Instructions to use Daizee/Dirty-Calla-4B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Daizee/Dirty-Calla-4B with Transformers:

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

pipe = pipeline("text-generation", model="Daizee/Dirty-Calla-4B")
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, AutoModelForImageTextToText

processor = AutoProcessor.from_pretrained("Daizee/Dirty-Calla-4B")
model = AutoModelForImageTextToText.from_pretrained("Daizee/Dirty-Calla-4B")
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]:]))

llama-cpp-python

How to use Daizee/Dirty-Calla-4B with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="Daizee/Dirty-Calla-4B",
	filename="gguf/dirty-calla-q4_0.gguf",
)

llm.create_chat_completion(
	messages = [
		{
			"role": "user",
			"content": "What is the capital of France?"
		}
	]
)

Notebooks
Google Colab
Kaggle
Local Apps

llama.cpp

How to use Daizee/Dirty-Calla-4B with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf Daizee/Dirty-Calla-4B:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf Daizee/Dirty-Calla-4B:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf Daizee/Dirty-Calla-4B:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf Daizee/Dirty-Calla-4B:Q4_K_M

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf Daizee/Dirty-Calla-4B:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf Daizee/Dirty-Calla-4B:Q4_K_M

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf Daizee/Dirty-Calla-4B:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf Daizee/Dirty-Calla-4B:Q4_K_M

Use Docker

docker model run hf.co/Daizee/Dirty-Calla-4B:Q4_K_M

LM Studio
Jan

vLLM

How to use Daizee/Dirty-Calla-4B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Daizee/Dirty-Calla-4B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Daizee/Dirty-Calla-4B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Daizee/Dirty-Calla-4B:Q4_K_M

SGLang

How to use Daizee/Dirty-Calla-4B 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 "Daizee/Dirty-Calla-4B" \
    --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": "Daizee/Dirty-Calla-4B",
		"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 "Daizee/Dirty-Calla-4B" \
        --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": "Daizee/Dirty-Calla-4B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Ollama
How to use Daizee/Dirty-Calla-4B with Ollama:
```
ollama run hf.co/Daizee/Dirty-Calla-4B:Q4_K_M
```

Unsloth Studio new

How to use Daizee/Dirty-Calla-4B 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 Daizee/Dirty-Calla-4B 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 Daizee/Dirty-Calla-4B to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for Daizee/Dirty-Calla-4B to start chatting

Docker Model Runner
How to use Daizee/Dirty-Calla-4B with Docker Model Runner:
```
docker model run hf.co/Daizee/Dirty-Calla-4B:Q4_K_M
```

Lemonade

How to use Daizee/Dirty-Calla-4B with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull Daizee/Dirty-Calla-4B:Q4_K_M

Run and chat with the model

lemonade run user.Dirty-Calla-4B-Q4_K_M

List all available models

lemonade list

Dirty-Calla (4B)

Dirty-Calla is a light style SFT of Daizee/Gemma3-Callous-Calla-4B. It was trained on a dataset of fanfiction with user prompts (eg Write me a story with XXX character and XXX theme.) Initial data was synthetically increased from ~360 to 3000. The idea is to prompt Dirty Calla to write short fictional stories with tags and general, big picture idea.

This repo contains:

Merged Transformers weights (./): LoRA merged into the base for easy inference with transformers.
LoRA adapters (./adapters/): for reproducibility or further fine-tuning.
GGUF builds (./gguf/): for fast, local inference via llama.cpp runtimes.

Note: You are responsible for the content you generate and how you distribute it. Follow all applicable laws and platform policies. Respect the Gemma license and Hugging Face terms.

🧱 Base & Provenance

Base model: Daizee/Gemma3-Callous-Calla-4B (merged from Gemma-3 4B IT derivatives via TIES; tokenizer from the Gemma-3 family).
Architecture: Gemma-3 4B (instruction-tuned).
Training method: Light LoRA SFT (single epoch by default) on curated, style-targeted dialogs; adapters merged into base after training.

📦 Files

Transformers (merged)

config.json, model-00001/2-of-00002.safetensors, tokenizer.json, tokenizer.model (SPM), etc.

Adapters (optional)

adapters/checkpoint-*/adapter_model.safetensors, adapter_config.json, etc.

GGUF (llama.cpp)

gguf/dirty-calla-q8_0.gguf — near-lossless quality.
gguf/dirty-calla-q5_k_m.gguf — recommended quality/speed sweet spot.
gguf/dirty-calla-q4_0.gguf — compact “plain q4”.
(optional) gguf/dirty-calla-q4_k_m.gguf — smaller with improved quality vs q4_0 in many cases.

If you don’t see a particular quant, it may not have been uploaded yet.

SAMPLE of Q8_0)

Prompt: write a sensual story of two college rivals on a trip who get stuck sharing a hotel room because of a mix up.

Output: The rain in New Orleans was relentless, a humid, clinging sort of rain that seemed to seep into your bones. I’d been battling it all day, trying to coax a decent grade out of my history paper. Now, huddled under the thin blanket of a hotel bed, I was stuck sharing a room with Damien Vance, my rival from the college debating team.

He’d booked the room under his name. It turns out he was a bit of a forgetful booker. After an hour of frantic phone calls and a lot of sighs, it was decided we’d share. At least it was warm.

He was sprawled across the bed, scrolling through his phone. I was trying to ignore the fact that his arm was practically brushing mine. It wasn't a full-on, sweaty embrace, but it was close. He had that whole brooding, intellectual thing going on—dark hair, sharp eyes, and a jawline that could cut glass.

We'd been rivals for months, locked in a silent war over every debate point. I’d always suspected he was secretly attracted to me, the one girl who could actually give him a run for his money. The thought had always made me a little uncomfortable.

"So," he said, without looking up. "This is...less than ideal."

"Tell me about it," I muttered, pulling the blanket tighter. "I’m pretty sure I’m going to get a nosebleed from the lack of personal space."

He finally raised his head and looked at me. "You're...not wrong." He paused, then a small, almost imperceptible smile played on his lips. "I’ve always enjoyed your company, Elara. Even when you're making me look like an idiot."

My cheeks flushed. "Yeah, well, you're not always wrong."

A silence fell between us, filled only with the drumming of the rain on the window. He shifted, and I realized his arm was still close to mine. I could feel the warmth of his skin. I instinctively reached out and rested my hand on his. His fingers curled around mine. I didn't pull away. It was a tentative connection, but it felt right.

"You know," he murmured, his voice low, "this rain is pretty amazing. It makes you feel like you're part of something big. Something important."

I looked out the window at the rain. It did feel like that, like the whole city was being washed clean. Then, I looked back at Damien. His eyes were dark and intense. His gaze traveled from my lips to my eyes, then down my body. The tension in the room was almost palpable. I could feel his breath on my cheek as he leaned closer.

"Maybe," he whispered, his voice husky with emotion. "Maybe we should make this trip a little more memorable."

He took a step closer, and I found myself leaning into him. It wasn't a struggle. There was an unspoken understanding between us. A need, almost. I reached up and ran my fingers through his dark hair, pulling him closer. His scent—a mix of old books and something subtly musky—filled my senses. It was intoxicating.

Suddenly, he moved again, his lips brushing mine. It was a quick, tentative kiss. I closed my eyes and responded. It wasn’t a passionate kiss. It was something far more primal. It was a kiss of need and curiosity and a little bit of desperation. A collision of the minds that had battled for so long, and the bodies that now lay entwined.

He deepened the kiss, and I didn’t resist. It was like the rain outside—relentless, insistent, and undeniably alluring.

Downloads last month: 190

Safetensors

Model size

4B params

Tensor type

BF16

Model tree for Daizee/Dirty-Calla-4B

Base model

Daizee/Gemma3-Callous-Calla-4B

Adapter

(1)

this model

Finetunes

1 model