Instructions to use Vortex5/Abyssal-Seraph-12B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Vortex5/Abyssal-Seraph-12B with Transformers:

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

pipe = pipeline("text-generation", model="Vortex5/Abyssal-Seraph-12B")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Vortex5/Abyssal-Seraph-12B")
model = AutoModelForCausalLM.from_pretrained("Vortex5/Abyssal-Seraph-12B", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Vortex5/Abyssal-Seraph-12B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Vortex5/Abyssal-Seraph-12B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Vortex5/Abyssal-Seraph-12B",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/Vortex5/Abyssal-Seraph-12B

SGLang

How to use Vortex5/Abyssal-Seraph-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 "Vortex5/Abyssal-Seraph-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": "Vortex5/Abyssal-Seraph-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 "Vortex5/Abyssal-Seraph-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": "Vortex5/Abyssal-Seraph-12B",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use Vortex5/Abyssal-Seraph-12B with Docker Model Runner:
```
docker model run hf.co/Vortex5/Abyssal-Seraph-12B
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

🌌 Abyssal-Seraph-12B

Where the light of the divine meets the poetry of the abyss.

🜂 Overview

Abyssal-Seraph-12B is a multi-stage creative merge designed for expressive storytelling, emotional depth, and lyrical dialogue.
It was crafted through a layered fusion using MergeKit:

🌙 LunaMaid × Vermilion-Sage — merged via NearSwap (t=0.0008) to unify LunaMaid’s balanced composure with Vermilion-Sage’s radiant prose.
🕯️ Dark-Quill × Mag-Mell-R1 — merged via NearSwap (t=0.0008) to draw forth mysticism, poetic darkness, and a sense of dreamlike gravity.
✨ Both intermediate results combined with the Karcher Mean — a geometric blend ensuring harmony between light and shadow.

🩶 Model Essence

Trait	Description
🧠 Core Nature	Philosophical, poetic, emotionally resonant
💬 Style	Fluid prose, vivid imagery, articulate reflection
💫 Tone	Dreamlike, balanced between divine warmth and abyssal calm
🎭 Best For	Roleplay, character dialogue, introspection, lore writing, creative prose

🧬 Merge Overview

Abyssal-Seraph-12B was created through a multi-stage, precision merge designed to blend expressive prose with poetic balance while maintaining model stability.

🌙 Stage 1

✨ Method: NearSwap (t = 0.0008)
🩵 Base: Vortex5/LunaMaid-12B
💮 Secondary: Vortex5/Vermilion-Sage-12B

Stage 1 Configuration

name: First
models:
- model: Vortex5/Vermilion-Sage-12B
merge_method: nearswap
base_model: Vortex5/LunaMaid-12B
parameters:
t: 0.0008
dtype: bfloat16
tokenizer:
source: Vortex5/LunaMaid-12B

🩶 Stage 2

⚙️ Method: NearSwap (t = 0.0008) 🖤 Base: Vortex5/Dark-Quill-12B 💫 Secondary: inflatebot/MN-12B-Mag-Mell-R1

Stage 2 Configuration

name: Second
models:
  - model: inflatebot/MN-12B-Mag-Mell-R1
merge_method: nearswap
base_model: Vortex5/Dark-Quill-12B
parameters:
  t: 0.0008
dtype: bfloat16`

🌌 Stage 3 — Final Merge

⚖️ Method: Karcher Mean (tol = 1e-9, max_iter = 20000) 🜂 Inputs: First + Second 💎 Purpose: To geometrically fuse both for coherence.

Final Merge Configuration

models:
- model: First
- model: Second
merge_method: karcher
dtype: bfloat16
parameters:
tol: 1e-9
max_iter: 20000
tokenizer:
source: First

🌑🜂 Acknowledgements 🜂🌑

⚙️ mradermacher — for static and imatrix quantization
🜛 DeathGodlike — for EXL3 quants
🩶 All original model authors and contributors whose work made this model possible.

Models merged in this creation:

Downloads last month: 12

Safetensors

Model size

12B params

Tensor type

BF16

Model tree for Vortex5/Abyssal-Seraph-12B

Vortex5/Dark-Quill-12B

Vortex5/LunaMaid-12B

Vortex5/Vermilion-Sage-12B

inflatebot/MN-12B-Mag-Mell-R1

Merge model

this model

Merges

7 models

Quantizations

4 models