inference for NAN

Question

inference for NAN

Opened this issue 3 years ago · 3 comments

runzeer commented 3 years ago

why the model forward generate the nan?

Answer 1 · 2022-03-28T00:10:23.000Z

I never encountered nan issue. What's your input

Answer 2 · 2024-05-22T13:56:52.000Z

I have encountered the same problem. Have you resolved it now

Answer 3 · 2024-05-23T03:22:04.000Z

I have solved this problem. My problem is that I used my own mask creator, which has a probability of generating all 0 masks, causing the path to calculate 0 and ultimately resulting in the generator parameter being nan.
I hope it will be helpful to you.