lyc6749
lyc6749
I use coco to pretrain a model with seg and track, but the loss of cost_volume is `nan.```
when I used the convert model for test, run with `editany_test.py`,I meet the problem 
> when I used the convert model for test, run with `editany_test.py`,I meet the problem  maybe only support for sd1.5?
I changed the code of `editany_test.py` , but the error is still appeared. 
