Commit
•
5293f64
1
Parent(s):
4c68f8e
Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,85 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: cc-by-4.0
|
| 3 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: cc-by-4.0
|
| 3 |
+
---
|
| 4 |
+
|
| 5 |
+
# #llama-3 #roleplay
|
| 6 |
+
|
| 7 |
+
GGUF-IQ-Imatrix quants for [cgato/L3-TheSpice-8b-v0.8.3](https://huggingface.co/cgato/L3-TheSpice-8b-v0.8.3).
|
| 8 |
+
|
| 9 |
+
> [!IMPORTANT]
|
| 10 |
+
> These quants have already been done after the fixes from [llama.cpp/pull/6920](https://github.com/ggerganov/llama.cpp/pull/6920). <br>
|
| 11 |
+
> Use **KoboldCpp version 1.64** or higher.
|
| 12 |
+
|
| 13 |
+
> [!NOTE]
|
| 14 |
+
> **Prompt formatting...** <br>
|
| 15 |
+
> Seems simple so I'd try Alpaca or ChatML and see how it goes.
|
| 16 |
+
|
| 17 |
+
# Original model information by the author:
|
| 18 |
+
|
| 19 |
+
Now not overtrained and with the tokenizer fix to base llama3. Trained for 3 epochs.
|
| 20 |
+
|
| 21 |
+
The latest TheSpice, dipped in Mama Liz's LimaRP Oil.
|
| 22 |
+
I've focused on making the model more flexible and provide a more unique experience.
|
| 23 |
+
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.
|
| 24 |
+
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.
|
| 25 |
+
|
| 26 |
+
|
| 27 |
+
## Datasets Used
|
| 28 |
+
|
| 29 |
+
* Capybara
|
| 30 |
+
* Claude Multiround 30k
|
| 31 |
+
* Augmental
|
| 32 |
+
* ToxicQA
|
| 33 |
+
* Yahoo Answers
|
| 34 |
+
* Airoboros 3.1
|
| 35 |
+
* LimaRP
|
| 36 |
+
|
| 37 |
+
## Features ( Examples from 0.1.1 because I'm too lazy to take new screenshots. Its tested tho. )
|
| 38 |
+
|
| 39 |
+
Narration
|
| 40 |
+
|
| 41 |
+
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.
|
| 42 |
+
|
| 43 |
+
# You can look at anything mostly as long as you end it with "What do I see?"
|
| 44 |
+
|
| 45 |
+
|
| 46 |
+
|
| 47 |
+
# You can also request to know what a character is thinking or planning.
|
| 48 |
+
|
| 49 |
+
|
| 50 |
+
|
| 51 |
+
# You can ask for a quick summary on the character as well.
|
| 52 |
+
|
| 53 |
+
|
| 54 |
+
|
| 55 |
+
# Before continuing the conversation as normal.
|
| 56 |
+
|
| 57 |
+
|
| 58 |
+
|
| 59 |
+
## Prompt Format: Chat ( The default Ooba template and Silly Tavern Template )
|
| 60 |
+
|
| 61 |
+
|
| 62 |
+
|
| 63 |
+
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.
|
| 64 |
+
|
| 65 |
+
|
| 66 |
+
```
|
| 67 |
+
{System Prompt}
|
| 68 |
+
|
| 69 |
+
Username: {Input}
|
| 70 |
+
BotName: {Response}
|
| 71 |
+
Username: {Input}
|
| 72 |
+
BotName: {Response}
|
| 73 |
+
|
| 74 |
+
```
|
| 75 |
+
## Presets
|
| 76 |
+
|
| 77 |
+
All screenshots above were taken with the below SillyTavern Preset.
|
| 78 |
+
## Recommended Silly Tavern Preset -> (Temp: 1.25, MinP: 0.1, RepPen: 1.05)
|
| 79 |
+
This is a roughly equivalent Kobold Horde Preset.
|
| 80 |
+
## Recommended Kobold Horde Preset -> MinP
|
| 81 |
+
|
| 82 |
+
|
| 83 |
+
# Disclaimer
|
| 84 |
+
|
| 85 |
+
Please prompt responsibly and take anything outputted by any Language Model with a huge grain of salt. Thanks!
|