Which parameter will be better for Stanford Dogs？

[vba34520]

First of all, thank you for your work!

Train on the Stanford Dogs Dataset

python train.py --init-lr 0 --progressive-resizing 224 224 --dataset-dir dataset/dogs --epochs 10000 --batch-size 32 --weights imagenet

Early stop at epoch 88 with the best_weight_EfficientNetB0_val_loss:0.018_val_acc:0.822_epoch:68.h5

Which parameter should I choose to improve the performance?

Looking forward to your reply !