Instructions to use OraRL/Video-ORA-4B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use OraRL/Video-ORA-4B with Transformers:

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

pipe = pipeline("image-text-to-text", model="OraRL/Video-ORA-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, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("OraRL/Video-ORA-4B")
model = AutoModelForMultimodalLM.from_pretrained("OraRL/Video-ORA-4B", 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 OraRL/Video-ORA-4B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "OraRL/Video-ORA-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": "OraRL/Video-ORA-4B",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/OraRL/Video-ORA-4B

SGLang

How to use OraRL/Video-ORA-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 "OraRL/Video-ORA-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": "OraRL/Video-ORA-4B",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

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 "OraRL/Video-ORA-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": "OraRL/Video-ORA-4B",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Docker Model Runner
How to use OraRL/Video-ORA-4B with Docker Model Runner:
```
docker model run hf.co/OraRL/Video-ORA-4B
```

Video-ORA-4B

File size: 2,934 Bytes

[build-system]
requires = ["setuptools>=77", "wheel"]
build-backend = "setuptools.build_meta"

[project]
name = "orarl"
version = "0.1.0"
description = "Annotations as rollouts for unified video MLLM reinforcement learning"
readme = "README.md"
requires-python = ">=3.10"
license = "Apache-2.0"
license-files = ["LICENSE", "NOTICE"]
keywords = [
    "reinforcement-learning",
    "video-understanding",
    "multimodal",
    "grpo",
]
authors = [
    { name = "Yunheng Li" },
    { name = "Guohong Mu" },
    { name = "Hao Li" },
    { name = "Shengsheng Qian" },
    { name = "Dingwen Zhang" },
    { name = "Qibin Hou" },
    { name = "Ming-Ming Cheng" },
]
classifiers = [
    "Development Status :: 3 - Alpha",
    "Programming Language :: Python :: 3",
    "Programming Language :: Python :: 3.10",
    "Programming Language :: Python :: 3.11",
    "Programming Language :: Python :: 3.12",
]
dependencies = [
    "numpy",
    "PyYAML",
    "torch",
]

[project.optional-dependencies]
test = ["pytest"]
lint = ["ruff"]
hf = ["huggingface_hub"]

[project.urls]
Homepage = "https://orarl.github.io/"
Documentation = "https://orarl.github.io/"

[project.scripts]
orarl-train = "orarl.cli.train:main"
orarl-eval = "orarl.cli.evaluate:main"
orarl-eval-data = "orarl.cli.eval_data:main"
orarl-prepare = "orarl.cli.prepare:main"

[tool.setuptools.packages.find]
where = ["."]
include = ["orarl*", "verl*"]

[tool.setuptools.data-files]
"share/orarl" = ["environment.yml", "requirements-cu129.txt"]
"share/orarl/configs" = ["configs/*.yaml"]
"share/orarl/docs" = ["docs/*.md"]
"share/orarl/scripts" = ["scripts/*.py", "scripts/*.sh"]
"share/orarl/eval" = ["eval/README.md"]
"share/orarl/eval/task" = ["eval/task/*.py", "eval/task/*.sh"]
"share/orarl/eval/task/mindcube" = ["eval/task/mindcube/*.py"]
"share/orarl/eval/task/mmsi" = ["eval/task/mmsi/*.py", "eval/task/mmsi/*.sh"]
"share/orarl/eval/task/revsi" = ["eval/task/revsi/*.py", "eval/task/revsi/*.sh"]
"share/orarl/eval/task/segmentation" = [
    "eval/task/segmentation/*.py",
    "eval/task/segmentation/*.sh",
]
"share/orarl/eval/task/spatial_grounding" = ["eval/task/spatial_grounding/*.py"]
"share/orarl/eval/task/spatial_temporal_grounding" = [
    "eval/task/spatial_temporal_grounding/*.py",
]
"share/orarl/eval/task/temporal_grounding" = [
    "eval/task/temporal_grounding/*.py",
    "eval/task/temporal_grounding/*.sh",
]
"share/orarl/eval/task/tracking" = ["eval/task/tracking/*.py"]

[tool.pytest.ini_options]
testpaths = ["tests"]
addopts = "-ra"

[tool.ruff]
target-version = "py310"
line-length = 100
# The evaluation runtime, the training runtime, and the checkpoint merger are
# ported verbatim from the stack that produced the published numbers. Several of
# their long lines are prompt literals, so reformatting them would change model
# inputs.
extend-exclude = ["eval/task", "verl", "scripts/model_merger.py"]

[tool.ruff.lint]
select = ["E", "F", "I", "W"]