File size: 2,049 Bytes
c5b533d
32f2e0b
 
 
 
 
 
c5b533d
 
 
9a95b06
32f2e0b
 
9a95b06
 
32f2e0b
11104b1
 
 
 
 
 
 
 
 
abbcd9a
11104b1
 
 
 
32f2e0b
 
 
7bfd536
 
4a74ed0
389d633
 
4a74ed0
32f2e0b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
812fe15
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
---
license: apache-2.0
base_model:
- mistralai/Mistral-Nemo-Instruct-2407
tags:
- roleplay
- conversational
language:
- en
---
# Nemo 12b RP Ink
![image/png](https://cdn-uploads.huggingface.co/production/uploads/634262af8d8089ebaefd410e/-yOYQdx9p3TjHLSq2RrRf.png)

A roleplay-focused LoRA finetune of Mistral Nemo Instruct. Methodology and hyperparams inspired by [SorcererLM](https://huggingface.co/rAIfle/SorcererLM-8x22b-bf16) and [Slush](https://huggingface.co/crestf411/Q2.5-32B-Slush).  
Renamed to Ink to distinguish from [insert every other rp tune ever], but it's the same data as was used in the [Teleut RP model](https://huggingface.co/allura-org/Teleut-7b-RP).

## Testimonials
> dude Fizz, Ink is fucking **gas**

\- inflatebot

> First impression, I put this higher than Rocinante and Magnum Picaro
> 
\- ShotMisser64

> Never tried 12B, but this one is GOAT
> 
\- Evoc


## Dataset
The worst mix of data you've ever seen. Like, seriously, you do not want to see the things that went into this model. It's bad.

"this is like washing down an adderall with a bottle of methylated rotgut" - inflatebot

## Quants
- [Static GGUFs (thanks auri!)](https://huggingface.co/allura-org/MN-12b-RP-Ink-GGUF)
- [Imatrix GGUFs (thanks bart!)](https://huggingface.co/bartowski/MN-12b-RP-Ink-GGUF)

## Recommended Settings
Chat template: Mistral v3-Tekken  
Recommended samplers (not the be-all-end-all, try some on your own!):
- Temp 1.25 / MinP 0.1
- Temp 1.03 / TopK 200 / MinP 0.05 / TopA 0.2

## Hyperparams
### General
- Epochs = 2
- LR = 6e-5
- LR Scheduler = Cosine
- Optimizer = Paged AdamW 8bit
- Effective batch size = 12
### LoRA
- Rank = 16
- Alpha = 32
- Dropout = 0.25 (Inspiration: [Slush](https://huggingface.co/crestf411/Q2.5-32B-Slush))

## Credits
Humongous thanks to the people who created the data. I would credit you all, but that would be cheating ;)  
Big thanks to all Allura members, especially Toasty, for testing and emotional support ilya /platonic  
Also special thanks to Bot for making the model card image here :3