Placeholder content - original file not found in upstream repository.
3.11 Restarting an Interrupted YOLO26 Training Session¶
{: #lesson-restarting-yolo}
Sometimes, long training runs can be interrupted due to runtime disconnects, power outages, or manual stops. Fortunately, YOLO26 automatically saves your training progress after every epoch to the runs/detect/train/weights/last.pt file.
You can easily resume an interrupted training session right where it left off.
How to Resume Training¶
When you initialize your YOLO26 model, instead of loading the base model (e.g., yolo26n.pt), you simply load your last saved weights. Then, you call the train() method with the resume=True argument.
Python Code Snippet¶
from ultralytics import YOLO
# 1. Load the last saved checkpoint from your runs directory
# (Make sure the path matches your specific run folder)
model = YOLO("runs/detect/train/weights/last.pt")
# 2. Resume training
# YOLO26 automatically remembers your epochs, batch size, and dataset from the checkpoint
results = model.train(resume=True)Once executed, YOLO26 will seamlessly restore the optimizer state, learning rates, and current epoch, and continue training until the original epoch limit is reached.