Update README.md
Browse files
README.md
CHANGED
@@ -1,6 +1,39 @@
|
|
1 |
---
|
2 |
library_name: peft
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
4 |
## Training procedure
|
5 |
|
6 |
|
|
|
1 |
---
|
2 |
library_name: peft
|
3 |
+
datasets:
|
4 |
+
- Squish42/bluemoon-fandom-1-1-rp-cleaned
|
5 |
+
- OpenLeecher/Teatime
|
6 |
+
- PygmalionAI/PIPPA
|
7 |
+
tags:
|
8 |
+
- not-for-all-audiences
|
9 |
+
- nsfw
|
10 |
---
|
11 |
+
## What is PetrolLM?
|
12 |
+
PetrolLM is [Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1) model fine-tune using QLoRA (4-bit precision) for the purposes of creative writing and roleplay.
|
13 |
+
|
14 |
+
The dataset consists of 5800 samples, with the composition as follows:
|
15 |
+
* AICG Logs (~17%)
|
16 |
+
* PygmalionAI/PIPPA (~17%)
|
17 |
+
* Squish42/bluemoon-fandom-1-1-rp-cleaned (~13%)
|
18 |
+
* OpenLeecher/Teatime (~2%)
|
19 |
+
|
20 |
+
These samples were then back-filled using gpt-4/gpt-3.5-turbo-16k or otherwise converted to fit the prompt format.
|
21 |
+
|
22 |
+
## Prompt Format
|
23 |
+
The model was finetuned with a prompt format similar to the original SuperHOT prototype:
|
24 |
+
```
|
25 |
+
---
|
26 |
+
style: roleplay
|
27 |
+
characters:
|
28 |
+
[char]: [description]
|
29 |
+
summary: [scenario]
|
30 |
+
---
|
31 |
+
<chat_history>
|
32 |
+
Format:
|
33 |
+
[char]: [message]
|
34 |
+
Human: [message]
|
35 |
+
```
|
36 |
+
|
37 |
## Training procedure
|
38 |
|
39 |
|