Instructions to use efficiencyx/Jun-LoRA-12B-Safetensor with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use efficiencyx/Jun-LoRA-12B-Safetensor with Transformers:

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

pipe = pipeline("text-generation", model="efficiencyx/Jun-LoRA-12B-Safetensor")
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)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("efficiencyx/Jun-LoRA-12B-Safetensor")
model = AutoModelForMultimodalLM.from_pretrained("efficiencyx/Jun-LoRA-12B-Safetensor", device_map="auto")
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=256)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use efficiencyx/Jun-LoRA-12B-Safetensor with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/efficiencyx/Jun-LoRA-12B-Safetensor

SGLang

How to use efficiencyx/Jun-LoRA-12B-Safetensor 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 "efficiencyx/Jun-LoRA-12B-Safetensor" \
    --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": "efficiencyx/Jun-LoRA-12B-Safetensor",
		"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 "efficiencyx/Jun-LoRA-12B-Safetensor" \
        --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": "efficiencyx/Jun-LoRA-12B-Safetensor",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use efficiencyx/Jun-LoRA-12B-Safetensor with Docker Model Runner:
```
docker model run hf.co/efficiencyx/Jun-LoRA-12B-Safetensor
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

Jun-12B Safetensor (v8)

Merged bf16 weights of the v8 Jun LoRA on Gemma 4 12B (QAT), for vLLM and transformers. This is what the Jun OS Colab notebook serves when it runs on a TPU. Same fine-tune as the v8 GGUF release, see that card for what changed in v8.

Model Variants & Repositories

Repository	Format	Description
`efficiencyx/Jun-LoRA-12B-Safetensor`	safetensors, bf16	This repo: v8, full precision
`efficiencyx/Jun-LoRA-12B-Safetensor-INT8`	safetensors, INT8 W8A8	v8 for TPUs with 16 GB (v5e-1)
`efficiencyx/Jun-LoRA-12B-GGUF`	GGUF	v8 for llama.cpp and Ollama
`efficiencyx/Jun-LoRA-12B-Safetensor-v7`	safetensors, bf16	Previous generation (v7)

Merge

The v8 adapter (r 32, rsLoRA, alpha folded to 181.02) merged into the base with PEFT 0.18.1: all 328 attention and MLP projections of the language model. The vision and audio input embedder are untouched, and the chat template is the base model's, byte for byte.

Usage

vllm serve efficiencyx/Jun-LoRA-12B-Safetensor --served-model-name jun --max-model-len 8192

22.3 GiB of weights: a TPU v6e-1 (32 GB) runs it, a v5e-1 (16 GB) needs the INT8 build.

Downloads last month: 56

Safetensors

Model size

12B params

Tensor type

BF16

Model tree for efficiencyx/Jun-LoRA-12B-Safetensor

Base model

google/gemma-4-12B

Finetuned

google/gemma-4-12B-it

Finetuned

unsloth/gemma-4-12B-it-qat-q4_0-unquantized

Adapter

(5)

this model

Quantizations

1 model