change backbone to dla102
Closed this issue · 2 comments
xinzaishi commented
Hi, [pengsida]
I want to change the backbone from dla34 to dla102. I meet the weights and input shape mismatch. To solve this, I add a conv kernel just like you define self.envolve_fuse = nn.Conv1d(128, 64, 1). And pust in the evolve_poly func. the training process seems like ok. Yet I am not sure is it ok?
pengsida commented
It is ok.
xiannsa commented
@xinzaishi Hello! I encountered the issue you described when I was trying to change "dla-34" to "dla60". Do you know how to resolve this error?