About AutoBatch

[z1069614715]

Search before asking

• I have searched the YOLOv5 issues and discussions and found no similar questions.

Question

Hi! I want to use your autobatch function in my code, but have something problem, Could you please take a look at it for me?

as shown above, You can see my gpu memory is not increasing, but this autobatch function in working order in yolov5 code.
Why?
Wait patiently for your answer!

Additional

No response

[github-actions]

👋 Hello @z1069614715, 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://ultralytics.com or email support@ultralytics.com.

Requirements

Python>=3.7.0 with all requirements.txt installed including PyTorch>=1.7. To get started:

git clone https://github.com/ultralytics/yolov5  # clone
cd yolov5
pip install -r requirements.txt  # install

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 and Kaggle notebooks with free GPU:
• Google Cloud Deep Learning VM. See GCP Quickstart Guide
• Amazon Deep Learning AMI. See AWS Quickstart Guide
• Docker Image. 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), validation (val.py), inference (detect.py) and export (export.py) on macOS, Windows, and Ubuntu every 24 hours and on every commit.

[robotaiguy]

Why does it say "Using batch-size 253731...", but it looks like it suggested batch-size of 16.

[glenn-jocher]

@z1069614715 can you confirm you are seeing this issue in master?

AutoBatch is experimental and may not work in all circumstances, so if you are getting unreliable results you should set your batch-size manually.

[glenn-jocher]

@z1069614715 I've pushed PR #9209 to protect from super high batch size predictions, but the reality is that CUDA memory utilisation is sometimes hard to reproduce and measure correctly, especially if previous trainings are running, have run or been terminated early.

[z1069614715]

@z1069614715 I've pushed PR #9209 to protect from super high batch size predictions, but the reality is that CUDA memory utilisation is sometimes hard to reproduce and measure correctly, especially if previous trainings are running, have run or been terminated early.

Thanks for your help, I have solved the problem, The reason is that (torch.backends.cudnn.benchmark) I set it to True, if True AutoBatch will lose efficacy.

[glenn-jocher]

@z1069614715 got it. Should we set benchmark(False) before running AutoBatch? Benchmark(True) during training should only occur on non-zero seeds it seems, though that doesn't make much sense. I've created PR #9259 to update this.