File size: 1,505 Bytes
5a10e60
 
fbcd3b9
 
 
 
 
5a10e60
fbcd3b9
5a10e60
fbcd3b9
5a10e60
fbcd3b9
5a10e60
fbcd3b9
5a10e60
fbcd3b9
5a10e60
fbcd3b9
5a10e60
fbcd3b9
5a10e60
429c1fb
5a10e60
fbcd3b9
 
 
 
 
 
 
 
5a10e60
fbcd3b9
5a10e60
fbcd3b9
 
5a10e60
fbcd3b9
 
 
5a10e60
fbcd3b9
5a10e60
fbcd3b9
 
 
 
5a10e60
fbcd3b9
5a10e60
fbcd3b9
5a10e60
fbcd3b9
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
---
library_name: transformers
license: llama3
base_model:
- nbeerbower/llama-3-Daredevil-Mahou-8B
datasets:
- flammenai/MahouMix-v1
---
![image/png](https://huggingface.co/flammenai/Mahou-1.0-mistral-7B/resolve/main/mahou1.png)

# Mahou-1.3-llama3-8B

Mahou is our attempt to build a production-ready conversational/roleplay LLM.

Future versions will be released iteratively and finetuned from flammen.ai conversational data.

### License

This model is based on Meta Llama-3-8B and is governed by the [META LLAMA 3 COMMUNITY LICENSE AGREEMENT](LICENSE).

### Chat Format

This model has been trained to use ChatML format. Note the additional tokens in [tokenizer_config.json](tokenizer_config.json).

```
<|im_start|>system
{{system}}<|im_end|>
<|im_start|>{{char}}
{{message}}<|im_end|>
<|im_start|>{{user}}
{{message}}<|im_end|>
```

### Roleplay Format

- Speech without quotes.
- Actions in `*asterisks*`

```
*leans against wall cooly* so like, i just casted a super strong spell at magician academy today, not gonna lie, felt badass.
```

### ST Settings

1. Use ChatML for the Context Template.
2. Enable Instruct Mode.
3. Use the [Mahou preset](https://huggingface.co/datasets/flammenai/Mahou-ST-ChatML-Instruct/raw/main/Mahou.json).
4. Recommended: Add newline as a stopping string: `["\n"]`

### Method

Finetuned for 10 epochs using an A100 on Google Colab.

[Fine-tune Llama 3 with ORPO](https://huggingface.co/blog/mlabonne/orpo-llama-3) - [Maxime Labonne](https://huggingface.co/mlabonne)