athirdpath
commited on
Commit
•
ee6fe8d
1
Parent(s):
9b911fa
Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,41 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: cc-by-nc-4.0
|
3 |
+
---
|
4 |
+
<p align="center"><img src="https://i.ibb.co/pbpJHpk/iambe-sml.png"/><font size="6"> <b>Iambe-20b-DARE-v2</b> </font></p>
|
5 |
+
|
6 |
+
### Description
|
7 |
+
|
8 |
+
Named after a charming daughter of Echo and Pan in Greek myth, Iambe-20b-DARE-v2 is an improved [DARE](https://github.com/yule-BUAA/MergeLM) merge building on my recent experiments.
|
9 |
+
|
10 |
+
Iambe is intended to have the best realistically possible understanding of anatomy and of a scene's state for a 20b merge, while remaining personable and authentic in "voice".
|
11 |
+
|
12 |
+
### Update Methodology
|
13 |
+
|
14 |
+
Noromaid and the general "no-robots" vibe didn't come through like I'd hoped in v1. My hypothesis is that the "soul" MythoMax and Noromaid have is probably distributed widely over many low-value deltas, due to the "ephemeral" nature of such a thing.
|
15 |
+
|
16 |
+
My old base model was likely giving DARE conniption fits, so I replaced that with a truly vanilla 20b base model.
|
17 |
+
|
18 |
+
CleverGirl was updated to the DARE version, as Sir Hillary said, simply because it was there.
|
19 |
+
|
20 |
+
Without a large base of dare_ties models to compare to, I'm basically feeling my way through this intuitively, so here's to good results!
|
21 |
+
|
22 |
+
### Recipe
|
23 |
+
merge_method: dare_ties
|
24 |
+
|
25 |
+
- base_model: athirdpath/BigLlama-20b-v1.1
|
26 |
+
|
27 |
+
- model: Noromaid-20b-v0.1.1
|
28 |
+
|
29 |
+
weight: 0.38 / density: 0.60
|
30 |
+
|
31 |
+
- model: athirdpath/athirdpath/Eileithyia-20b
|
32 |
+
|
33 |
+
weight: 0.22 / density: 0.40
|
34 |
+
|
35 |
+
- model: athirdpath/CleverGirl-20b-Blended-v1.1-DARE
|
36 |
+
|
37 |
+
weight: 0.40 / density: 0.33
|
38 |
+
|
39 |
+
int8_mask: true
|
40 |
+
|
41 |
+
dtype: bfloat16
|