You can deploy this model and start earning money today!

powow/
nougat-swe

Name: nougat-swe
Availability: InStock

transformers

safetensors

vision-encoder-decoder

image-to-text

endpoints_compatible

Views

No views yet

Nougat for Swedish

Fine-tuned version of facebook/nougat-small. Primarily trained on mathematical expressions and Swedish text.

Related code at: https://github.com/will-berg/nougat-swe

Inference

Code for inference on a pdf is in the GitHub repo, but can be done quickly on a provided image (of a pdf page) using the following code:

python

from PIL import Image
from transformers import NougatProcessor, VisionEncoderDecoderModel
import torch

processor = NougatProcessor.from_pretrained("powow/nougat-swe")
model = VisionEncoderDecoderModel.from_pretrained("powow/nougat-swe")
device = "cuda" if torch.cuda.is_available() else "cpu"

model.to(device)

image = Image.open(image_path)
pixel_values = processor(image, return_tensors="pt").pixel_values

outputs = model.generate(
  pixel_values.to(device),
  min_length=1,
  max_new_tokens=3584,
  bad_words_ids=[[processor.tokenizer.unk_token_id]],
)

sequence = processor.batch_decode(outputs, skip_special_tokens=True)[0]
sequence = processor.post_process_generation(sequence, fix_markdown=False)

print(sequence)