Hey i have an issues il my training i tried to use the model roboflow and RTDETR but il both the training stopped after ~30 epochs. but i want 100 epochs how to do this ?because the graphes does not finish to converge.
Hi @Maizane_ALI_AHAMADA!
I am sorry you’re running into this issue. I wanted to let you know that I am looking into it, and will find a solution.
To help me further troubleshoot, do you providing the project id and a version number? Additionally, do you mind recreating the issue, and providing screenshots with your Dev console displayed, any DevTools error messages, etc.?
Here is a guide on how to open your dev tools: Finding your browser’s developer console
salut merci pour votre reponse l identifiant du projet c’est : under-hook-detection. numero de version de quoi du dataset? car environ tous mes entreinement sur roboflow s’arrete à environ une trentaine d’epoch les courbes n’ont pas le temps de converger. je pense c’est regler dans kes hyperparametre mais nous on a pas acces. je veux que l entrainement fasses au moins 100 epoch . En vous remerciant d’avance
Hi @Maizane_ALI_AHAMADA!
Thank you for your patience, I am going to loop in the appropriate team to troubleshoot this further. I will keep you updated.
Best Regards,
Ford
Hi,
Thank You for your help, yes keep me updated.
Best Regards.
Hi @Maizane_ALI_AHAMADA!
Thank you for your patience! I talked with the team and we have logic in place that stops training if the model is no longer improving to conserve your credits. To improve the training process, you can either add more data to your dataset, or if you are willing to provide your training graphs, I’d be happy to troubleshoot further.
Please let me know if I can help in any other way!
Best Regards,
Ford
Hey,
Thank you for your respons. i can’t force the training into 100 epoch.
I really want it.
You will find here the matrice confusion and the graphe.
Best regards,
I increase the data but it still stoppped and why befor it was not stopp and i had less data than now.
Hi @Maizane_ALI_AHAMADA!
I have looped in a member of our models team to investigate this further. I will keep you updated on the progress.
Thank you again for your continued patience.
Best Regards,
Ford