File size: 3,970 Bytes
5bdcb66 dd09949 5bdcb66 059f3c7 dd09949 328beff 80466a4 dd09949 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 |
---
license: cc-by-4.0
inference: false
tags:
- gguf
- mistral
- roleplay
---
This repository hosts GGUF-IQ-Imatrix quants for [cgato/TheSpice-7b-v0.1.1](https://huggingface.co/cgato/TheSpice-7b-v0.1.1).
*The return of a cult classic.*
Quants:
```python
quantization_options = [
"Q4_K_M", "Q4_K_S", "IQ4_XS", "Q5_K_M", "Q5_K_S",
"Q6_K", "Q8_0", "IQ3_M", "IQ3_S", "IQ3_XXS"
]
```
**What does "Imatrix" mean?**
It stands for **Importance Matrix**, a technique used to improve the quality of quantized models.
The **Imatrix** is calculated based on calibration data, and it helps determine the importance of different model activations during the quantization process.
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.
[[1]](https://github.com/ggerganov/llama.cpp/discussions/5006) [[2]](https://github.com/ggerganov/llama.cpp/discussions/5263#discussioncomment-8395384)
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.
**Steps:**
```
Base⇢ GGUF(F16)⇢ Imatrix-Data(F16)⇢ GGUF(Imatrix-Quants)
```
*Using the latest llama.cpp at the time.*
# Submitted model image:
# Original model information:
Officially rebranding Thespis to **TheSpice**. Why? Cause its a cooler, simpler name.
I've focused on making the model more flexible and provide a more unique experience.
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.
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.
## Datasets Used
* Dolphin
* Ultrachat
* Capybara
* Augmental
* ToxicQA
* Yahoo Answers
* Airoboros 3.1
## Features
Narration
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.
# You can look at anything mostly as long as you end it with "What do I see?"
# You can also request to know what a character is thinking or planning.
# You can ask for a quick summary on the character as well.
# Before continuing the conversation as normal.
## Prompt Format: Chat ( The default Ooba template and Silly Tavern Template )
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.
```
{System Prompt}
Username: {Input}
BotName: {Response}
Username: {Input}
BotName: {Response}
```
## Presets
All screenshots above were taken with the below SillyTavern Preset.
## Recommended Silly Tavern Preset -> (Temp: 1.25, MinP: 0.1, RepPen: 1.05)
This is a roughly equivalent Kobold Horde Preset.
## Recommended Kobold Horde Preset -> MinP
# Disclaimer
Please prompt responsibly and take anything outputted by any Language Model with a huge grain of salt. Thanks! |