I have trained a classification model using yolo11n , now i have to write its inference script , but the problem is for development and deployment I have python version 3.6, but yolo and ultralytics supports python version 3.8 and above only , I have tried to convert the classification model i.e .pt file into torchscript to remove this dependency issue , but converting into torchscript is lowering the accuracy of the model , is there any way , or any source I can refer to resolve this issue? My requirement is to use .pt file directly to infer my model on python version 3.6 without loading model using ultralytics. I tried different approches Converting the classification .pt file into torchscript but is is lowering the accuracy. I tries rebuilding the architecture for yolo11 and then extarcting weights Continue reading...
