When I run detect.py, I got a prediction of nan.

[Ramsden1]

❔Question

I run the detect.py by

python detect.py --classes 2 --weights best.py --device 0

The code has no error, but the output image of this code has no bounding box annotated.
Then I tried to print this pred,

pred = model(img, augment=opt.augment)[0]

I find that the output tensor has a lot of nan values and some float numbers. But when I change device to CPU, this problem appeared.
So now I can only run it by CPU, can anyone tell me where is the wrong?

Thank you.

[github-actions]

Hello @Ramsden1, thank you for your interest in 🚀 YOLOv5! Please visit our ⭐️ Tutorials to get started, where you can find quickstart guides for simple tasks like Custom Data Training all the way to advanced concepts like Hyperparameter Evolution.

If this is a 🐛 Bug Report, please provide screenshots and minimum viable code to reproduce your issue, otherwise we can not help you.

If this is a custom training ❓ Question, please provide as much information as possible, including dataset images, training logs, screenshots, and a public link to online W&B logging if available.

For business inquiries or professional support requests please visit https://www.ultralytics.com or email Glenn Jocher at glenn.jocher@ultralytics.com.

Requirements

Python 3.8 or later with all requirements.txt dependencies installed, including torch>=1.7. To install run:

$ pip install -r requirements.txt

Environments

YOLOv5 may be run in any of the following up-to-date verified environments (with all dependencies including CUDA/CUDNN, Python and PyTorch preinstalled):

• Google Colab Notebook with free GPU:
• Kaggle Notebook with free GPU: https://www.kaggle.com/ultralytics/yolov5
• Google Cloud Deep Learning VM. See GCP Quickstart Guide
• Docker Image https://hub.docker.com/r/ultralytics/yolov5. See Docker Quickstart Guide

Status

If this badge is green, all YOLOv5 GitHub Actions Continuous Integration (CI) tests are currently passing. CI tests verify correct operation of YOLOv5 training (train.py), testing (test.py), inference (detect.py) and export (export.py) on MacOS, Windows, and Ubuntu every 24 hours and on every commit.

[wudashuo]

Yes, I also found this problem. You can find more details in #1625.
Are you using Windows + Pytorch1.7? If yes:
A bad solution is set cudnn.benchmark = True, this will bring another problem, the results are pretty disastrous, according @glenn-jocher said.
My solution is install pytorch1.6.0+CU101 by pip install torch==1.6.0+cu101 torchvision==0.7.0+cu101 -f https://download.pytorch.org/whl/torch_stable.html, and there's no need to downgrade your CUDA if you are using CUDA10.2 or CUDA11.x, it works fine with my env(CUDA11.1, Python3.8, Pytorch1.6.0+CU101)

[Ramsden1]

Yes, I also found this problem. You can find more details in #1625.
Are you using Windows + Pytorch1.7? If yes:
A bad solution is set cudnn.benchmark = True, this will bring another problem, the results are pretty disastrous, according @glenn-jocher said.
My solution is install pytorch1.6.0+CU101 by pip install torch==1.6.0+cu101 torchvision==0.7.0+cu101 -f https://download.pytorch.org/whl/torch_stable.html, and there's no need to downgrade your CUDA if you are using CUDA10.2 or CUDA11.x, it works fine with my env(CUDA11.1, Python3.8, Pytorch1.6.0+CU101)

Yes, My env is Windows + pytorch1.7 + cuda11.
I change the version of pytorch as you said, and it works for me.
Thank you!