Commit
•
011d917
1
Parent(s):
4f4a764
Update README.md
Browse files
README.md
CHANGED
@@ -1,42 +1,59 @@
|
|
1 |
---
|
2 |
license: cc-by-nc-4.0
|
3 |
---
|
4 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
5 |
|
6 |
-
|
7 |
|
8 |
-
|
9 |
|
10 |
-
|
|
|
|
|
|
|
|
|
|
|
11 |
|
12 |
- model: NeverSleep/Noromaid-20b-v0.1.1
|
13 |
-
|
14 |
-
|
15 |
-
|
16 |
-
weight: 0.41
|
17 |
-
|
18 |
-
density: 0.50
|
19 |
|
20 |
- model: athirdpath/Eileithyia-20b
|
21 |
-
|
22 |
-
|
23 |
-
|
24 |
-
weight: 0.18
|
25 |
-
|
26 |
-
density: 0.30
|
27 |
|
28 |
- model: athirdpath/CleverGirl-20b-Blended
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
-
weight: 0.41
|
33 |
-
|
34 |
-
density: 0.50
|
35 |
|
36 |
|
37 |
int8_mask: true
|
38 |
|
39 |
dtype: bfloat16
|
40 |
|
41 |
-
|
42 |
-
Thanks to brucethemoose for the recipe.
|
|
|
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-L2-32b</b> </font></p>
|
5 |
+
|
6 |
+
## Description and Role
|
7 |
+
|
8 |
+
Named after a charming daughter of Echo and Pan in Greek myth, Iambe-20b-DARE is a [DARE](https://github.com/yule-BUAA/MergeLM) merge building on my recent experiments.
|
9 |
+
Iambe is intended to be a model that has the best realisticly possible understanding of the scene state
|
10 |
+
remains personable and "human-like" in RP scenarios,
|
11 |
+
|
12 |
+
## Prompting and Context
|
13 |
+
|
14 |
+
Iambe-20b-DARE uses Alpaca formatting, and has an effective context size of 4096 tokens. This model is uncensored, and the output/deployment of this model is the responsibility of the user.
|
15 |
+
|
16 |
+
## Method and Hypothesis
|
17 |
+
|
18 |
+
Based on my extended vanilla model [BigLlama](https://github.com/yule-BUAA/MergeLM), this adds elements of:
|
19 |
+
|
20 |
+
- NeverSleep/Noromaid-20b-v0.1.1 - Addded to adapt the excellent writing and "soul" that come from the datasets backing this model.
|
21 |
+
|
22 |
+
- athirdpath/Eileithyia-20b - Added at low weight and density to capture anatomical data and its relation to fiction without the model's other... quirks.
|
23 |
+
|
24 |
+
- athirdpath/CleverGirl-20b-Blended - Added to capture CleverGirl's problem-solving abilities.
|
25 |
+
|
26 |
+
## Examples
|
27 |
+
|
28 |
+
Coming soon.
|
29 |
|
30 |
+
## Testing and Conclusions
|
31 |
|
32 |
+
VERY Impressed so far, concrete data coming soonish.
|
33 |
|
34 |
+
Does still have some confusion (at q5_k_m), but has instantly become my daily driver.
|
35 |
+
|
36 |
+
## Recipe
|
37 |
+
merge_method: dare_ties
|
38 |
+
|
39 |
+
- base_model: athirdpath/BigLlama-20B
|
40 |
|
41 |
- model: NeverSleep/Noromaid-20b-v0.1.1
|
42 |
+
|
43 |
+
weight: 0.41 / density: 0.50
|
|
|
|
|
|
|
|
|
44 |
|
45 |
- model: athirdpath/Eileithyia-20b
|
46 |
+
|
47 |
+
weight: 0.18 / density: 0.30
|
|
|
|
|
|
|
|
|
48 |
|
49 |
- model: athirdpath/CleverGirl-20b-Blended
|
50 |
+
|
51 |
+
weight: 0.41 / density: 0.50
|
|
|
|
|
|
|
|
|
52 |
|
53 |
|
54 |
int8_mask: true
|
55 |
|
56 |
dtype: bfloat16
|
57 |
|
58 |
+
## Gratitude
|
59 |
+
Thanks to brucethemoose for the recipe. Thanks to Undi95 and IkariDev at NeverSleep for Noromaid, as well as lots of inspiration. Thanks to Sao10K for half of CleverGirl.
|