Instructions to use NeuralNovel/Tanuki-7B-v0.1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use NeuralNovel/Tanuki-7B-v0.1 with Transformers:

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

pipe = pipeline("text-generation", model="NeuralNovel/Tanuki-7B-v0.1")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("NeuralNovel/Tanuki-7B-v0.1")
model = AutoModelForCausalLM.from_pretrained("NeuralNovel/Tanuki-7B-v0.1")
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=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use NeuralNovel/Tanuki-7B-v0.1 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "NeuralNovel/Tanuki-7B-v0.1"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "NeuralNovel/Tanuki-7B-v0.1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/NeuralNovel/Tanuki-7B-v0.1

SGLang

How to use NeuralNovel/Tanuki-7B-v0.1 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 "NeuralNovel/Tanuki-7B-v0.1" \
    --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": "NeuralNovel/Tanuki-7B-v0.1",
		"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 "NeuralNovel/Tanuki-7B-v0.1" \
        --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": "NeuralNovel/Tanuki-7B-v0.1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use NeuralNovel/Tanuki-7B-v0.1 with Docker Model Runner:
```
docker model run hf.co/NeuralNovel/Tanuki-7B-v0.1
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

NeuralNovel/Tanuki-7B-v0.1

Designed to generate instructive and narrative text, with a specific focus on roleplay & short storytelling. This fine-tune has been tailored to provide detailed and creative responses in the context of complex narrative.

Full-parameter fine-tune (FFT) of Mistral-7B-Instruct-v0.2, with apache-2.0 license, suitable for commercial or non-commercial use.

Data-set

The model was finetuned using the Neural-Story-v1 and Creative-Logic-v1 datasets.

Summary

Fine-tuned with the intention of generating creative and narrative text, making it more suitable for creative writing prompts and storytelling.

Out-of-Scope Use

The model may not perform well in scenarios unrelated to instructive and narrative text generation. Misuse or applications outside its designed scope may result in suboptimal outcomes.

Bias, Risks, and Limitations

The model may exhibit biases or limitations inherent in the training data. It is essential to consider these factors when deploying the model to avoid unintended consequences.

This model and its datasets serves as an excellent starting point for testing language models, users are advised to exercise caution, as there might be some inherent genre or writing bias.

Hardware and Training

Trained using NVIDIA Tesla T40 24 GB.


 n_epochs = 4, # increased from 3
 n_checkpoints = 2,
 batch_size = 6, # decreased from 20
 learning_rate = 1e-5,

Sincere appreciation to Techmind for their generous sponsorship.

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	64.74
AI2 Reasoning Challenge (25-Shot)	62.80
HellaSwag (10-Shot)	83.14
MMLU (5-Shot)	60.54
TruthfulQA (0-shot)	66.33
Winogrande (5-shot)	75.85
GSM8k (5-shot)	39.80