Instructions to use Casual-Autopsy/G4-MeroMero-31B-StyleSwap with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Casual-Autopsy/G4-MeroMero-31B-StyleSwap with Transformers:

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

pipe = pipeline("text-generation", model="Casual-Autopsy/G4-MeroMero-31B-StyleSwap")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("Casual-Autopsy/G4-MeroMero-31B-StyleSwap")
model = AutoModelForMultimodalLM.from_pretrained("Casual-Autopsy/G4-MeroMero-31B-StyleSwap")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
inputs = processor.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(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Casual-Autopsy/G4-MeroMero-31B-StyleSwap with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Casual-Autopsy/G4-MeroMero-31B-StyleSwap"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Casual-Autopsy/G4-MeroMero-31B-StyleSwap",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Casual-Autopsy/G4-MeroMero-31B-StyleSwap

SGLang

How to use Casual-Autopsy/G4-MeroMero-31B-StyleSwap 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 "Casual-Autopsy/G4-MeroMero-31B-StyleSwap" \
    --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": "Casual-Autopsy/G4-MeroMero-31B-StyleSwap",
		"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 "Casual-Autopsy/G4-MeroMero-31B-StyleSwap" \
        --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": "Casual-Autopsy/G4-MeroMero-31B-StyleSwap",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Casual-Autopsy/G4-MeroMero-31B-StyleSwap with Docker Model Runner:
```
docker model run hf.co/Casual-Autopsy/G4-MeroMero-31B-StyleSwap
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

MeroMero-31B Style Swap

An expirimental tensor swap merge targeting only one tensor: lm_head.weight
The merge consist of two models:

The base: zerofata/G4-MeroMero-31B
The donor: Gryphe/Gemma-4-31B-StyleTune

The theory behind this is that since Gryphe's tune touches what your typical fine-tune doesn't: meging through tensor swapping should be practically loseless.
It's also possible that it would make for good merging fodder post-FTing, but that's a theory for someone more knowledgeable in training to dive into.

If you're interested in Gryphe's tuning method, I'd suggest reading the model card of the tune.

Downloads last month: 19

Safetensors

Model size

33B params

Tensor type

BF16

Model tree for Casual-Autopsy/G4-MeroMero-31B-StyleSwap

Gryphe/Gemma-4-31B-StyleTune

zerofata/G4-MeroMero-31B

Merge model

this model

Quantizations

4 models