search

loolzaaa / faster-rcnn-pytorch

A PyTorch implementation of Faster R-CNN
MIT License
17 stars 5 forks source link

import _C #9

Closed edehino closed 1 year ago

edehino commented 2 years ago

I saw that in your ipynb file, you have import _C. May I know what it is? I am ready familiar with it

loolzaaa commented 2 years ago

When you run python setup.py develop, it uses Extension package from PyTorch for compile and future use some utils (NMS, ROI_Align, ROI_Pool, etc.) written on C++. This line on code of setup.py.

All C++ sources you can find in ./lib/csrc/

I am ready familiar with it

May be you need to update some python dependencies from notebook, i used it long time ago.

edehino commented 2 years ago

Thank you for that!

Also, I am trying to replicate your installation: !pip install torch==1.4.0+cu100 torchvision==0.5.0+cu100 -f https://download.pytorch.org/whl/torch_stable.html

But when I try to, it prompts me that the earliest I could get it torch==1.7.1+cu101

Would this still work?

On Wed, May 4, 2022 at 2:43 PM Andrey Korsakov @.***> wrote:

When you run python setup.py develop, it uses Extension package from PyTorch for compile and future use some utils (NMS, ROI_Align, ROI_Pool, etc.) written on C++. This line on code https://github.com/loolzaaa/faster-rcnn-pytorch/blob/110e9e36704da378990ab80b2b3382345ca5d030/lib/setup.py#L30 of setup.py.

All C++ sources you can find in ./lib/csrc/

I am ready familiar with it

May be you need to update some python dependencies from notebook, i used it long time ago.

— Reply to this email directly, view it on GitHub https://github.com/loolzaaa/faster-rcnn-pytorch/issues/9#issuecomment-1116977967, or unsubscribe https://github.com/notifications/unsubscribe-auth/APN24GCWA7KSWI7WKKPEXWLVIIL73ANCNFSM5VAMOB4Q . You are receiving this because you authored the thread.Message ID: @.***>

-- The information contained in the email (any attachment and response string) is strictly confidential and for the use of the addressee only, unless otherwise indicated. If you are not the intended recipient, please do not read, copy, use or disclose to others this message or any attachment. Please also notify the sender by replying to this email and then delete the email and any copies of it. Any views or opinions expressed within the email are solely those of the sender and do not necessarily represent those of the University of the Philippines.

loolzaaa commented 2 years ago

There is no problem with Cuda 10.1

I ran this notebook when PyTorch 1.5.0 was latest release. Previously, version 1.5.0 could work without changes, but I can't say the same about versions 1.5.0+

The PyTorch team sometimes makes breaking changes to their framework. If you decide to use version 1.7.1, then you may have to make changes to the neural network code.

Feel free to pull requests for fixes to support the new version.