AttributeError: Can't get attribute 'Hardswish' about yolo_tracking HOT 6 CLOSED

I resolved the issue and ran the script successfully in Windows. Initially I was trying to run the script in PyCharm. But this time I created an environment in Anaconda Powershell Prompt using: conda create -n python*3.7 deepsort. Then open the requirements.txt and remove the lines torch>=1.6.0 and torchvision>=0.7.0 then go to pytorch and select Build, OS, Package, Language etc. Then copy the line after Run this Command: and paste it into your command line. At the time of writing this comment I got this line pip install torch===1.7.0 torchvision===0.8.1 torchaudio===0.7.0 -f https://download.pytorch.org/whl/torch_stable.html. Then follow the instructions in the repository for installing the rest of the requirements and obtaining the weights and checkpoint files and then run the track.py script.

from yolo_tracking.

Check that you fulfill all the requirements. I seems to me that you pytorch is not >=1.7

from yolo_tracking.

I installed the latest version even tried the nightly build of pytorch. I tried the same in Linux and it worked though but in Windows I have yet to make it work. Are you running this on Windows?

from yolo_tracking.

This repo has never been run on a windows machine. But this issue may help ultralytics/yolov5#773

from yolo_tracking.

I've seen that and that's not a real solution. Telling me to use Google Colab or Kaggle Notebook or Docker because it's "Tested" is an excuse worst than the crime itself. Anyways it looks like it's a pytorch problem since the same code works in Linux. I've heard somewhere that a workaround would be to replace the Hardswish with LeakyReLU and retrain the network but that's a huge undertaking. Anyways thanks for the response.

from yolo_tracking.

Yolov5 uses CI for testing its basic functionality: "including training (train.py), testing (test.py), inference (detect.py) and export (export.py) on MacOS, Windows, and Ubuntu"

Using a verified environment like a DockerContainer to get stuff working is a software development standard. Specially if it was already tested and recommended by its creator. I would say it is even more important if you have problems getting the code running.

from yolo_tracking.