How to resume/continue from already trained YOLO5 Training Epoch?

Question

I have started the Yolo5 Training with custom data

The command I have used:

!python train.py --img 640 --batch-size 32 --epochs 5 --data /content/drive/MyDrive/yolov5_dataset/dataset_Trafic/data.yaml --cfg /content/drive/MyDrive/yolov5/models/yolov5s.yaml  --name Model

Training started as below & completed:

For resuming/continue for more epoch I have below command

!python train.py --img 640 --batch-size 32 --epochs 6 --data /content/drive/MyDrive/yolov5_dataset/dataset_Trafic/data.yaml --weights /content/drive/MyDrive/yolov5/runs/train/Model/weights/best.pt --cache --exist-ok

But still the Training start from the scratch. How to continue from previous epoch.

Also I tried with resume command

!python train.py --epochs 10 --resume

but I am getting below error message

When you specify a weight in command, it initialize network weighs with specified weight file. So it does not start from scratch. Only the number of epochs starts from zero. — Mohammad Khoshbin, Apr 17 '22 at 21:33
Thanks.But if we compare the time taken for each epoch , there is no much difference between these two methods/command — thangaraj1980, Apr 19 '22 at 01:54
Time taken for each epoch depends on your data and your model. So if you don't change them, duration of epoch won't change — Mohammad Khoshbin, Apr 19 '22 at 07:28
Found few solution from video : https://www.youtube.com/watch?v=c9WS5jga0Lg — thangaraj1980, Aug 11 '22 at 04:22

score 2 · Answer 1 · edited Dec 26 '22 at 09:01

2

Open runs/train/Model/opt.yaml file and change epochs no. try again

!python train.py  --weights /content/drive/MyDrive/yolov5/runs/train/Model/weights/last.pt  --resume

edited Dec 26 '22 at 09:01

David Arenburg

91,361
17
137
196

answered Jul 09 '22 at 14:41

Ashish1KD

21
4

1

The provided solution is not clear for readers. Please, clarify it. – Hamed Baziyad Jul 10 '22 at 13:09
2

Your answer could be improved with additional supporting information. Please [edit] to add further details, such as citations or documentation, so that others can confirm that your answer is correct. You can find more information on how to write good answers [in the help center](/help/how-to-answer). – Community Jul 11 '22 at 13:28
1

This works for YOLOv7 too – David Arenburg Dec 26 '22 at 09:15

Tanishq Sharma · Answer 2 · 2022-10-08T16:47:27.883

Try this: In the runs folder(where your previous weights are stored), find a file with name "args.yaml" inside the file, augment the epochs count by the no. of epochs you want to train your model now. For example: If previously your model was trained for 100 epochs then, the args.yaml file would have an attribute: "epochs: 100" Lets say you was to continue training for 50 more epochs then change epochs to 150 i.e. "epochs: 150" save the file and then execute training command.

How to resume/continue from already trained YOLO5 Training Epoch?

2 Answers2