acc on unet for image segmentation is not rising when is trained on fp16

Question

i am doing some image segmentation whit deep learning, when i do it on fp32 i get val acc about 97% but when i try to train on fp16 i get stuck on 21% i\'ve changed the lear