Instructions to use decem/Dionysus-Mistral-m3-v6 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use decem/Dionysus-Mistral-m3-v6 with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="decem/Dionysus-Mistral-m3-v6")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("decem/Dionysus-Mistral-m3-v6")
model = AutoModelForCausalLM.from_pretrained("decem/Dionysus-Mistral-m3-v6")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use decem/Dionysus-Mistral-m3-v6 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "decem/Dionysus-Mistral-m3-v6"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "decem/Dionysus-Mistral-m3-v6",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/decem/Dionysus-Mistral-m3-v6

SGLang

How to use decem/Dionysus-Mistral-m3-v6 with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "decem/Dionysus-Mistral-m3-v6" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "decem/Dionysus-Mistral-m3-v6",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "decem/Dionysus-Mistral-m3-v6" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "decem/Dionysus-Mistral-m3-v6",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use decem/Dionysus-Mistral-m3-v6 with Docker Model Runner:
```
docker model run hf.co/decem/Dionysus-Mistral-m3-v6
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

Dionysus-Mistral-m3-v6 - A Fine-tuned Language Model

Overview

Dionysus-Mistral-m3-v6 is a language model fine-tuned on the AIDC-ai-business/Marcoroni-7B-v3 architecture. This model is mainly used for roleplay.

Model Details

Developer: DECEM
Base Model: AIDC-ai-business/Marcoroni-7B-v3
Fine-tuning Method: SFT
Language: English

Prompting

Prompt Template for alpaca style

{system prompt} (without the {})

### Instruction:
{user_name}: {prompt} (without the {})

### Response:
{character_name}: (without the {})

stop_words: </s>

Open LLM Leaderboard Evaluation Details

https://huggingface.co/datasets/open-llm-leaderboard/details_decem__Dionysus-Mistral-m3-v6

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	67.10
AI2 Reasoning Challenge (25-Shot)	63.14
HellaSwag (10-Shot)	84.51
MMLU (5-Shot)	62.82
TruthfulQA (0-shot)	49.49
Winogrande (5-shot)	78.45
GSM8k (5-shot)	64.22

Downloads last month: 205

Safetensors

Model size

7B params

Tensor type

BF16

Model tree for decem/Dionysus-Mistral-m3-v6

Quantizations

4 models

Spaces using decem/Dionysus-Mistral-m3-v6 9

Evaluation results

normalized accuracy on AI2 Reasoning Challenge (25-Shot)
test set Open LLM Leaderboard

63.140
normalized accuracy on HellaSwag (10-Shot)
validation set Open LLM Leaderboard

84.510
accuracy on MMLU (5-Shot)
test set Open LLM Leaderboard

62.820
mc2 on TruthfulQA (0-shot)
validation set Open LLM Leaderboard

49.490
accuracy on Winogrande (5-shot)
validation set Open LLM Leaderboard

78.450
accuracy on GSM8k (5-shot)
test set Open LLM Leaderboard

64.220