Instructions to use thesysdev/OUI-1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use thesysdev/OUI-1 with Transformers:

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

pipe = pipeline("text-generation", model="thesysdev/OUI-1")
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("thesysdev/OUI-1")
model = AutoModelForMultimodalLM.from_pretrained("thesysdev/OUI-1", 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=40)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use thesysdev/OUI-1 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "thesysdev/OUI-1"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "thesysdev/OUI-1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/thesysdev/OUI-1

SGLang

How to use thesysdev/OUI-1 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 "thesysdev/OUI-1" \
    --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": "thesysdev/OUI-1",
		"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 "thesysdev/OUI-1" \
        --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": "thesysdev/OUI-1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use thesysdev/OUI-1 with Docker Model Runner:
```
docker model run hf.co/thesysdev/OUI-1
```

OUI-1 / adapter /README.md

andhakanoon

adapter card: license is Apache 2.0, same as the Gemma 4 base

68e8006 verified 7 days ago

preview code

Raw

History Blame Contribute Delete

1.09 kB

metadata

base_model: unsloth/diffusiongemma-26B-A4B-it
library_name: peft
license: apache-2.0
tags:
  - lora
  - diffusion-language-model
  - generative-ui
  - openui-lang

OUI-1 LoRA adapter

The PEFT adapter that produced the merged weights in the parent repo, thesysdev/OUI-1. Rank 64, alpha 128, 205 target modules (the decoder's q, k, v, o, gate, up and down projections), trained on unsloth/diffusiongemma-26B-A4B-it, a mirror of google/diffusiongemma-26B-A4B-it.

It is a tied LoRA: DiffusionGemma's encoder and decoder share weight storage, so merging the decoder deltas adapts both passes. PeftModel.from_pretrained(base, adapter) on its own wraps the decoder only and gives a different model; call .merge_and_unload() to reproduce the merged weights, or use the merged checkpoint in the parent repo, which is what every published number was measured on.

Published by Thesys under Apache 2.0, the same license as the base model (Gemma 4 license: https://ai.google.dev/gemma/docs/gemma_4_license). See the parent repo's model card for results, serving settings and usage.