Instructions to use Sao10K/Llama-3.3-70B-Vulpecula-r1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Sao10K/Llama-3.3-70B-Vulpecula-r1 with Transformers:

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

pipe = pipeline("text-generation", model="Sao10K/Llama-3.3-70B-Vulpecula-r1")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Sao10K/Llama-3.3-70B-Vulpecula-r1")
model = AutoModelForCausalLM.from_pretrained("Sao10K/Llama-3.3-70B-Vulpecula-r1", 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=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Inference
Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Sao10K/Llama-3.3-70B-Vulpecula-r1 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Sao10K/Llama-3.3-70B-Vulpecula-r1"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Sao10K/Llama-3.3-70B-Vulpecula-r1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Sao10K/Llama-3.3-70B-Vulpecula-r1

SGLang

How to use Sao10K/Llama-3.3-70B-Vulpecula-r1 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 "Sao10K/Llama-3.3-70B-Vulpecula-r1" \
    --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": "Sao10K/Llama-3.3-70B-Vulpecula-r1",
		"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 "Sao10K/Llama-3.3-70B-Vulpecula-r1" \
        --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": "Sao10K/Llama-3.3-70B-Vulpecula-r1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Sao10K/Llama-3.3-70B-Vulpecula-r1 with Docker Model Runner:
```
docker model run hf.co/Sao10K/Llama-3.3-70B-Vulpecula-r1
```

Character's thoughts inside <think> tags instead of LLM's reflection

by Ubr - opened Mar 27, 2025

Discussion

Ubr

Mar 27, 2025

•

edited Mar 28, 2025

The DeepSeek R1 models usually reflect inside the <think> tags how to create a perfect character reply. Vulpecula-R1 instead shows the character's thoughts inside the <think> tags. Is this intended behavior or am I prompting it wrongly?

Sao10K

Owner Mar 28, 2025

•

edited Mar 28, 2025

I've been mainly using LeCeption by Steelskull with thinking:

https://huggingface.co/Steelskull/L3.3-Electra-R1-70b/blob/main/LeCeption-XML-V2.json

And the Euryale v2.1 Prompt without thinking:

https://huggingface.co/Sao10K/L3-70B-Euryale-v2.1/blob/main/Euryale-v2.1-Llama-3-Instruct.json

Unfortunately can't really help you much, I spent a few hours testing and never ran into that issue on my own.

Ubr

Mar 28, 2025

•

edited Mar 28, 2025

I tried LeCeption in the exact version you provided. And if I add the <think> prefix in SillyTaver, the replies will contain <think> content with the characters thoughts. Do you actually get narrator thoughts inside the <think> tags in your setup?

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment