Instructions to use tokennext/llama-3-8b-elyza-ja-werewolf with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use tokennext/llama-3-8b-elyza-ja-werewolf with Transformers:

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

pipe = pipeline("text-generation", model="tokennext/llama-3-8b-elyza-ja-werewolf")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("tokennext/llama-3-8b-elyza-ja-werewolf")
model = AutoModelForCausalLM.from_pretrained("tokennext/llama-3-8b-elyza-ja-werewolf", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=256)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use tokennext/llama-3-8b-elyza-ja-werewolf with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "tokennext/llama-3-8b-elyza-ja-werewolf"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "tokennext/llama-3-8b-elyza-ja-werewolf",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/tokennext/llama-3-8b-elyza-ja-werewolf

SGLang

How to use tokennext/llama-3-8b-elyza-ja-werewolf 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 "tokennext/llama-3-8b-elyza-ja-werewolf" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "tokennext/llama-3-8b-elyza-ja-werewolf",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

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 "tokennext/llama-3-8b-elyza-ja-werewolf" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "tokennext/llama-3-8b-elyza-ja-werewolf",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use tokennext/llama-3-8b-elyza-ja-werewolf with Docker Model Runner:
```
docker model run hf.co/tokennext/llama-3-8b-elyza-ja-werewolf
```

🔧 DPO（人狼ゲーム用）

このモデルは、オリジナルの elyza/Llama-3-ELYZA-JP-8B をベースに、人狼ゲームのセリフ生成タスク向けに DPO（Direct Preference Optimization） を用いて学習を施したものです。

学習目的

人狼ゲームにおける「取り乱す村人」や「しらを切る人狼」、「情けない行動」など、特定の感情表現を伴うセリフを高品質に生成できるようにすることを目的としました。

データについて

データ数は少数（百数ペア）であり、ハッカソン形式で迅速に作成されたものです。

使用手法

DPO (Direct Preference Optimization)
LoRAなどの軽量手法と併用
トレーニングには trl ライブラリを使用

注意点

データ量が少ないため、本モデルはあくまで試験的なモデルです。

Downloads last month: 9

Safetensors

Model size

8B params

Tensor type

F32