README.md

---
pipeline_tag: image-text-to-text
language:
- multilingual
tags:
- deepseek
- vision-language
- ocr
- custom_code
license: mit
---
<div align="center">
  <img src="https://github.com/deepseek-ai/DeepSeek-V2/blob/main/figures/logo.svg?raw=true" width="60%" alt="DeepSeek AI" />
</div>
<hr>
<div align="center">
  <a href="https://www.deepseek.com/" target="_blank">
    <img alt="Homepage" src="https://github.com/deepseek-ai/DeepSeek-V2/blob/main/figures/badge.svg?raw=true" />
  </a>
  <a href="https://huggingface.co/deepseek-ai/DeepSeek-OCR" target="_blank">
    <img alt="Hugging Face" src="https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-DeepSeek%20AI-ffc107?color=ffc107&logoColor=white" />
  </a>

</div>

<div align="center">

  <a href="https://discord.gg/Tc7c45Zzu5" target="_blank">
    <img alt="Discord" src="https://img.shields.io/badge/Discord-DeepSeek%20AI-7289da?logo=discord&logoColor=white&color=7289da" />
  </a>
  <a href="https://twitter.com/deepseek_ai" target="_blank">
    <img alt="Twitter Follow" src="https://img.shields.io/badge/Twitter-deepseek_ai-white?logo=x&logoColor=white" />
  </a>

</div>



<p align="center">
  <a href="https://github.com/deepseek-ai/DeepSeek-OCR"><b>🌟 Github</b></a> |
  <a href="https://huggingface.co/deepseek-ai/DeepSeek-OCR"><b>📥 Model Download</b></a> |
  <a href="https://github.com/deepseek-ai/DeepSeek-OCR/blob/main/DeepSeek_OCR_paper.pdf"><b>📄 Paper Link</b></a> |
  <a href="https://arxiv.org/abs/2510.18234"><b>📄 Arxiv Paper Link</b></a> |
</p>
<h2>
<p align="center">
  <a href="">DeepSeek-OCR: Contexts Optical Compression</a>
</p>
</h2>
<p align="center">
<img src="assets/fig1.png" style="width: 1000px" align=center>
</p>
<p align="center">
<a href="">Explore the boundaries of visual-text compression.</a>       
</p>

## Usage
Inference using Huggingface transformers on NVIDIA GPUs. Requirements tested on python 3.12.9 + CUDA11.8：

```
torch==2.6.0
transformers==4.46.3
tokenizers==0.20.3
einops
addict 
easydict
pip install flash-attn==2.7.3 --no-build-isolation
```

```python
from transformers import AutoModel, AutoTokenizer
import torch
import os
os.environ["CUDA_VISIBLE_DEVICES"] = '0'
model_name = 'deepseek-ai/DeepSeek-OCR'

tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
model = AutoModel.from_pretrained(model_name, _attn_implementation='flash_attention_2', trust_remote_code=True, use_safetensors=True)
model = model.eval().cuda().to(torch.bfloat16)

# prompt = "<image>\nFree OCR. "
prompt = "<image>\n<|grounding|>Convert the document to markdown. "
image_file = 'your_image.jpg'
output_path = 'your/output/dir'

# infer(self, tokenizer, prompt='', image_file='', output_path = ' ', base_size = 1024, image_size = 640, crop_mode = True, test_compress = False, save_results = False):

# Tiny: base_size = 512, image_size = 512, crop_mode = False
# Small: base_size = 640, image_size = 640, crop_mode = False
# Base: base_size = 1024, image_size = 1024, crop_mode = False
# Large: base_size = 1280, image_size = 1280, crop_mode = False

# Gundam: base_size = 1024, image_size = 640, crop_mode = True

res = model.infer(tokenizer, prompt=prompt, image_file=image_file, output_path = output_path, base_size = 1024, image_size = 640, crop_mode=True, save_results = True, test_compress = True)
```

## vLLM
Refer to [🌟GitHub](https://github.com/deepseek-ai/DeepSeek-OCR/) for guidance on model inference acceleration and PDF processing, etc.<!--  -->

[2025/10/23] 🚀🚀🚀 DeepSeek-OCR is now officially supported in upstream [vLLM](https://docs.vllm.ai/projects/recipes/en/latest/DeepSeek/DeepSeek-OCR.html#installing-vllm).
```shell
uv venv
source .venv/bin/activate
# Until v0.11.1 release, you need to install vLLM from nightly build
uv pip install -U vllm --pre --extra-index-url https://wheels.vllm.ai/nightly
```

```python
from vllm import LLM, SamplingParams
from vllm.model_executor.models.deepseek_ocr import NGramPerReqLogitsProcessor
from PIL import Image

# Create model instance
llm = LLM(
    model="deepseek-ai/DeepSeek-OCR",
    enable_prefix_caching=False,
    mm_processor_cache_gb=0,
    logits_processors=[NGramPerReqLogitsProcessor]
)

# Prepare batched input with your image file
image_1 = Image.open("path/to/your/image_1.png").convert("RGB")
image_2 = Image.open("path/to/your/image_2.png").convert("RGB")
prompt = "<image>\nFree OCR."

model_input = [
    {
        "prompt": prompt,
        "multi_modal_data": {"image": image_1}
    },
    {
        "prompt": prompt,
        "multi_modal_data": {"image": image_2}
    }
]

sampling_param = SamplingParams(
            temperature=0.0,
            max_tokens=8192,
            # ngram logit processor args
            extra_args=dict(
                ngram_size=30,
                window_size=90,
                whitelist_token_ids={128821, 128822},  # whitelist: <td>, </td>
            ),
            skip_special_tokens=False,
        )
# Generate output
model_outputs = llm.generate(model_input, sampling_param)

# Print output
for output in model_outputs:
    print(output.outputs[0].text)
```


## Visualizations
<table>
<tr>
<td><img src="assets/show1.jpg" style="width: 500px"></td>
<td><img src="assets/show2.jpg" style="width: 500px"></td>
</tr>
<tr>
<td><img src="assets/show3.jpg" style="width: 500px"></td>
<td><img src="assets/show4.jpg" style="width: 500px"></td>
</tr>
</table>


## Acknowledgement

We would like to thank [Vary](https://github.com/Ucas-HaoranWei/Vary/), [GOT-OCR2.0](https://github.com/Ucas-HaoranWei/GOT-OCR2.0/), [MinerU](https://github.com/opendatalab/MinerU), [PaddleOCR](https://github.com/PaddlePaddle/PaddleOCR), [OneChart](https://github.com/LingyvKong/OneChart), [Slow Perception](https://github.com/Ucas-HaoranWei/Slow-Perception) for their valuable models and ideas.

We also appreciate the benchmarks: [Fox](https://github.com/ucaslcl/Fox), [OminiDocBench](https://github.com/opendatalab/OmniDocBench).

---

## 🔍 Summary of Changes

### 1. **Formatting & Clean-Up**

* Removed extra spaces, blank lines, and inconsistent indentation.
* Fixed small style issues (like missing spaces in comments).
* Added missing newline at the end of the file.

---

### 2. **Device and Dtype Handling**

* Added automatic device detection:

  ```python
  model_device = next(self.parameters()).device
  ```
* Added adaptive dtype logic:

  ```python
  image_dtype = torch.bfloat16 if model_device.type == "cuda" else torch.float32
  ```
* Replaced all hardcoded `.cuda()` and `.to(torch.bfloat16)` with:

  ```python
  .to(model_device)
  .to(image_dtype)
  ```

✅ **Now works automatically on both GPU and CPU**, without device mismatch errors.

---

### 3. **Autocast and Inference Improvements**

* Wrapped generation in a conditional autocast block:

  ```python
  use_autocast = model_device.type == "cuda"
  if use_autocast:
      with torch.autocast("cuda", dtype=torch.bfloat16):
          with torch.no_grad():
              ...
  else:
      with torch.no_grad():
          ...
  ```
* Reduces memory usage and speeds up inference on GPU.
* Added `torch.no_grad()` for safer evaluation (no gradient tracking).

---

### 4. **Image Preprocessing**

* All image tensors now use:

  ```python
  .to(image_dtype)
  ```

  instead of hardcoded `torch.bfloat16`.
* Improves flexibility and prevents dtype errors when running on CPU.

---

### 5. **Generation Parameter Updates**

* Adjusted text generation settings:

  ```python
  do_sample = False
  num_beams = 1
  max_new_tokens = 4096  # was 8192
  min_new_tokens = 1
  repetition_penalty = 1.2
  pad_token_id = tokenizer.pad_token_id or tokenizer.eos_token_id
  ```

🧠 Results: Faster, more controlled generation; avoids repetitive or runaway outputs.

---

### 6. **Safer Decoding**

* Cleaned up decoding logic:

  ```python
  input_length = input_ids.unsqueeze(0).to(model_device).shape[1]
  outputs = tokenizer.decode(output_ids[0, input_length:])
  ```

✅ Avoids CUDA-specific assumptions, consistent across devices.

---

### 7. **Miscellaneous**

* Added helpful comments for clarity.
* Improved readability around image transformation and saving results.
* Added extra blank lines for cleaner structure.

---

## ⚙️ Overall Impact

| Category        | Before                | After                                    |
| --------------- | --------------------- | ---------------------------------------- |
| Device handling | Hardcoded `.cuda()`   | Auto-detected and flexible               |
| Dtype           | Always bfloat16       | Adaptive: bfloat16 (GPU) / float32 (CPU) |
| Inference       | Could crash on CPU    | Runs safely everywhere                   |
| Generation      | Unbounded, repetitive | Tuned and stable                         |
| Readability     | Mixed formatting      | Clean and consistent                     |

---

## Citation
```bibtex
@article{wei2025deepseek,
  title={DeepSeek-OCR: Contexts Optical Compression},
  author={Wei, Haoran and Sun, Yaofeng and Li, Yukun},
  journal={arXiv preprint arXiv:2510.18234},
  year={2025}
}