slow training (learning) of new nnUNetv2

[mehrnia]

I was using nnUNetv2 version that I downloaded in April 2024. it was converging in iteration less than 100. for the exact same dataset, I'm using nnUNetv2 which I downloaded in August 2024, now it goes to 1000 to converge. and the dice score is same, what is the reason for this slow performance? I see the learning rate decreases where it may not be. and how to make it faster. I'm afraid to increase the leaning rate then to sabotage the model.