Instructions to use tencent/HunyuanOCR with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use tencent/HunyuanOCR with Transformers:

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

pipe = pipeline("image-text-to-text", model="tencent/HunyuanOCR")
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("tencent/HunyuanOCR")
model = AutoModelForMultimodalLM.from_pretrained("tencent/HunyuanOCR", 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 tencent/HunyuanOCR with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "tencent/HunyuanOCR"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "tencent/HunyuanOCR",
		"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/tencent/HunyuanOCR

SGLang

How to use tencent/HunyuanOCR 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 "tencent/HunyuanOCR" \
    --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": "tencent/HunyuanOCR",
		"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 "tencent/HunyuanOCR" \
        --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": "tencent/HunyuanOCR",
		"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 tencent/HunyuanOCR with Docker Model Runner:
```
docker model run hf.co/tencent/HunyuanOCR
```

哥们，这效果也太差了

#34

by AlexLee111 - opened 12 days ago

Discussion

AlexLee111

12 days ago

•

edited 12 days ago

不是哥们，咱能别过拟合测评集吗？

不说别的，就说pdf解析，真实场景上跑你们这个pdf析速度慢的离谱，我一开始还纳闷，拿结果一看一堆复读机样本，效果差的和前几天百度unlimt ocr有的一拼，两个都是复读🐔

那技术报告写的花里胡哨的，你们自己信吗？

调你们这个模型我周末都在加班，调出来这效果，你说叫人蚌埠崩溃？？？？？

麻烦拿点有诚意的东西，不然我以后再也不碰浑圆

VirtualLUO

10 days ago

可以同步一下推理代码，以及部署方式
如果目前使用llamacpp部署，会因为EOS默认对齐120020导致重复预测
推荐使用vllm部署

qifengliang07

9 days ago

确实太拉了，用它官方的推理代码，测一张身份证简单的信息提取，生成的东西乱七八糟。

VirtualLUO

6 days ago

感谢两位反馈，check后发现hf的eos符设置错误，导致模型重复预测幻觉输出，目前hf的model config已经fix
同时transformer input的连接与训练不一致，也会在今日进行仓库更新
欢迎持续反馈

AAOBA

6 days ago

+1，卡了我巨长时间以为自己哪配错了;(

Guenyu

5 days ago

不止，hg的transformer测试代码，processer没有禁用use_fast属性，会导致图片预处理完全错误，导致胡乱输出。

AAOBA

5 days ago

不止，hg的transformer测试代码，processer没有禁用use_fast属性，会导致图片预处理完全错误，导致胡乱输出。

我操哥们，这bug怎么de出来的，总算能用了🤣

psp-dada

Tencent org 3 days ago

•

edited about 1 hour ago

不止，hg的transformer测试代码，processer没有禁用use_fast属性，会导致图片预处理完全错误，导致胡乱输出。

你好，感谢关注，我们也注意到了这一点，也能复现这个问题，可以参考官方 GitHub 仓库里的推理代码，这份是正确的
https://github.com/Tencent-Hunyuan/HunyuanOCR/blob/main/inference/transformers/infer_hf_8gpu_hyocr15.py

对于 fast processer 的问题修复的 pull request 在这： https://github.com/huggingface/transformers/pull/47499
可以本地执行这些修改，尝试一下问题是否修复

snake7gun

Tencent org 3 days ago

请问在upstream的transformer可以正常工作吗？

AlexLee111

about 8 hours ago

请问现在正常了吗？问题已经全部修复了？我直接拉最新的代码就能正常跑了吗？

psp-dada

Tencent org about 6 hours ago

•

edited about 6 hours ago

请问现在正常了吗？问题已经全部修复了？我直接拉最新的代码就能正常跑了吗？

你好，我们近期又进行了一轮测试，也推出了一个同时支持 Transformers、vLLM、DFlash 的推理环境，可以参考以下链接进行配置和使用
https://github.com/Tencent-Hunyuan/HunyuanOCR/blob/main/docs/inference/inference_zh.md
我们推荐使用 vLLM 引擎进行推理
如果机器没有 CUDA13，可以参考下面这个环境单独配置 CUDA12 的 vLLM 推理
https://github.com/Tencent-Hunyuan/HunyuanOCR/blob/main/docs/inference/archive/vLLM_zh.md

在使用前请重新拉取或者更新 tencent/HunyuanOCR 模型，以保证配置文件为最新版

GitHub 仓库中所有公布的环境均经过测试，如果在使用过程中仍发现问题，欢迎进一步讨论

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment