ultralytics / ultralytics

Ultralytics YOLO11 🚀
https://docs.ultralytics.com
GNU Affero General Public License v3.0
32.48k stars 6.25k forks source link

about resume train with yolo11 #17492

Open huhuhu5798 opened 1 day ago

huhuhu5798 commented 1 day ago

Search before asking

Question

At first, I set my epochs to 200 and completed the training, but I want to change my epochs to 500 based on the finished training. What should I do?

Additional

No response

UltralyticsAssistant commented 1 day ago

👋 Hello @huhuhu5798, thank you for your interest in Ultralytics 🚀! We're excited to help you with your training query. While an Ultralytics engineer will assist you soon, here's some guidance to get started:

If your question is related to resuming training with YOLO11, we suggest referring to the Model Training Tips for best practices and strategies.

To resume training your model from where you left off, typically, you can adjust your training configuration to set a higher epoch count and continue the training with the saved model weights. Please ensure you have all your previous training parameters saved to maintain consistency.

In the meantime, if this doesn't solve your issue, or you encounter any difficulties, please provide more details, such as the training configuration used and any logs that could help troubleshoot the issue further.

For community support, join us on Discord 🎧 for real-time chat, explore our Discourse for in-depth discussions, or visit our Subreddit to share with the community.

Upgrade

Make sure you're using the latest version of ultralytics with:

pip install -U ultralytics

Environments

YOLO can be run in verified environments. For a smooth experience, we recommend:

Status

Keep an eye on our Ultralytics CI to ensure all tests pass successfully across various platforms and conditions.

We're here to help, so don't hesitate if you have more questions! 😊

Y-T-G commented 1 day ago

You can start training from last.pt.

https://www.reddit.com/r/Ultralytics/comments/1eolwl8/the_correct_way_to_train_from_a_previously/

But it's not "resuming", it's a new training starting from your finished model.

purples123 commented 18 hours ago

I meet the same problem. I set the resume=True, and model.load('/yolov11/runs/train/exp3/weights/last.pt') ,but ite just start training from 1.Is there any method to start training from last training epoch?

Y-T-G commented 18 hours ago

No. If the training is finished, you can't resume. You can only restart.