why different optimizer train get different result

[tank1530532]

Search before asking

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

Question

pertrain weights: yolov5s.pt
dataset: coco-2017 with below classes
classes:[0:person,1:car,2:motorcycle,3:bus,4:truck]
optimizer: SGD-redline Adam-blueline

I train yolov5 model with 2 different optimizes and I got two different results . Why this two results lookes like this?

Additional

[UltralyticsAssistant]

👋 Hello @tank1530532, thank you for your interest in YOLOv5 🚀! This is an automated response to help guide you, and an Ultralytics engineer will also assist soon.

For differences in training results using various optimizers like SGD and Adam, this is quite common due to the differing optimization strategies and hyperparameters inherent to each optimizer.

If this is a 🐛 Bug Report, please provide a minimum reproducible example to help us diagnose the issue effectively.

For training-related inquiries, ensure your environment meets the following:

1. Python version 3.8.0 or higher
2. All dependencies installed via requirements.txt
3. Check if you have CUDA and PyTorch correctly configured, especially when leveraging GPU for training.

You may want to try running on different environments or using verified setups such as Google Colab or Kaggle Notebooks, which preinstall necessary prerequisites.

Moreover, you might find our Training Tips guide useful in troubleshooting and optimizing your training results.

Lastly, we encourage you to explore our latest model YOLOv8, which may offer improved performance and features for your tasks.

Looking forward to assisting you further! 😊

[pderrenger]

@tank1530532 different optimizers can lead to varying results due to their distinct approaches to adjusting learning rates and handling gradients, which can affect convergence and model performance.