Kooten commited on
Commit
a6c17ec
1 Parent(s): 6879422

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +46 -0
README.md ADDED
@@ -0,0 +1,46 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: cc-by-nc-4.0
3
+ tags:
4
+ - not-for-all-audiences
5
+ - nsfw
6
+ ---
7
+ ## Description
8
+
9
+ Exllama 2 quant of [Undi95/U-Amethyst-20B](https://huggingface.co/Undi95/U-Amethyst-20B)
10
+
11
+ 3 BPW
12
+
13
+ ## Prompt template
14
+
15
+ [Recommended reading](https://huggingface.co/lemonilia/LimaRP-Llama2-13B-v3-EXPERIMENT)
16
+
17
+ You can follow these instruction format settings in SillyTavern. Replace `tiny` with
18
+ your desired response length:
19
+
20
+ ![settings](https://files.catbox.moe/6lcz0u.png)
21
+
22
+ ### Message length control
23
+ Inspired by the previously named "Roleplay" preset in SillyTavern, starting from this
24
+ version of LimaRP it is possible to append a length modifier to the response instruction
25
+ sequence, like this:
26
+
27
+ ```
28
+ ### Input
29
+ User: {utterance}
30
+
31
+ ### Response: (length = medium)
32
+ Character: {utterance}
33
+ ```
34
+
35
+ This has an immediately noticeable effect on bot responses. The available lengths are:
36
+ `tiny`, `short`, `medium`, `long`, `huge`, `humongous`, `extreme`, `unlimited`. **The
37
+ recommended starting length is `medium`**. Keep in mind that the AI may ramble
38
+ or impersonate the user with very long messages.
39
+
40
+ The length control effect is reproducible, but the messages will not necessarily follow
41
+ lengths very precisely, rather follow certain ranges on average, as seen in this table
42
+ with data from tests made with one reply at the beginning of the conversation:
43
+
44
+ ![lengths](https://files.catbox.moe/dy39bt.png)
45
+
46
+ Response length control appears to work well also deep into the conversation.