---
library_name: transformers
license: llama3
base_model:
- nbeerbower/llama-3-Stheno-Mahou-8B
datasets:
- flammenai/FlameMix-DPO-v1
- flammenai/Grill-preprod-v1_chatML
- flammenai/Grill-preprod-v2_chatML
---
**Exllamav2** quant (**exl2** / **2.2 bpw**) made with ExLlamaV2 v0.0.21
Other EXL2 quants:
| **Quant** | **Model Size** | **lm_head** |
| ----- | ---------- | ------- |
|
**[2.2](https://huggingface.co/Zoyd/flammenai_Mahou-1.2a-llama3-8B-2_2bpw_exl2)** | 3250 MB | 6 |
|**[2.5](https://huggingface.co/Zoyd/flammenai_Mahou-1.2a-llama3-8B-2_5bpw_exl2)** | 3479 MB | 6 |
|**[3.0](https://huggingface.co/Zoyd/flammenai_Mahou-1.2a-llama3-8B-3_0bpw_exl2)** | 3893 MB | 6 |
|**[3.5](https://huggingface.co/Zoyd/flammenai_Mahou-1.2a-llama3-8B-3_5bpw_exl2)** | 4311 MB | 6 |
|**[3.75](https://huggingface.co/Zoyd/flammenai_Mahou-1.2a-llama3-8B-3_75bpw_exl2)** | 4518 MB | 6 |
|**[4.0](https://huggingface.co/Zoyd/flammenai_Mahou-1.2a-llama3-8B-4_0bpw_exl2)** | 4727 MB | 6 |
|**[4.25](https://huggingface.co/Zoyd/flammenai_Mahou-1.2a-llama3-8B-4_25bpw_exl2)** | 4935 MB | 6 |
|**[5.0](https://huggingface.co/Zoyd/flammenai_Mahou-1.2a-llama3-8B-5_0bpw_exl2)** | 5557 MB | 6 |
|**[6.0](https://huggingface.co/Zoyd/flammenai_Mahou-1.2a-llama3-8B-6_0bpw_exl2)** | 6496 MB | 8 |
|**[6.5](https://huggingface.co/Zoyd/flammenai_Mahou-1.2a-llama3-8B-6_5bpw_exl2)** | 6902 MB | 8 |
|**[8.0](https://huggingface.co/Zoyd/flammenai_Mahou-1.2a-llama3-8B-8_0bpw_exl2)** | 8131 MB | 8 |
![image/png](https://huggingface.co/flammenai/Mahou-1.0-mistral-7B/resolve/main/mahou1.png)
# Mahou-1.2a-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.
### 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.
```
### ST Settings
1. Use ChatML for the Context Template.
2. Turn on Instruct Mode for ChatML.
3. Use the following stopping strings: `["<", "|", "<|", "\n"]`
### Method
Finetuned using an A100 on Google Colab.
[Fine-tune a Mistral-7b model with Direct Preference Optimization](https://towardsdatascience.com/fine-tune-a-mistral-7b-model-with-direct-preference-optimization-708042745aac) - [Maxime Labonne](https://huggingface.co/mlabonne)
### Configuration
LoRA, model, and training settings:
```python
# LoRA configuration
peft_config = LoraConfig(
r=16,
lora_alpha=16,
lora_dropout=0.05,
bias="none",
task_type="CAUSAL_LM",
target_modules=['k_proj', 'gate_proj', 'v_proj', 'up_proj', 'q_proj', 'o_proj', 'down_proj']
)
# Model to fine-tune
model = AutoModelForCausalLM.from_pretrained(
model_name,
torch_dtype=torch.bfloat16,
load_in_4bit=True
)
model.config.use_cache = False
# Reference model
ref_model = AutoModelForCausalLM.from_pretrained(
model_name,
torch_dtype=torch.bfloat16,
load_in_4bit=True
)
# Training arguments
training_args = TrainingArguments(
per_device_train_batch_size=4,
gradient_accumulation_steps=4,
gradient_checkpointing=True,
learning_rate=5e-5,
lr_scheduler_type="cosine",
max_steps=2000,
save_strategy="no",
logging_steps=1,
output_dir=new_model,
optim="paged_adamw_32bit",
warmup_steps=100,
bf16=True,
report_to="wandb",
)
# Create DPO trainer
dpo_trainer = DPOTrainer(
model,
ref_model,
args=training_args,
train_dataset=dataset,
tokenizer=tokenizer,
peft_config=peft_config,
beta=0.1,
force_use_ref_model=True
)
# Fine-tune model with DPO
dpo_trainer.train()
```