Instructions to use soob3123/GrayLine-Gemma3-12B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use soob3123/GrayLine-Gemma3-12B with Transformers:

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

pipe = pipeline("text-generation", model="soob3123/GrayLine-Gemma3-12B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("soob3123/GrayLine-Gemma3-12B")
model = AutoModelForCausalLM.from_pretrained("soob3123/GrayLine-Gemma3-12B")
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

vLLM

How to use soob3123/GrayLine-Gemma3-12B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "soob3123/GrayLine-Gemma3-12B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "soob3123/GrayLine-Gemma3-12B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/soob3123/GrayLine-Gemma3-12B

SGLang

How to use soob3123/GrayLine-Gemma3-12B 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 "soob3123/GrayLine-Gemma3-12B" \
    --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": "soob3123/GrayLine-Gemma3-12B",
		"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 "soob3123/GrayLine-Gemma3-12B" \
        --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": "soob3123/GrayLine-Gemma3-12B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use soob3123/GrayLine-Gemma3-12B with Docker Model Runner:
```
docker model run hf.co/soob3123/GrayLine-Gemma3-12B
```

Not accurate and doesn't check answers or logic.

by Ambiguous666 - opened May 9, 2025

Discussion

Ambiguous666

May 9, 2025

•

edited May 9, 2025

A quick use test showed lack of accuracy or reason or checking wrong answers like Gemma3. Amoral model doesn't have same bahaviour.

soob3123

Owner May 9, 2025

Can you give me an example?

Ambiguous666

May 9, 2025

•

edited May 9, 2025

Hope image works. Doesn't seem major, but I've found when a model behaves like this - I can't work with it, kind of fundamental issue.

soob3123

Owner May 9, 2025

are you comparing the qwen3 to gemma3?

this is still useful information, but just wondering whats the rationale here.

soob3123

Owner May 9, 2025

•

edited May 9, 2025

qwen3-amoral even got thinking on and its in q6(quant might not much of a issue tho)
Have you compared this to amoral-gemma3? do you get the same issue?

soob3123 changed discussion status to closed May 9, 2025

soob3123 changed discussion status to open May 9, 2025

Ambiguous666

May 9, 2025

•

edited May 9, 2025

Fair enough. (openwebui says "thinking" even if it it's set to disable")
Here's a comparison with the amoral-gemma3-12B-v2.Q4_K_M

Seems the Gray doesn't have the typical Gemma "You're right to question my previous response." behaviour.

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment