Describe your question/issue here! (delete this when you post)
- Project Type: YOLO26 Instance Segmentation (Small)
- Operating System & Browser: chome, window
- Project Universe Link or Workspace/Project ID: yamato-dnvf7/aluminium-ingot-pallet
- Do you grant Roboflow Support permission to access your Workspace for troubleshooting? (Yes/No): yes
Hello Roboflow Support Team,
I am contacting you regarding an ongoing YOLO26 Instance Segmentation (Small) training job.
Initially, Roboflow estimated approximately 5 hours of training time and around 10 credits. However, the training has now been running for over 12 hours, and the estimated remaining time is still showing “3 days, 9 hours remaining.”
Additionally, the training has already consumed 28.14 credits, which is significantly higher than the original estimate.
My dataset contains 2,924 images after augmentation, with preprocessing set to Resize: Fit Within 640×640.
Given the relatively small dataset size and image resolution, I am concerned that this training job is taking an unusually long time. I suspect there may be an issue with the training server or infrastructure.
I would appreciate clarification on the following questions:
-
Is this training duration normal for a YOLO26 Small segmentation model with 2,924 images?
-
Could this delay be caused by an issue on Roboflow’s servers?
-
Should I stop the current training job and request a refund of the consumed credits before restarting?
-
Or should I leave the training running for the estimated additional 3 days, even though the credit usage continues to increase?
I am unsure whether continuing this training job is the right decision, especially given the unexpected training time and credit consumption.
Could you please investigate this issue and advise me on how to proceed? I would also appreciate your consideration of a credit refund if the problem is related to the training infrastructure.
Thank you for your assistance.
Best regards