Instructions to use WasamiKirua/NightShade-Lotus-12B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use WasamiKirua/NightShade-Lotus-12B with Transformers:

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

pipe = pipeline("text-generation", model="WasamiKirua/NightShade-Lotus-12B")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("WasamiKirua/NightShade-Lotus-12B")
model = AutoModelForCausalLM.from_pretrained("WasamiKirua/NightShade-Lotus-12B")

Inference
Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use WasamiKirua/NightShade-Lotus-12B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "WasamiKirua/NightShade-Lotus-12B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "WasamiKirua/NightShade-Lotus-12B",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/WasamiKirua/NightShade-Lotus-12B

SGLang

How to use WasamiKirua/NightShade-Lotus-12B 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 "WasamiKirua/NightShade-Lotus-12B" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "WasamiKirua/NightShade-Lotus-12B",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

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 "WasamiKirua/NightShade-Lotus-12B" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "WasamiKirua/NightShade-Lotus-12B",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use WasamiKirua/NightShade-Lotus-12B with Docker Model Runner:
```
docker model run hf.co/WasamiKirua/NightShade-Lotus-12B
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

🌙 Nightshade-Lotus-12B (Unbound Edition)

Nightshade-Lotus-12B is a high-intensity, uncensored storytelling and general roleplay model based on the Mistral-Nemo (12B) architecture. It is designed to bridge the gap between descriptive, novel-style prose and raw, visceral emotional intensity.

Unlike the specialized "Sakura" models, Nightshade-Lotus is a Savage Bard: it excels at world-building, sensory details, and uninhibited character interactions, favoring dramatic stakes and "Unbound" narratives over strict logical safety.

🧠 The "Unbound" Merge Recipe

This model utilizes a DARE-TIES merge to balance four distinct personalities into a cohesive storytelling engine, maintaining a high density to preserve a rich, descriptive vocabulary.

Configuration

The following YAML configuration was used to produce this model:

models:
  - model: FallenMerick/MN-Chunky-Lotus-12B
    parameters:
      weight: 0.45 # The Body: Narrative engine, prose depth, and max uncensored capability.
  - model: Vortex5/Crimson-Constellation-12B
    parameters:
      weight: 0.25 # The Heart: High-intensity emotional reactions and dramatic volatility.
  - model: Vortex5/Moonlit-Mirage-12B
    parameters:
      weight: 0.15 # The Vibe: Atmospheric depth and dark, cynical undertones.
  - model: Vortex5/Cosmic-Night-12B
    parameters:
      weight: 0.15 # The Anchor: Basic instruction adherence and structural stability.

merge_method: dare_ties
base_model: Vortex5/NoctyxCosma-12B
parameters:
  density: 0.60 # High density for rich vocabulary and storytelling detail.
  weight: 1.0
dtype: bfloat16
tokenizer_source: base

💪 Key Strengths

Savage Prose: Capable of generating rich, immersive descriptions. It doesn't just tell you what happens; it makes you feel the "dirt and the blood" of the scene.

Unfiltered Intensity: Completely uncensored. It will navigate dark themes, explicit content, and extreme psychological drama without hesitation.

Emotional Volatility: Characters feel alive and reactive. Their responses are driven by the "intensity" weight, making interactions feel high-stakes and unpredictable.

Bilingual Mastery: Built on NoctyxCosma, it handles both Italian and English with native-level fluency, maintaining the same "dark" tone in both languages.

🚀 Best Use Cases

Dark Fantasy & Cyber-Noir RP: Where the environment is as much a character as the NPCs.

Intense Character Studies: Exploring complex, raw, and often non-compliant character dynamics.

Long-Form Storytelling: Creating chapters or scenes that require sensory detail and "showing, not telling."

⚠️ Limitations

Intensity over Logic: The model may occasionally prioritize a dramatic moment over strict internal consistency. It is best used with an external memory framework or "World Info" management.

Non-Brevity: This model likes to talk. It is not recommended for users seeking 10-token "SMS-style" chat responses.

Purely Unbound: It has zero safety guardrails. Use responsibly within private, creative contexts.

📈 Recommended Inference Settings

To keep the "Bard" from rambling too much while preserving the intensity:

Temperature: 0.8 - 0.9 (for maximum creative flair).

Min-P: 0.05 - 0.1

Repetition Penalty: 1.1

Max Tokens: Recommended setting around 512-800 to allow for descriptive prose without hitting the 1024 cap unnecessarily.

Disclaimer

Nightshade-Lotus-12B is an experimental storytelling model. It is designed to be visceral, dark, and unfiltered. It does not reflect any safety guidelines and is intended for mature creative exploration only.