Possible solutions to decrease contrastive Loss

[ines321]

Hello, like the official paper has mention , Loss must decrease over epcohs, my contrastive loss does not decrease and still around 5 . I use temperature = 0.1 batch_size = 128 learning_rate = 1e-1 epoch = 500 crop_size = [80, 80] input_size = [224,224]

What are possible solutions to decrease the contrastive loss ?