Mahou
Collection
flammen.ai's production model for casual conversation and character roleplay
•
24 items
•
Updated
•
4
Mahou-1.5-llama3.1-70B
Mahou is designed to provide short messages in a conversational context. It is capable of casual conversation and character roleplay.
Chat Format
This model has been trained to use ChatML format.
<|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.
SillyTavern Settings
- Use ChatML for the Context Template.
- Enable Instruct Mode.
- Use the Mahou ChatML Instruct preset.
- Use the Mahou Sampler preset.
Method
ORPO finetuned with 2x H100 for 3 epochs.
Open LLM Leaderboard Evaluation Results
| Metric | Value |
|---|---|
| Avg. | 36.04 |
| IFEval (0-Shot) | 71.47 |
| BBH (3-Shot) | 52.37 |
| MATH Lvl 5 (4-Shot) | 13.14 |
| GPQA (0-shot) | 13.87 |
| MuSR (0-shot) | 23.71 |
| MMLU-PRO (5-shot) | 41.66 |
- Downloads last month
- 20
Inference Providers
NEW
This model is not currently available via any of the supported third-party Inference Providers, and
the model is not deployed on the HF Inference API.
Model tree for flammenai/Mahou-1.5-llama3.1-70B
Base model
mlabonne/Hermes-3-Llama-3.1-70B-lorablated
Finetuned
nbeerbower/Llama3.1-Gutenberg-Doppel-70B
Finetuned
flammenai/Llama3.1-Flammades-70B
Dataset used to train flammenai/Mahou-1.5-llama3.1-70B
Collection including flammenai/Mahou-1.5-llama3.1-70B
Evaluation results
- strict accuracy on IFEval (0-Shot)Open LLM Leaderboard71.470
- normalized accuracy on BBH (3-Shot)Open LLM Leaderboard52.370
- exact match on MATH Lvl 5 (4-Shot)Open LLM Leaderboard13.140
- acc_norm on GPQA (0-shot)Open LLM Leaderboard13.870
- acc_norm on MuSR (0-shot)Open LLM Leaderboard23.710
- accuracy on MMLU-PRO (5-shot)test set Open LLM Leaderboard41.660