Image-Text-to-Text
Transformers
Safetensors
Chinese
English
paddleocr_vl
table-recognition
paddleocr-vl
verl
sft
dapo
conversational
custom_code
Instructions to use CheneyWong/paddleocr-vl16-sft800-rl-repro with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use CheneyWong/paddleocr-vl16-sft800-rl-repro with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("image-text-to-text", model="CheneyWong/paddleocr-vl16-sft800-rl-repro", trust_remote_code=True) 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("CheneyWong/paddleocr-vl16-sft800-rl-repro", trust_remote_code=True) model = AutoModelForMultimodalLM.from_pretrained("CheneyWong/paddleocr-vl16-sft800-rl-repro", trust_remote_code=True, 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 CheneyWong/paddleocr-vl16-sft800-rl-repro with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "CheneyWong/paddleocr-vl16-sft800-rl-repro" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "CheneyWong/paddleocr-vl16-sft800-rl-repro", "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/CheneyWong/paddleocr-vl16-sft800-rl-repro
- SGLang
How to use CheneyWong/paddleocr-vl16-sft800-rl-repro 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 "CheneyWong/paddleocr-vl16-sft800-rl-repro" \ --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": "CheneyWong/paddleocr-vl16-sft800-rl-repro", "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 "CheneyWong/paddleocr-vl16-sft800-rl-repro" \ --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": "CheneyWong/paddleocr-vl16-sft800-rl-repro", "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 CheneyWong/paddleocr-vl16-sft800-rl-repro with Docker Model Runner:
docker model run hf.co/CheneyWong/paddleocr-vl16-sft800-rl-repro
PaddleOCR-VL-1.6 SFT800:权重与实验复现包
这是 CheneyWong 的表格识别实验快照,整理日期 2026-11-16 UTC。根目录为已选定的 SFT step800 完整 HF 权重;不是本轮 RL 的最终模型。仓库按所有者要求公开,便于另一台机器继续迭代。
**第一次使用请看 迁移与环境安装**,再看 前情提要、reward 与 VeRL 参数、crop/fullpage 评测、Codex 与 GPT 配置。CPU 验证记录说明已检查的范围。
| 内容 | 位置 |
|---|---|
| SFT800 权重、Tokenizer、Processor | 仓库根目录 |
| 当前 VeRL 运行源码快照 | repro/vendor/verl/ |
| Paddle 适配、reward、DAPO / 类 DAPO 训练配置 | repro/workspace/train/ |
| 修订后的评测脚本、公式归一化、布局缓存及 GT 覆盖规则 | repro/workspace/eval/ |
| 原始 SFT 数据,16,091 条 | repro/workspace/train/selected-formula-v3/sft-train.parquet |
| 当前两种 RL 共用数据,2,499 条 | repro/workspace/train/adaptive-rl-v2/train.parquet |
| 独立 heldout,82 条 | repro/workspace/train/opd-rl-pilot/data/val.parquet |
| crop 评测,173,141 条 / 32 shards | repro/workspace/datasets/table-recognition-eval-173k/ |
| fullpage 评测,2,864 页 | repro/workspace/datasets/table-recognition-fullpage-eval/ |
| teacher 专项评测原始集合、排除规则、参考分数 | repro/workspace/eval/teacher-benchmark-v5/ |
| PP-DocLayoutV3 权重 | assets/PP-DocLayoutV3/ |
| 环境记录、逐文件校验 | environment/、MANIFEST.json |
下载
python -m pip install huggingface_hub
hf download CheneyWong/paddleocr-vl16-sft800-rl-repro \
--local-dir /workspace/paddle-sft800-repro --max-workers 16
python /workspace/paddle-sft800-repro/scripts/verify_bundle.py
完整快照约 16.7 GB(十进制);不重复存放由原评测集派生的短/长 crop 分片。只需要权重时:
hf download CheneyWong/paddleocr-vl16-sft800-rl-repro \
--local-dir /workspace/sft800 \
--include '*.safetensors' --include '*.json' --include '*.model' \
--include '*.jinja' --include '*.py' \
--exclude 'repro/*' --exclude 'assets/*' --exclude 'scripts/*' \
--exclude 'environment/*' --exclude 'examples/*'
新机器采用与原实验相同的 /workspace/train、/workspace/eval、/workspace/datasets 和 /venv/main 路径。安装脚本遇到已有目录会拒绝覆盖;不要直接在正在训练的机器执行迁移安装。
边界
- 本包保存 SFT800 的推理/继续微调权重,不包含其 step800 优化器、调度器和 RNG;从它开始新 RL 可复现初始化,不能据此逐位恢复当时 SFT 的中间训练状态。
- 运行代码包含本项目的 Paddle/VeRL 适配和 checkpoint 空间保护。当前安装环境存在上游依赖声明冲突,安装说明明确记录,不能直接用最新版包替换。
- 部分训练与专项评测标签来自原数据或机器伪标签,并非全部人工金标。比较模型时必须冻结同一版图片、GT、备用 GT 和布局修正。
licenses/及 vendored 源码中保留原许可;上游模型与各数据源仍遵循各自使用条款。本仓库未对全部混合数据重新授予统一许可证。- 不含 API/HF 密钥、登录文件、原始 Codex 会话、Qwen 大模型权重及当前 RL checkpoint。生成脚本的示例不会自动调用付费 API。
原模型:PaddlePaddle/PaddleOCR-VL-1.6。实际文件范围以 MANIFEST.json 为准,远程上传完成情况见 UPLOAD_STATUS.json。
- Downloads last month
- 27
Model tree for CheneyWong/paddleocr-vl16-sft800-rl-repro
Base model
PaddlePaddle/PaddleOCR-VL-1.6