the perfect layers to choose to unfreeze for efficientnetv2

I am working on creating a classification model, and I chose tf_efficientnetv2_b0 from timm library in pytorch with my images but I don’t know if the image size 224 is suitable for this model, also I want to ask about what is the perfect layers to choose to unfreeze for efficientnetv2

I don’t find any good source about the image size issue in the Internet, but about the unfreezing, I print the modules of the model and find this: 1-conv_stem 2-bn1 3-blocks 4-conv_head 5-bn2 6-global_pool 7-classifier I chose only for fine-tuning and unfreezing these layers (3,4,7) since the bn is needed to be frozen and layer 1 is very initial and contains global features from Imagenet, what I want to ask is if my implementation is correct.