Lewdiculous
commited on
Commit
•
dd09949
1
Parent(s):
5bdcb66
Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,103 @@
|
|
1 |
---
|
2 |
license: cc-by-4.0
|
|
|
|
|
|
|
|
|
|
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
license: cc-by-4.0
|
3 |
+
inference: false
|
4 |
+
tags:
|
5 |
+
- gguf
|
6 |
+
- mistral
|
7 |
+
- roleplay
|
8 |
---
|
9 |
+
This repository hosts GGUF-IQ-Imatrix quants for [ResplendentAI/Persephone_7B](https://huggingface.co/ResplendentAI/Persephone_7B).
|
10 |
+
|
11 |
+
*The return of a cult classic.*
|
12 |
+
|
13 |
+
Quants:
|
14 |
+
```python
|
15 |
+
quantization_options = [
|
16 |
+
"Q4_K_M", "Q4_K_S", "IQ4_XS", "Q5_K_M", "Q5_K_S",
|
17 |
+
"Q6_K", "Q8_0", "IQ3_M", "IQ3_S", "IQ3_XXS"
|
18 |
+
]
|
19 |
+
```
|
20 |
+
|
21 |
+
**What does "Imatrix" mean?**
|
22 |
+
|
23 |
+
It stands for **Importance Matrix**, a technique used to improve the quality of quantized models.
|
24 |
+
The **Imatrix** is calculated based on calibration data, and it helps determine the importance of different model activations during the quantization process.
|
25 |
+
The idea is to preserve the most important information during quantization, which can help reduce the loss of model performance, especially when the calibration data is diverse.
|
26 |
+
[[1]](https://github.com/ggerganov/llama.cpp/discussions/5006) [[2]](https://github.com/ggerganov/llama.cpp/discussions/5263#discussioncomment-8395384)
|
27 |
+
|
28 |
+
For imatrix data generation, kalomaze's `groups_merged.txt` with added roleplay chats was used, you can find it [here](https://huggingface.co/Lewdiculous/Datura_7B-GGUF-Imatrix/blob/main/imatrix-with-rp-format-data.txt). This was just to add a bit more diversity to the data.
|
29 |
+
|
30 |
+
**Steps:**
|
31 |
+
|
32 |
+
```
|
33 |
+
Base⇢ GGUF(F16)⇢ Imatrix-Data(F16)⇢ GGUF(Imatrix-Quants)
|
34 |
+
```
|
35 |
+
*Using the latest llama.cpp at the time.*
|
36 |
+
|
37 |
+
# Original model information:
|
38 |
+
|
39 |
+
Officially rebranding Thespis to **TheSpice**. Why? Cause its a cooler, simpler name.
|
40 |
+
I've focused on making the model more flexible and provide a more unique experience.
|
41 |
+
I'm still working on cleaning up my dataset, but I've shrunken it down a lot to focus on a "less is more" approach.
|
42 |
+
This is ultimate a return to form of the way I used to train Thespis, with more of a focus on a small hand edited dataset.
|
43 |
+
|
44 |
+
|
45 |
+
## Datasets Used
|
46 |
+
|
47 |
+
* Dolphin
|
48 |
+
* Ultrachat
|
49 |
+
* Capybara
|
50 |
+
* Augmental
|
51 |
+
* ToxicQA
|
52 |
+
* Yahoo Answers
|
53 |
+
* Airoboros 3.1
|
54 |
+
|
55 |
+
## Features
|
56 |
+
|
57 |
+
Narration
|
58 |
+
|
59 |
+
If you request information on objects or characters in the scene, the model will narrate it to you. Most of the time, without moving the story forward.
|
60 |
+
|
61 |
+
# You can look at anything mostly as long as you end it with "What do I see?"
|
62 |
+
|
63 |
+
![image/png](https://cdn-uploads.huggingface.co/production/uploads/64dd7cda3d6b954bf7cdd922/VREY8QHtH6fCL0fCp8AAC.png)
|
64 |
+
|
65 |
+
# You can also request to know what a character is thinking or planning.
|
66 |
+
|
67 |
+
![image/png](https://cdn-uploads.huggingface.co/production/uploads/64dd7cda3d6b954bf7cdd922/U3RTAgbaB2m1ygfZGJ-SM.png)
|
68 |
+
|
69 |
+
# You can ask for a quick summary on the character as well.
|
70 |
+
|
71 |
+
![image/png](https://cdn-uploads.huggingface.co/production/uploads/64dd7cda3d6b954bf7cdd922/uXFd6GhnXS8w_egUEfcAp.png)
|
72 |
+
|
73 |
+
# Before continuing the conversation as normal.
|
74 |
+
|
75 |
+
![image/png](https://cdn-uploads.huggingface.co/production/uploads/64dd7cda3d6b954bf7cdd922/dYTQUdCshUDtp_BJ20tHy.png)
|
76 |
+
|
77 |
+
## Prompt Format: Chat ( The default Ooba template and Silly Tavern Template )
|
78 |
+
|
79 |
+
![image/png](https://cdn-uploads.huggingface.co/production/uploads/64dd7cda3d6b954bf7cdd922/59vi4VWP2d0bCbsW2eU8h.png)
|
80 |
+
|
81 |
+
If you're using Ooba in verbose mode as a server, you can check if you're console is logging something that looks like this.
|
82 |
+
![image/png](https://cdn-uploads.huggingface.co/production/uploads/64dd7cda3d6b954bf7cdd922/mB3wZqtwN8B45nR7W1fgR.png)
|
83 |
+
|
84 |
+
```
|
85 |
+
{System Prompt}
|
86 |
+
|
87 |
+
Username: {Input}
|
88 |
+
BotName: {Response}
|
89 |
+
Username: {Input}
|
90 |
+
BotName: {Response}
|
91 |
+
|
92 |
+
```
|
93 |
+
## Presets
|
94 |
+
|
95 |
+
All screenshots above were taken with the below SillyTavern Preset.
|
96 |
+
## Recommended Silly Tavern Preset -> (Temp: 1.25, MinP: 0.1, RepPen: 1.05)
|
97 |
+
This is a roughly equivalent Kobold Horde Preset.
|
98 |
+
## Recommended Kobold Horde Preset -> MinP
|
99 |
+
|
100 |
+
|
101 |
+
# Disclaimer
|
102 |
+
|
103 |
+
Please prompt responsibly and take anything outputted by any Language Model with a huge grain of salt. Thanks!
|