Image-Text-to-Text
Transformers
Safetensors
qwen3_5
vllm
video
multimodal
reinforcement-learning
temporal-grounding
object-tracking
video-segmentation
visual-question-answering
spatial-reasoning
qwen3.5
conversational
Instructions to use OraRL/Video-ORA-9B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use OraRL/Video-ORA-9B with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("image-text-to-text", model="OraRL/Video-ORA-9B") 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-9B") model = AutoModelForMultimodalLM.from_pretrained("OraRL/Video-ORA-9B", 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-9B with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "OraRL/Video-ORA-9B" # 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-9B", "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-9B
- SGLang
How to use OraRL/Video-ORA-9B 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-9B" \ --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-9B", "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-9B" \ --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-9B", "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-9B with Docker Model Runner:
docker model run hf.co/OraRL/Video-ORA-9B
File size: 3,586 Bytes
53c10a4 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 | <p align="right"><a href="evaluation.md">English</a></p>
# 评测 Video-ORA
OraRL 使用七个发布任务族的规范配置评测任务原生答案,对应运行时位于
`eval/task/`。运行前请先
[验证评测节点](environment_zh.md#验证仅用于评测的节点)。
## 1. 准备模型和数据
安装 Hugging Face CLI 并下载发布文件:
```bash
python -m pip install -U huggingface_hub
hf download OraRL/Video-ORA-9B \
--local-dir "$PWD/models/Video-ORA-9B"
hf download OraRL/OraRL-Data \
--repo-type dataset \
--include "OraRL-eval-data/**" \
--local-dir "$PWD/OraRL-Data"
```
完整评测数据体积较大,下载支持断点续传;
`OraRL-eval-data/assets.jsonl` 是权威文件清单。
本地数据结构为:
```text
OraRL-Data/OraRL-eval-data/
├── datasets.jsonl # benchmark、prompt、parser、metric 与采样配置
├── assets.jsonl # 发布文件清单
├── annotations/ # 规范化 JSONL 标注
└── media/ # 原始图像、视频和字幕
```
发布数据有意排除衍生预处理缓存;不存在兼容缓存时,评测器会直接解码清单中声明的
原始媒体。
## 2. 运行论文规范配置
先预览解析后的命令:
```bash
orarl-eval \
--model "$PWD/models/Video-ORA-9B" \
--tasks paper \
--dataset "$PWD/OraRL-Data/OraRL-eval-data" \
--summary "$PWD/outputs/Video-ORA-9B/evaluation.json"
```
`orarl-eval` 默认为 dry run。检查模型、数据、评测器、任务配置和输出路径后,
增加 `--run`。
可先运行限制样本数的 smoke test:
```bash
orarl-eval \
--model "$PWD/models/Video-ORA-9B" \
--tasks videomme \
--dataset "$PWD/OraRL-Data/OraRL-eval-data" \
--max-samples 8 \
--summary "$PWD/outputs/Video-ORA-9B/videomme-smoke.json" \
--run
```
Smoke test 分数仅用于验证流程,不应作为 benchmark 结果报告。
## 3. 组合任务集
`--tasks paper` 选择所有发布任务,`--tasks video_qa` 选择 7 个视频问答
benchmark;多个单项任务可用逗号分隔。
| 任务族 | 任务名 |
| --- | --- |
| 视频问答 | `videomme`, `videommev2`, `mvbench`, `mmvu`, `videoholmes`, `longvideobench`, `mlvu` |
| 空间智能 | `vsi`, `mmsi`, `mindcube`, `revsi` |
| 时序定位 | `temporal_grounding` |
| 空间定位 | `spatial_grounding` |
| 跟踪 | `tracking` |
| 时空定位 | `stvg` |
| 分割 | `segmentation` |
`datasets.jsonl` 中的规范配置固定了帧采样、分辨率、prompt、parser 和 metric;
修改这些值即代表不同的评测设置。ReVSI 单独报告,不计入三个 benchmark 的空间
智能平均分。
## 4. 添加分割后处理
分割推理默认不执行 SAM2 后处理。启用 `--segmentation-run-sam2` 时还需要:
- SAM2 权重(`SEGMENTATION_SAM2_CKPT`)
- 匹配的 Hydra 配置(`SEGMENTATION_SAM2_CFG`)
- OneThinker 官方 `seg_post_sam2.py`
(`SEGMENTATION_POSTPROCESSOR_PATH`)
- `sam2` Python 包
OraRL 会在启动前检查三个路径。
## 5. 保存可报告的输出
每个任务评测器先写入原生 summary,随后 `orarl-eval` 生成指定的聚合 JSON,
其中包含请求、完成和缺失的任务、返回码以及官方 metric。输出按论文任务族组织在
`outputs/<model>/` 下。
正式报告结果时请保留:
- OraRL 源码版本
- 模型和数据版本
- 完整命令与聚合 summary
- 软件版本和加速卡类型/数量
- 所有非默认配置或命令行覆盖项
评测器映射、视频解码后端和 checkpoint 格式说明见
[`../eval/README.md`](../eval/README.md)。
|