Instructions to use Kaoeiri/Magpantheonse-lark-v4x1.6.2-Cydonia-vXXX-22B-3 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Kaoeiri/Magpantheonse-lark-v4x1.6.2-Cydonia-vXXX-22B-3 with Transformers:

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

pipe = pipeline("text-generation", model="Kaoeiri/Magpantheonse-lark-v4x1.6.2-Cydonia-vXXX-22B-3")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Kaoeiri/Magpantheonse-lark-v4x1.6.2-Cydonia-vXXX-22B-3")
model = AutoModelForCausalLM.from_pretrained("Kaoeiri/Magpantheonse-lark-v4x1.6.2-Cydonia-vXXX-22B-3", 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]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Kaoeiri/Magpantheonse-lark-v4x1.6.2-Cydonia-vXXX-22B-3 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Kaoeiri/Magpantheonse-lark-v4x1.6.2-Cydonia-vXXX-22B-3"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Kaoeiri/Magpantheonse-lark-v4x1.6.2-Cydonia-vXXX-22B-3",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Kaoeiri/Magpantheonse-lark-v4x1.6.2-Cydonia-vXXX-22B-3

SGLang

How to use Kaoeiri/Magpantheonse-lark-v4x1.6.2-Cydonia-vXXX-22B-3 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 "Kaoeiri/Magpantheonse-lark-v4x1.6.2-Cydonia-vXXX-22B-3" \
    --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": "Kaoeiri/Magpantheonse-lark-v4x1.6.2-Cydonia-vXXX-22B-3",
		"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 "Kaoeiri/Magpantheonse-lark-v4x1.6.2-Cydonia-vXXX-22B-3" \
        --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": "Kaoeiri/Magpantheonse-lark-v4x1.6.2-Cydonia-vXXX-22B-3",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Kaoeiri/Magpantheonse-lark-v4x1.6.2-Cydonia-vXXX-22B-3 with Docker Model Runner:
```
docker model run hf.co/Kaoeiri/Magpantheonse-lark-v4x1.6.2-Cydonia-vXXX-22B-3
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

merge

This is a merge of pre-trained language models created using mergekit.

Merge Details

Merge Method

This model was merged using the DARE TIES merge method using anthracite-org/magnum-v4-22b as a base.

Models Merged

The following models were included in the merge:

Configuration

The following YAML configuration was used to produce this model:

models:
  - model: anthracite-org/magnum-v4-22b
    parameters:
      weight: 1.0         # Magnum as the primary writing style model
      density: 0.85      # Slightly lower density to allow for blending with other models
  - model: TheDrummer/Cydonia-22B-v1.3
    parameters:
      weight: 0.3         # Reduced weight for creativity to avoid too much overlap
      density: 0.7       # Reduced density to ensure balance in creativity influence
  - model: TheDrummer/Cydonia-22B-v1.2
    parameters:
      weight: 0.2         # Extra creativity without overwhelming the narrative
      density: 0.65      # Reduced density for further balance
  - model: TheDrummer/Cydonia-22B-v1.1
    parameters:
      weight: 0.25        # Lower weight for accuracy and specific features (evil/trolling)
      density: 0.7       # Moderate density for accuracy retention without interference
  - model: Gryphe/Pantheon-RP-Pure-1.6.2-22b-Small
    parameters:
      weight: 0.35        # Increased weight for storytelling and RP focus
      density: 0.8       # Moderate density to ensure a good storytelling influence
  - model: allura-org/MS-Meadowlark-22B
    parameters:
      weight: 0.3         # Increased weight for creativity and balanced writing
      density: 0.7       # Moderate density to ensure creativity enhances the text
  - model: spow12/ChatWaifu_v2.0_22B
    parameters:
      weight: 0.3         # Balanced weight for anime-style RP and conversational tone
      density: 0.7       # Moderate density to contribute anime style without overshadowing
  - model: Saxo/Linkbricks-Horizon-AI-Japanese-Superb-V1-22B
    parameters:
      weight: 0.3         # Balanced weight for Japanese context handling
      density: 0.75      # Higher density to retain Japanese context without excess influence

merge_method: dare_ties  # Using dare_ties for smoother blending across models
base_model: anthracite-org/magnum-v4-22b
parameters:
  density: 0.85          # General density to ensure an overall balanced mix
  epsilon: 0.1           # Maximum change in drop probabilities (affects model output blending)
  lambda: 1.2            # Scaling factor for final merged deltas
dtype: bfloat16