Darkknight535
commited on
Commit
•
f4fa10f
1
Parent(s):
a294a80
Update README.md
Browse files
README.md
CHANGED
@@ -3,14 +3,12 @@ library_name: transformers
|
|
3 |
tags:
|
4 |
- not-for-all-audiences
|
5 |
---
|
6 |
-
###
|
7 |
---
|
8 |
-
![Waifu](https://huggingface.co/Darkknight535/
|
9 |
---
|
10 |
-
**Improved Version of OpenCrystal v1**
|
11 |
- **Enhancements**:
|
12 |
-
- **Fine-tuning**: The model has undergone fine-tuning
|
13 |
-
- **Merging**: It has been merged with the base model for improved functionality.
|
14 |
|
15 |
---
|
16 |
|
@@ -28,7 +26,7 @@ Temp : 1.2
|
|
28 |
Min P : 0.05
|
29 |
Repetition Penalty : 1.05
|
30 |
|
31 |
-
[And everything else neutral]
|
32 |
```
|
33 |
|
34 |
## Normal(Precise)
|
@@ -37,27 +35,23 @@ Temp : 0.6 - 0.8
|
|
37 |
Min P : 0.1
|
38 |
Repetition Penalty : 1.1
|
39 |
|
40 |
-
[And everything else neutral]
|
41 |
```
|
42 |
|
43 |
-
|
44 |
-
### Pro Tip
|
45 |
-
**Use Example Messages for multiturn conversations**
|
46 |
-
|
47 |
### Features
|
48 |
-
-
|
49 |
-
-
|
50 |
-
-
|
|
|
51 |
- Output feels similar to using Character.ai
|
52 |
-
- Less Perplexity than v1
|
53 |
- Improved adherence to prompts.
|
54 |
-
-
|
55 |
- Capable of summarizing and generating image prompts well. (The above image's prompt was generated in a roleplay by this model; possible due to the Llama-3-instruct base.)
|
56 |
|
57 |
### System Prompt
|
58 |
```
|
59 |
-
|
60 |
```
|
61 |
|
62 |
### FeedBack
|
63 |
-
[FeedBack here](https://huggingface.co/Darkknight535/
|
|
|
3 |
tags:
|
4 |
- not-for-all-audiences
|
5 |
---
|
6 |
+
### MoonLight-L3-15B
|
7 |
---
|
8 |
+
![Waifu](https://huggingface.co/Darkknight535/)
|
9 |
---
|
|
|
10 |
- **Enhancements**:
|
11 |
+
- **Fine-tuning**: The model has undergone fine-tuning on different RP datasheets and some dark nsfw datasheets.
|
|
|
12 |
|
13 |
---
|
14 |
|
|
|
26 |
Min P : 0.05
|
27 |
Repetition Penalty : 1.05
|
28 |
|
29 |
+
[And everything else neutral] (DRY Sampler Breaks it)
|
30 |
```
|
31 |
|
32 |
## Normal(Precise)
|
|
|
35 |
Min P : 0.1
|
36 |
Repetition Penalty : 1.1
|
37 |
|
38 |
+
[And everything else neutral](DRY Sampler Breaks it)
|
39 |
```
|
40 |
|
|
|
|
|
|
|
|
|
41 |
### Features
|
42 |
+
- Supports 16k through rope scaling (finetuned on a 16k datasheet)
|
43 |
+
- Automatically speaks as other NPCs.
|
44 |
+
- Creative output(Like the OpenCrystal 12b but more coherent).
|
45 |
+
- Coherent responses(Accuatlly having 15B model logic).
|
46 |
- Output feels similar to using Character.ai
|
|
|
47 |
- Improved adherence to prompts.
|
48 |
+
- No hallucinations (15B).
|
49 |
- Capable of summarizing and generating image prompts well. (The above image's prompt was generated in a roleplay by this model; possible due to the Llama-3-instruct base.)
|
50 |
|
51 |
### System Prompt
|
52 |
```
|
53 |
+
Develop the plot slowly, always stay in character. Describe all actions in full, elaborate, explicit, graphic, and vivid detail. Mention all relevant sensory perceptions. Keep the story immersive and engaging. Speak as other person when needed and prefix with the name of person you're speaking as except {{user}}.
|
54 |
```
|
55 |
|
56 |
### FeedBack
|
57 |
+
[FeedBack here](https://huggingface.co/Darkknight535/Darkknight535/Moonlight-L3-15B-16k/discussions/1)
|