When train my custom data, how to load pretrained model? It says not match with default weights.

[h030162]

🚀 Feature

Motivation

Pitch

Alternatives

Additional context

[github-actions]

Hello @h030162, thank you for your interest in our work! Please visit our Custom Training Tutorial to get started, and see our Jupyter Notebook , Docker Image, and Google Cloud Quickstart Guide for example environments.

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 model or data training question, please note that Ultralytics does not provide free personal support. As a leader in vision ML and AI, we do offer professional consulting, from simple expert advice up to delivery of fully customized, end-to-end production solutions for our clients, such as:

• Cloud-based AI systems operating on hundreds of HD video streams in realtime.
• Edge AI integrated into custom iOS and Android apps for realtime 30 FPS video inference.
• Custom data training, hyperparameter evolution, and model exportation to any destination.

For more information please visit https://www.ultralytics.com.

[glenn-jocher]

@h030162 can you supply code to reproduce your error please?

[glenn-jocher]

@h030162 I just tested this on our colab notebook, and everything works correctly. You can load up an 80 class model with pretrained weights if you are training on coco, or you can change the class count of your model and dataset as well. The only requirement is that the dataset and model class counts must match each other:

Dataset:

yolov5/data/coco128.yaml

Lines 14 to 15 in ad71d2d

	# number of classes
	nc: 80

Model:

yolov5/models/yolov5s.yaml

Lines 1 to 2 in ad71d2d

	# parameters
	nc: 80 # number of classes

If you change nc, everything still works correctly. In this case the model will be composed of pretrained weights except for the output layers, which are no longer the same shape as the pretrained output layers. The output layers will remain initialized by random weights.

[h030162]

./yolov5_checkpoint/yolov5_models/yolov5x.pt is not compatible with ./models/yolov5x.yaml. Specify --weights '' or specify a --cfg compatible with ./yolov5_checkpoint/yolov5_models/yolov5x.pt.

I just change yolov5x.yaml nc from 80 to 2. and change yaml file in data .nc from 80 to2. names:["dog", "cat"]

[glenn-jocher]

@h030162 I tested this again (a second time) in colab, and everything is still working fine. I updated yolov5x.yaml and coco128.yaml to 160 classes (duplicated the names to reach 160 names), and started training with the following command, all is correct with no problems.

python train.py --img 640 --batch 16 --epochs 5 --data ./data/coco128.yaml --cfg ./models/yolov5x.yaml --weights yolov5x.pt

See https://docs.ultralytics.com/yolov5/tutorials/train_custom_data

[ou525]

@glenn-jocher I also have this problem

[glenn-jocher]

@ou525 if this is a bug report, please provide screenshots and minimum viable code to reproduce your issue, otherwise we can not help you.

[github-actions]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.