maxenceliu
maxenceliu
hallo? Really wondering...
oh ,got it. mxnet is different from numpy in function random. numpy use different random seed durning each run however mxnet use same seed....
but still interesting, since during each train ,it use different m, this constraint won't change the final direction however
2-3secs on GTX1080
after the newest commit, total_loss explode after 350 iterations due to the rpn_cls_loss exploded.
result is not stable, this time, regular_loss become Nan ater 500 iters...
Yes, But I don't know which commit correct this problem... @CharlesShang would you explain this for us? Just for study. FYI, MobileNets is simple amazing!
In FILE darknet.py Change Line 283 for i, start in enumerate(range(0, len(keys), 5)): for i, start in enumerate(range(0, len(keys), 6)):
@sunbinbin1991 Hi, Have you resolved the problem. I think i met the exact the same problem as yours. And I use the mxnet-scala in Java.
> I did not understand this code, can anyone help me understand this code?  Have you ever make clear what does this equation mean? I can't understand this either....