loss_diff decrease to zero very fast

Question

loss_diff decrease to zero very fast

rshaojimmy opened this issue 4 years ago · 2 comments

rshaojimmy commented 4 years ago

Thanks for your nice codes!

During my training, the loss of differences decrease to 0 within the first 100 steps and remain 0 afterwards.

May I ask do you also encounter such case?

Thanks.

Answer 1 · 2021-05-24T05:56:02.000Z

Hi, guys, have your solved it ? I meet same situation like this... T T
loss_dif and loss_simse are always keep 0 from epoch 0 to 99,it made me so confused :(

Answer 2 · 2021-09-12T14:07:39.000Z

I also encountered the same problem, and found that private code learn nothing, they are all zeros....， i didn't directly use the original experimental datasets, but i think it doesn't matter too much....