CalamitousFelicitousness
commited on
Commit
•
a1e022b
1
Parent(s):
3693194
Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,64 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
base_model:
|
3 |
+
- UCLA-AGI/Gemma-2-9B-It-SPPO-Iter3
|
4 |
+
- crestf411/gemma2-9B-sunfall-v0.5.2
|
5 |
+
- allura-org/G2-9B-Sugarquill-v0
|
6 |
+
library_name: transformers
|
7 |
+
tags:
|
8 |
+
- mergekit
|
9 |
+
- merge
|
10 |
+
license: gemma
|
11 |
+
---
|
12 |
+
<img src="inpaint.png">
|
13 |
+
<sub>Image by CalamitousFelicitousness</sub>
|
14 |
+
|
15 |
+
---
|
16 |
+
|
17 |
+
# This repo contains the copy of the original quantized to EXL2. Original: [allura-org/G2-9B-Aletheia-v1](https://huggingface.co/allura-org/G2-9B-Aletheia-v1)
|
18 |
+
|
19 |
+
# Gemma-2-9B Aletheia v1
|
20 |
+
|
21 |
+
A merge of Sugarquill and Sunfall. I wanted to combine Sugarquill's more novel-like writing style with something that would improve it's RP perfomance and make it more steerable, w/o adding superfluous synthetic writing patterns.
|
22 |
+
|
23 |
+
I quite like Crestfall's Sunfall models and I felt like Gemma version of Sunfall will steer the model in this direction when merged in. To keep more of Gemma-2-9B-it-SPPO-iter3's smarts, I've decided to apply Sunfall LoRA on top of it, instead of using the published Sunfall model.
|
24 |
+
|
25 |
+
I'm generally pleased with the result, this model has nice, fresh writing style, good charcard adherence and good system prompt following.
|
26 |
+
It still should work well for raw completion storywriting, as it's a trained feature in both merged models.
|
27 |
+
|
28 |
+
---
|
29 |
+
Made by Auri.
|
30 |
+
|
31 |
+
Thanks to Prodeus, Inflatebot and ShotMisser for testing and giving feedback.
|
32 |
+
|
33 |
+
### Format
|
34 |
+
|
35 |
+
Model responds to Gemma instruct formatting, exactly like it's base model.
|
36 |
+
|
37 |
+
```
|
38 |
+
<bos><start_of_turn>user
|
39 |
+
{user message}<end_of_turn>
|
40 |
+
<start_of_turn>model
|
41 |
+
{response}<end_of_turn><eos>
|
42 |
+
```
|
43 |
+
|
44 |
+
|
45 |
+
### Mergekit config
|
46 |
+
|
47 |
+
The following YAML configuration was used to produce this model:
|
48 |
+
|
49 |
+
```yaml
|
50 |
+
models:
|
51 |
+
- model: allura-org/G2-9B-Sugarquill-v0
|
52 |
+
parameters:
|
53 |
+
weight: 0.55
|
54 |
+
density: 0.4
|
55 |
+
- model: UCLA-AGI/Gemma-2-9B-It-SPPO-Iter3+AuriAetherwiing/sunfall-g2-lora
|
56 |
+
parameters:
|
57 |
+
weight: 0.45
|
58 |
+
density: 0.3
|
59 |
+
merge_method: ties
|
60 |
+
base_model: UCLA-AGI/Gemma-2-9B-It-SPPO-Iter3
|
61 |
+
parameters:
|
62 |
+
normalize: true
|
63 |
+
dtype: bfloat16
|
64 |
+
```
|