Sergiy

> You're welcome! ☺️ Please let us know if you face any more problems. I'm closing this issue I am still getting errors with `Ultralytics YOLOv8.0.5 🚀 Python-3.8.15 torch-1.10.1` Using...

> @Robotatron can please confirm if this issue is reproducible in the colab example? thanks! Yes, it is reproducible if using `imgsz=600` in Colab, same error and workaround as here...

Thanks for a prompt fix!

@dungviet112 is there any advantage of using `v5loader=True` ?

> we'll release the 1280 model soon 1) Any chance you are planning to release a 1280 segmentation model as well? 2) For the current segmentation models (trained on imgz...

So what is the answer for this?

What was your console command to start training? You are probably providing the path for the model incorrectly

Thanks Ayush. I made two runs with exact same hyper parameters (except for num_epochs) (thanks for integrating ClearML :) ):  But the results are very, very different (same dataset,...

> Yes maps for the test set.. The maps will be different it you set different number of epochs because lr scheduler works on number of epochs.. So you'll only...

> test set doubles the train batch size as it only runs on a single GPU 1. Not sure I follow. When training on several GPUs or a single one,...