upsample top index xxx out of range

[alexgkendall]

upsample top index xxx out of range

[Jinming-Su]

@jorgyz Thank you too. The later person who would come across can comment out line 127 and line 128.

[jorgyz]

@SpecialM2580 Having the same problem here...were you able to figure it out? Thanks

[jaykay2000]

@SpecialM2580
Hi, I ran into the same problem, have you guys figured out how to fix this problem? The funny thing is that by changing the net.prototxt the index "number" changes...

[jorgyz]

Interestingly, if you comment out that check in the code and recompile, it runs just fine and the results seem to look as expected.

[SpecialM2580]

I'm sorry，I almost forget this issue…It seems that this problem only occurs in CPU mode…I try running it in K40，it works just fine…I choose the batch size =3 (reach the maximum memory)and change the classification type to 32 ,using the Camvid dataset,after 15000 iteration ,and it didn't converge...the result is not good-A-,i'm new in deep learning ,may be my parameter chioce is bad,(batch size)or the iteration is not enough...

[khufkens]

@SpecialM2580 How was this issue resolved I can't make heads nor tails from this message chain.

[codecolony]

@jorgyz Thanks, your solution seems to work!

[5argon]

@jorgyz Thank you too, for those who came later and want to do that please note that there are 2 spots in the code that is throwing that fatal exception. You have to comment both out. And this seems to be a bug, since it is unlikely that just by switching from GPU to CPU will cause layer shapes to change...