unitable
unitable copied to clipboard
UniTable: Towards a Unified Table Foundation Model
I'm parsing some pdfs with openparse which uses unitable. This error occured and I have no idea what it even means , I would appreciate any insights _image_to_tensor(image.crop(bbox), size=(112, 448))...
``` %load_ext autoreload %autoreload 2 from IPython.display import display, HTML from typing import Tuple, List, Sequence, Optional, Union from pathlib import Path import re import torch import tokenizers as tk...
During HTML training, somethings the loss occasionally becomes inf, causing the total loss to remain inf after accumulation. Why doesn't gradient clipping take effect, and what is causing this issue?
The model I obtained through fine-tuning, which script should I use to load it for inference?