Instructions to use Delta-Vector/Nanuq-R1-14B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Delta-Vector/Nanuq-R1-14B with Transformers:

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

pipe = pipeline("text-generation", model="Delta-Vector/Nanuq-R1-14B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Delta-Vector/Nanuq-R1-14B")
model = AutoModelForCausalLM.from_pretrained("Delta-Vector/Nanuq-R1-14B")
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

vLLM

How to use Delta-Vector/Nanuq-R1-14B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Delta-Vector/Nanuq-R1-14B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Delta-Vector/Nanuq-R1-14B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Delta-Vector/Nanuq-R1-14B

SGLang

How to use Delta-Vector/Nanuq-R1-14B 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 "Delta-Vector/Nanuq-R1-14B" \
    --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": "Delta-Vector/Nanuq-R1-14B",
		"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 "Delta-Vector/Nanuq-R1-14B" \
        --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": "Delta-Vector/Nanuq-R1-14B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Delta-Vector/Nanuq-R1-14B with Docker Model Runner:
```
docker model run hf.co/Delta-Vector/Nanuq-R1-14B
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

Created by Delta-Vector →

Model Information

Nanuq-R1 14B

GRPO Experiment Q3-235B-8B Merge/heal Competant assistant with decent writing!

A sequel! The new Nanuq series is meant to be as a testing grounds for my GRPO experiments, This model is a full post-train heal of Snwy's Frankenmerge between Q3 235B and Q3 8B.

Pretrained for 2 epochs on 1B tokens of Creative Writing data, Then SFT with alot of my own and Pocketdoc's Instruct dataset, and then GRPO'd with the Claude-2.7K dataset in an attempt to align it to be more like Claude with POLARS and Verifiers

There's alot of things i could do different, As the reward almost falls flat as soon as you get out of warm-up but this model was pretty decent so decided to release it(Esp considering it's starting place), Hope people enjoy it!

Quantized Versions

Available Downloads

GGUF FormatFor use with LLama.cpp & Forks(Coming Soon!)
EXL2 FormatFor use with TabbyAPI (Coming soon!)

Prompting

Model has been tuned with the ChatML formatting. A typical input would look like this:

"""<|im_start|>user
Hi there!<|im_end|>
<|im_start|>assistant
Nice to meet you!<|im_end|>
<|im_start|>user
Can I ask a question?<|im_end|>
<|im_start|>assistant
"""

Training

The training was done for 2 epochs of Pretraining and 2 epochs of SFT and finally 500 steps of GRPO using Verifiers with 8 x H200s GPUs for the fine-tuning of the model.