|
--- |
|
library_name: transformers |
|
license: llama3 |
|
base_model: |
|
- flammenai/Mahou-1.3-llama3-8B |
|
datasets: |
|
- flammenai/MahouMix-v1 |
|
- flammenai/FlameMix-DPO-v1 |
|
--- |
|
![image/png](https://huggingface.co/flammenai/Mahou-1.0-mistral-7B/resolve/main/mahou1.png) |
|
|
|
# Mahou-1.3a-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 3 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) |