nsfw

Instructions to use AIgotahole/Gewwa-2-9B-wtf with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use AIgotahole/Gewwa-2-9B-wtf with Transformers:

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

pipe = pipeline("text-generation", model="AIgotahole/Gewwa-2-9B-wtf")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForMultimodalLM

tokenizer = AutoTokenizer.from_pretrained("AIgotahole/Gewwa-2-9B-wtf")
model = AutoModelForMultimodalLM.from_pretrained("AIgotahole/Gewwa-2-9B-wtf")
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 AIgotahole/Gewwa-2-9B-wtf with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "AIgotahole/Gewwa-2-9B-wtf"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "AIgotahole/Gewwa-2-9B-wtf",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/AIgotahole/Gewwa-2-9B-wtf

SGLang

How to use AIgotahole/Gewwa-2-9B-wtf 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 "AIgotahole/Gewwa-2-9B-wtf" \
    --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": "AIgotahole/Gewwa-2-9B-wtf",
		"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 "AIgotahole/Gewwa-2-9B-wtf" \
        --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": "AIgotahole/Gewwa-2-9B-wtf",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use AIgotahole/Gewwa-2-9B-wtf with Docker Model Runner:
```
docker model run hf.co/AIgotahole/Gewwa-2-9B-wtf
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

For RP & story gen, a nice fine-tuning of Gemma-2-9B could surprise you with some highly creative and authentic expressions way surpassing its size, which Gemma-3-12B even got no match. Yet the glitches are obvious too, and hard to ignore. As it's like breaking a perfect sentence with one word so weird that it may totally come from another language... Among tons of works trying to stabilize the bitch, I enjoy grimjim/Magnolia-v3-Gemma2-8k-9B the most. So I picked the rich recoilme/recoilme-gemma-2-9B-v0.2 plus the strong lemon07r/Gemma-2-Ataraxy-v4d-9B to tame it with one last merge. And failed again... It's just slightly smarter, more sensitive to NSFW directions with a little rebellious tendency. So keep retrying and editing. It's 9B, after all.
"It feels few steps to perfection, 'cause it's google."

For RP & story gen,
a nice fine-tuning of Gemma-2-9B could surprise you with some highly creative and authentic expressions way surpassing its size, which Gemma-3-12B even got no match.
Yet the glitches are obvious too, and hard to ignore.
As it's like breaking a perfect sentence with one word so weird
that it may totally come from another language...

Among tons of works trying to stabilize the bitch,
I enjoy grimjim/Magnolia-v3-Gemma2-8k-9B the most.
So I picked the rich recoilme/recoilme-gemma-2-9B-v0.2 plus the strong lemon07r/Gemma-2-Ataraxy-v4d-9B to tame it with one last merge.
And failed again...

It's just slightly smarter, more sensitive to NSFW directions with a little rebellious tendency.
So keep retrying and editing.
It's 9B, after all.

"It feels few steps to perfection, 'cause it's google."

models:
  - model: grimjim/Magnolia-v3-Gemma2-8k-9B
  - model: recoilme/recoilme-gemma-2-9B-v0.2
    parameters:
      density: [0.5, 0.7, 0.6, 0.7, 0.5]
      epsilon: [0.05, 0.07, 0.06, 0.07, 0.05]
      weight: [-0.01150, 0.01793, -0.01034, 0.01855, -0.01876]
  - model: lemon07r/Gemma-2-Ataraxy-v4d-9B
    parameters:
      density: [0.5, 0.3, 0.4, 0.3, 0.5]
      epsilon: [0.05, 0.03, 0.04, 0.03, 0.05]
      weight: [0.01763, -0.01992, 0.01975, -0.01096, 0.01951]
merge_method: della
base_model: grimjim/Magnolia-v3-Gemma2-8k-9B
parameters:
    normalize: false
    lambda: 0.66
tokenizer_source: base
dtype: bfloat16