Problem training 'Dog' model - recommended params?

Can anybody help me by imparting their knowledge on good training parameters for Yolov8?

I’ve trained a few roboflow datasets on dogs (all of them 10,000 images or more) and the datasets seem fine (i have browsed through the images) but the end result in terms of the trained model, the inference is always poor.

Can anybody rcommnd an optimal learning rate or learning rate range with decay and the recommended number of Epochs. I’ve tried a few different params as separate sets of training and also as fine tuning from checkpoints from the previous train, but nothing is yielding good results.

Any feedback or help would be much appreciated.

Many Thanks