Idefics2-OCR

Fine-tuned the HuggingFaceM4/idefics2-8b model on the nielsr/docvqa_1200_examples_donut dataset for document VQA pairs. Checkout Idefics2-OCR on Hugging Face.

Find the rest of training details here.

Finetune

Set your wanb token in the .env file as WANDB_API.

• Install the requirements

pip install -r requirements.txt

• Finetune Idefics

python3 idefics2.py --wandb True

Run the app

You don't have to finetune for running the app the model is loaded from Hugging Face.

python3 app.py