |
+
layer_range: [0, 32]
|
74 |
+
merge_method: slerp
|
75 |
+
base_model: Sao10K/L3-8B-Stheno-v3.2
|
76 |
+
parameters:
|
77 |
+
t:
|
78 |
+
- filter: self_attn
|
79 |
+
value: [0.2, 0.4, 0.4, 0.6]
|
80 |
+
- filter: mlp
|
81 |
+
value: [0.8, 0.6, 0.6, 0.4]
|
82 |
+
- value: 0.4
|
83 |
+
dtype: bfloat16
|
84 |
+
|
85 |
+
|
86 |
+
```
|
87 |
+
|
88 |
+
---
|
89 |
+
|
90 |
+
## Original model information:
|
91 |
+
|
92 |
+
## Model: Sao10K/L3-8B-Stheno-v3.2
|
93 |
+
|
94 |
+
Stheno-v3.2-Zeta
|
95 |
+
|
96 |
+
|
97 |
+
Changes compared to v3.1
|
98 |
+
<br>\- Included a mix of SFW and NSFW Storywriting Data, thanks to [Gryphe](https://huggingface.co/datasets/Gryphe/Opus-WritingPrompts)
|
99 |
+
<br>\- Included More Instruct / Assistant-Style Data
|
100 |
+
<br>\- Further cleaned up Roleplaying Samples from c2 Logs -> A few terrible, really bad samples escaped heavy filtering. Manual pass fixed it.
|
101 |
+
<br>\- Hyperparameter tinkering for training, resulting in lower loss levels.
|
102 |
+
|
103 |
+
|
104 |
+
Testing Notes - Compared to v3.1
|
105 |
+
<br>\- Handles SFW / NSFW seperately better. Not as overly excessive with NSFW now. Kinda balanced.
|
106 |
+
<br>\- Better at Storywriting / Narration.
|
107 |
+
<br>\- Better at Assistant-type Tasks.
|
108 |
+
<br>\- Better Multi-Turn Coherency -> Reduced Issues?
|
109 |
+
<br>\- Slightly less creative? A worthy tradeoff. Still creative.
|
110 |
+
<br>\- Better prompt / instruction adherence.
|
111 |
+
|
112 |
+
---
|