File size: 2,055 Bytes
14c036f
 
 
 
 
0a7ea71
 
14c036f
 
 
 
 
 
 
 
 
 
2139a26
14c036f
 
 
 
 
 
 
4a3dbb2
0a7ea71
14c036f
 
 
 
 
 
1809c23
14c036f
 
 
 
 
 
 
 
2139a26
95f92ed
2139a26
14c036f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
license: cc-by-nc-4.0
tags:
- not-for-all-audiences
- nsfw
language:
- en
---
## Miqu PlayMaid 70B v0.1

<center>
Image to be replaced :P
<div style="width: 100%;">
    <img src="https://cdn-uploads.huggingface.co/production/uploads/63ab1241ad514ca8d1430003/tPFdudSae6SCDNvhe1lC9.png" style="display: block; margin: auto;">
</div>
</center>

This model uses the Alpaca **prompting format**. Unlike the base model this one is tuned on **16384 sequence length**. This was due to memory constraints.

This is a roleplay focused finetune using MiquMaid v2 70B DPO by [Undi95](https://huggingface.co/Undi95) and [IkariDev](https://huggingface.co/IkariDev) as base.

My goal was to further push MiquMaid into creative writing and roleplay, hopefully adding some spice to it. Ultimately I want to follow Undi's and Ikari's approach and combine it into a 2x70B alongside their Miqu 70B DPO model.

## Credits:
- Netrve
- Special thanks to everyone on TheBloke's Discord who helped me!

Credits for MiquMaid:
- Undi
- IkariDev

## Description

This repo contains BFP16 files of Miqu-PlayMaid-70B-v0.1

## Training data used:
- [AndrijDavid Roleplay](https://huggingface.co/datasets/Undi95/andrijdavid_roleplay-conversation-sharegpt)
- [LimaRP Augmented](https://huggingface.co/datasets/grimulkan/LimaRP-augmented)
- [Airoboros 3.2](https://huggingface.co/datasets/jondurbin/airoboros-3.2)
- [Bluemoon Fandom RP](https://huggingface.co/datasets/Squish42/bluemoon-fandom-1-1-rp-cleaned)

### Format:

I can only guarantee that it works with a maximum context length of 16384. Might work with 32768, but no promises. 

```
### Instruction:
{system prompt}
### Input:
{input}
### Response:
{reply}
```

## Support

- Netrve: If you like what I did, feel free to support future work [here](https://ko-fi.com/netrve)

Don't forget to send some love to the original masterminds behind MiquMaid:

- Undi: If you want to support us, you can [here](https://ko-fi.com/undiai)
- IkariDev: Visit my [retro/neocities style website](https://ikaridevgit.github.io/) please kek