search

DS4SD / docling

🥚 Transform PDF to JSON or Markdown with ease and speed 🐣
MIT License
454 stars 46 forks source link

Not work well on windows #66

Open cuikai-ai opened 1 week ago

cuikai-ai commented 1 week ago

1725933869049

To fix this you could try to:

  1. loosen the range of package versions you've specified
  2. remove package versions to allow pip to attempt to solve the dependency conflict

ERROR: ResolutionImpossible: for help visit https://pip.pypa.io/en/latest/topics/dependency-resolution/#dealing-with-dependency-conflicts

cuikai-ai commented 1 week ago

atfer install deepsearch-glm==0.2.3

PeterStaar-IBM commented 1 week ago

@cuikai-ai Yes, our next (dev) step is to support windows. We need the latest deepsearch-glm, but we are on it. We would really appreciate it if you could give us feedback, to verify all works as expected!

sdspieg commented 6 days ago

I works fine on Windows under wsl for me... It also uses my GPU (albeit not very efficiently yet)...
image I guess I'll have to see if I can optimize the Batch Size and maybe also play with concurrent.futures. Any recommendations for that?

maxmnemonic commented 1 day ago

I works fine on Windows under wsl for me... It also uses my GPU (albeit not very efficiently yet)... I guess I'll have to see if I can optimize the Batch Size and maybe also play with concurrent.futures. Any recommendations for that?

Underlying models use GPU, and currently there is a big benefit of using GPU, albeit not as big as it could be, because of relatively low utilization. We are looking into ways of improving utilization, but it’s not very straight forward.

You can play with general batching and concurrency settings here: settings.py to get better use of your CPU and RAM, but I doubt you could load GPU more efficiently this way, please let us know if this would influence your GPU utilization as well.