Model Training Terminating Prematurely (Early Stopping Issue)

🤝 Community Help
1

I am reaching out because I have tried to train my model three separate times now, and each time the training terminates prematurely around epoch 30, despite me setting the limit to 300 epochs.

Because of this issue, I have exhausted all of my training tokens on three incomplete, clearly undertrained models.

  • Project ID / Link: slum-wnqjt

  • The Issue: The training stops at ~39 epochs. The model is still heavily undertrained at this point, but it seems an aggressive early stopping mechanism is triggering anyway.

  • The Impact: I have used up my tokens on three failed attempts and currently cannot proceed with my project.

Could you please look into why the training loop is cutting off so early for my dataset?

Thank you for your help, and I look forward to your guidance on how to prevent this on the next run.

image
image1392×387 39.3 KB

  • Project Type:Instance Segmantation
  • Operating System & Browser: Windows 11, Google Chrome Version 148.0.7778.168 (64-bit)
  • **Project Universe Link or Workspace/Project ID:**slum-wnqjt
  • Do you grant Roboflow Support permission to access your Workspace for troubleshooting? (Yes/No):Yes