search

VikParuchuri / marker

Convert PDF to markdown quickly with high accuracy
https://www.datalab.to
GNU General Public License v3.0
14.65k stars 763 forks source link

Illegal hardware instruction #99

Closed amuricys closed 2 months ago

amuricys commented 3 months ago

Might have to do with #97, no idea. I had to downgrade my Python version and install Ray globally, then get into the shell and try converting a single file. Other than that I just followed the README. I'm getting the following:

(marker-pdf-py3.12) ➜ marker git:(master) ✗ python3 convert_single.py reservoir-tikhonov.pdf reservoir-tikhonov.md Loaded texify model to mps with torch.float16 dtype [1] 48315 illegal hardware instruction python3 convert_single.py

Got this with another PDF too.

amuricys commented 3 months ago

After fiddling with brew installations for a while, it seems that one of the dependencies (ocrmypdf) depends on python 3.12, which is not supported by ray. Dependency hell I guess

brew install ocrmypdf libmagic tesseract tesseract-lang
==> Downloading https://ghcr.io/v2/homebrew/core/ocrmypdf/manifests/16.1.2
##################################################################### 100.0%
==> Fetching dependencies for ocrmypdf: python@3.12, glib, cairo, harfbuzz, libraqm, pillow, img2pdf, pango, tesseract, libass, ffmpeg and unpaper
halghasra commented 3 months ago

I'm also getting this issue and it is doing my head in!! I've been fiddling with brew for hours and days and I'm still getting the same error every single time! @VikParuchuri any insights on this issue?

zsh: illegal hardware instruction python convert_single.py docfiles/destination/output.md

VikParuchuri commented 2 months ago

This will be fixed in the new version (to be merged shortly). I removed ocrmypdf as a required dependency, and bumped the ray version.