Instructions to use lokiii005/Aurora with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use lokiii005/Aurora with PEFT:

from peft import PeftModel
from transformers import AutoModelForCausalLM

base_model = AutoModelForCausalLM.from_pretrained("unsloth/gemma-2-2b-it-bnb-4bit")
model = PeftModel.from_pretrained(base_model, "lokiii005/Aurora")

Transformers

How to use lokiii005/Aurora with Transformers:

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

pipe = pipeline("text-generation", model="lokiii005/Aurora")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("lokiii005/Aurora", dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use lokiii005/Aurora with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "lokiii005/Aurora"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "lokiii005/Aurora",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/lokiii005/Aurora

SGLang

How to use lokiii005/Aurora 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 "lokiii005/Aurora" \
    --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": "lokiii005/Aurora",
		"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 "lokiii005/Aurora" \
        --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": "lokiii005/Aurora",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Unsloth Studio new

How to use lokiii005/Aurora with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for lokiii005/Aurora to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for lokiii005/Aurora to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for lokiii005/Aurora to start chatting

Load model with FastModel

pip install unsloth
from unsloth import FastModel
model, tokenizer = FastModel.from_pretrained(
    model_name="lokiii005/Aurora",
    max_seq_length=2048,
)

Docker Model Runner
How to use lokiii005/Aurora with Docker Model Runner:
```
docker model run hf.co/lokiii005/Aurora
```

AURORA — rule-governor (LoRA on Gemma 2B)

AURORA is a text-generation policy (Gemma-2B + LoRA) trained to choose ACCEPT or REJECT in response to a text description of a simulated “civilization” state (active rule, resources, stability, belief mismatch, etc.). It is a governance policy for a research simulator, not a general chat model.

Hugging Face model: lokiii005/Aurora
Code & environment: inland-taipen/Aurora

Model details


Developed by	SuperSymmetry
Model type	Causal LM + PEFT/LoRA adapter
Base model	`unsloth/gemma-2-2b-it` (Unsloth + 4-bit training stack)
Language	English (state prompts)
License	Comply with Gemma / Google terms; adapter subject to this repo’s terms (edit if you add a LICENSE)
Finetuned from	`unsloth/gemma-2-2b-it` via Unsloth + TRL SFT

Training

Data: Synthetic rollouts from the in-repo grid environment; random ACCEPT/REJECT exploration, formatted as (state, action, reward) for SFT.
Method: Supervised fine-tuning (SFT) with trl / Unsloth on GPU (e.g. Google Colab T4).
Episodes (example run): ~30 (increase to 50+ to retrain with more data).
See code: training/train_aurora.py for LoRA rank, max_seq_length, and training steps.

Intended use

In scope: With the same AURORA state prompt style as in training, output ACCEPT or REJECT for a proposed rule change in the simulation.
Out of scope: Real-world control, high-stakes decisions, unconstrained open chat. Research / demo only.

Limitations and risks

Trained on synthetic env rollouts, not a broad human dataset; not safety-certified.
Gemma policy limitations apply. Parse ACCEPT/REJECT in code; generate may add extra text.

How to load

Use the base_model_name_or_path in adapter_config.json on the Files tab if it differs. Typical pattern:

from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel
import torch

base_id = "unsloth/gemma-2-2b-it"  # match adapter_config
adapter_id = "lokiii005/Aurora"

tokenizer = AutoTokenizer.from_pretrained(adapter_id, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(
    base_id, torch_dtype=torch.bfloat16, device_map="auto", trust_remote_code=True,
)
model = PeftModel.from_pretrained(model, adapter_id)
# Prompt format: see demo/app.py and training/train_aurora.py

Evaluation

Reproducible scripts: GitHub scripts/, training/train_aurora.py. No fixed leaderboard; report your own numbers if you re-run.

Environmental impact (rough)

Estimate GPU hours × hardware with the ML CO2 impact calculator if needed for a report.

Citation

Cite this model and the AURORA repository.

Contact: anika.24bcs104092sst.scaler.com

Framework note: PEFT; see adapter_config.json in this model repo for exact library versions.

Downloads last month: 30

Model tree for lokiii005/Aurora

Base model

unsloth/gemma-2-2b-it

Adapter

(259)

this model

lokiii005
/

Aurora