The Roboflow Model has converged, but it is still training and gives the same time estimate as the start of training

Describe your question/issue here! (delete this when you post)

  • Project Type: Object Detection
  • Operating System & Browser: Windows and Chrome
  • Project Universe Link or Workspace/Project ID: The project id is coralclassify


Above shows my current dashboard. As you can see, the model accuracy has largely converged and is seeing no improvements. However, there are still 6 hours left. For reference, model training was estimated to take 7 hours, and that was around 6 hours ago. I would ideally like to simply use the model as-is and stop training given the convergence. I am also concerned that training will never terminate given how the time estimate has been extended already.

This topic was automatically closed 21 days after the last reply. New replies are no longer allowed.