File size: 2,909 Bytes
0be4d34 2a70c3d a435fd4 2a70c3d 60d03d8 a435fd4 60d03d8 fea9c34 60d03d8 00043aa 60d03d8 2a70c3d a435fd4 2a70c3d 335e715 2a70c3d 335e715 2a70c3d 335e715 2a70c3d 335e715 70fbef0 335e715 90e0293 335e715 70fbef0 335e715 31aa84f 2a70c3d |
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 |
---
license: apache-2.0
---
# OFA-tiny
## Introduction
This is the **tiny** version of OFA pretrained model. OFA is a unified multimodal pretrained model that unifies modalities (i.e., cross-modality, vision, language) and tasks (e.g., image generation, visual grounding, image captioning, image classification, text generation, etc.) to a simple sequence-to-sequence learning framework.
The directory includes 4 files, namely `config.json` which consists of model configuration, `vocab.json` and `merge.txt` for our OFA tokenizer, and lastly `pytorch_model.bin` which consists of model weights. There is no need to worry about the mismatch between Fairseq and transformers, since we have addressed the issue yet.
## How to use
To use it in transformers, please refer to https://github.com/OFA-Sys/OFA/tree/feature/add_transformers. Install the transformers and download the models as shown below.
```bash
git clone --single-branch --branch feature/add_transformers https://github.com/OFA-Sys/OFA.git
pip install OFA/transformers/
git clone https://huggingface.co/OFA-Sys/OFA-tiny
```
After, refer the path to OFA-tiny to `ckpt_dir`, and prepare an image for the testing example below. Also, ensure that you have pillow and torchvision in your environment.
```python
>>> from PIL import Image
>>> from torchvision import transforms
>>> from transformers import OFATokenizer, OFAModel
>>> from generate import sequence_generator
>>> mean, std = [0.5, 0.5, 0.5], [0.5, 0.5, 0.5]
>>> resolution = 256
>>> patch_resize_transform = transforms.Compose([
lambda image: image.convert("RGB"),
transforms.Resize((resolution, resolution), interpolation=Image.BICUBIC),
transforms.ToTensor(),
transforms.Normalize(mean=mean, std=std)
])
>>> tokenizer = OFATokenizer.from_pretrained(ckpt_dir)
>>> txt = " what does the image describe?"
>>> inputs = tokenizer([txt], return_tensors="pt").input_ids
>>> img = Image.open(path_to_image)
>>> patch_img = patch_resize_transform(img).unsqueeze(0)
# using the generator of fairseq version
>>> model = OFAModel.from_pretrained(ckpt_dir, use_cache=True)
>>> generator = sequence_generator.SequenceGenerator(
tokenizer=tokenizer,
beam_size=5,
max_len_b=16,
min_len=0,
no_repeat_ngram_size=3,
)
>>> data = {}
>>> data["net_input"] = {"input_ids": inputs, 'patch_images': patch_img, 'patch_masks':torch.tensor([True])}
>>> gen_output = generator.generate([model], data)
>>> gen = [gen_output[i][0]["tokens"] for i in range(len(gen_output))]
# using the generator of huggingface version
>>> model = OFAModel.from_pretrained(ckpt_dir, use_cache=False)
>>> gen = model.generate(inputs, patch_images=patch_img, num_beams=5, no_repeat_ngram_size=3)
>>> print(tokenizer.batch_decode(gen, skip_special_tokens=True))
```
|