---
license: mit
language:
- zh
tags:
- gpt2
- vit
---
# 模型介绍

![](https://ankur3107.github.io/assets/images/vision-encoder-decoder.png)


1. vit对图像做encoder，然后再用gpt2做decoder
2. vit模型使用的是`google/vit-base-patch16-224`, gpt2使用的是`yuanzhoulvpi/gpt2_chinese`
3. 本模型支持中文


# 训练代码

[https://github.com/yuanzhoulvpi2017/zero_nlp/tree/main/vit-gpt2-image-chinese-captioning](https://github.com/yuanzhoulvpi2017/zero_nlp/tree/main/vit-gpt2-image-chinese-captioning)






# 推理代码
# infer

```python
from transformers import (VisionEncoderDecoderModel, 
                          AutoTokenizer,ViTImageProcessor)
import torch
from PIL import Image

```


```python
vision_encoder_decoder_model_name_or_path = "yuanzhoulvpi/vit-gpt2-image-chinese-captioning"#"vit-gpt2-image-chinese-captioning/checkpoint-3200"

processor = ViTImageProcessor.from_pretrained(vision_encoder_decoder_model_name_or_path)
tokenizer = AutoTokenizer.from_pretrained(vision_encoder_decoder_model_name_or_path)
model = VisionEncoderDecoderModel.from_pretrained(vision_encoder_decoder_model_name_or_path)
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model.to(device)
```


```python
max_length = 16
num_beams = 4
gen_kwargs = {"max_length": max_length, "num_beams": num_beams}


def predict_step(image_paths):
    images = []
    for image_path in image_paths:
        i_image = Image.open(image_path)
        if i_image.mode != "RGB":
            i_image = i_image.convert(mode="RGB")

        images.append(i_image)

    pixel_values = processor(images=images, return_tensors="pt").pixel_values
    pixel_values = pixel_values.to(device)

    output_ids = model.generate(pixel_values, **gen_kwargs)

    preds = tokenizer.batch_decode(output_ids, skip_special_tokens=True)
    preds = [pred.strip() for pred in preds]
    return preds


predict_step(['bigdata/image_data/train-1000200.jpg'])

```


# 效果
## example 1
![](images/images1.png)
## example 2
![](images/images2.png)