yolo task=detect mode=train imgsz=640 data=CustomCharacterPath.yaml epochs=500 batch=30 device=cpu name=UniversalOne
#Missing last parts after following 100% guide. Cant train images getting 'TORCH_USE_CUDA_DSA'
stuck lion
Im running a 3060ti just for giggles to test but ill run that line and see if same error happens. Ill post update later.
stuck lion
okay update:
i no longer have the 'TORCH_USE_CUDA_DSA' error
i checked my labels and instead of having a 0 it had position of 1 after being generated by makesense, explains why it wasnt working. its training on my cutom images now. Thank you for support 👍
stuck lion
okay so i test with 5 images and it worked butr i went up to 100 images and i get this error
IndexError: index 1 is out of bounds for dimension 1 with size 1
foggy scaffold
IndexError: index 1 is out of bounds for dimension 1 with size 1
Some of your images are corrupted or you are messing labels from the images and your distribution of images is a mess
Rewatch the tutorial, I've repeated this issue to a million human already
stuck lion
Okay i was able to fix the issue.
foggy scaffold
Okay i was able to fix the issue.
Did you redo the model?
stuck lion
Yes it can do image training now but im still missing onnx 1.12.0
I followed the whole tutorial vidoe etc step by step and as i watching the requirements install i dont see onnx as a installed requirement. When i do open the requirements.txt i see its listed as an install packet.
This is litetally the last missing part for .pt to .onnx conversion
Im just missing onnx 1.12.0 now new ver. Is 1.15.0
Ive installed both and still shows missing onnx when doing conversion
foggy scaffold
Try installing what's in #❓・aimmy-faq
stuck lion
i ran install tutuorial on my 2nd PC and it worked im able to train anad convert on my other PC so we are good. Thx for help DOC.