self.init_query_bbox = nn.Embedding(self.num_query, 10) . This parameter doesn't seem to have been updated.

[zbaishancha]

Hello dear author, thank you very much for your open source work and your promptness in answering questions. I was looking at your code and found this problem as follows:

During each forward, the query bbox is using self.init_query_box.weight.clone, and by doing so it seems that the parameters of self.init_query_box will not be updated, so what is the significance of setting self.init_query_box to nn.Embedding?

[zbaishancha]

@afterthat97 Best wishes to you :)

[afterthat97]

Note that clone != detach, so the gradient will still be propagated to init_query_box.weight

[zbaishancha]

Sorry, I may have misunderstood, I thought that after clone, query bbox and self.init_query_box.weight are two different variables (occupying different memory addresses despite having the same value). So, even though thequery bbox is involved in budgeting the gradient in the decoder layer, the gradient doesn't seem to trickle down to init_query_box.weight?

[zbaishancha]

Also, I tried to experiment and found that self.init_query_box.weight doesn't seem to change after each forward and backward.

[zbaishancha]

@afterthat97

[zbaishancha]

sry, my understanding was wrong and I am very sorry for the trouble!