get bug of memory leaks

[zuotianyoumeng]

hi I get some memory leaks when I run this code, I try to add sess.graph.finalize() to frozen the graph, but it still exists, Can anyone give some advice? thanks.

[zuotianyoumeng]

my cudnn version is 6.0 , cuda is 8.0

[kil77]

hello! I meet the same problem with you and it confused me several days, now i found it`s because in the function extract_batch ,too many data are read into the memory ,I decreased 512 to 128 and 4096 to 256,and the problem solved! hope that can help you

[HJanefer]

hello! I meet the same problem with you and it confused me several days, now i found it`s because in the function extract_batch ,too many data are read into the memory ,I decreased 512 to 128 and 4096 to 256,and the problem solved! hope that can help you

您好，我发现我改了之后还是内存爆了，如果方便的话，能给些建议吗？万分感激。

[kil77]

hello! I meet the same problem with you and it confused me several days, now i found it`s because in the function extract_batch ,too many data are read into the memory ,I decreased 512 to 128 and 4096 to 256,and the problem solved! hope that can help you

您好，我发现我改了之后还是内存爆了，如果方便的话，能给些建议吗？万分感激。

你可以分步调试，就会找到出问题的那条语句

[HJanefer]

hello! I meet the same problem with you and it confused me several days, now i found it`s because in the function extract_batch ,too many data are read into the memory ,I decreased 512 to 128 and 4096 to 256,and the problem solved! hope that can help you

您好，我发现我改了之后还是内存爆了，如果方便的话，能给些建议吗？万分感激。

你可以分步调试，就会找到出问题的那条语句

好滴，太谢谢您了！不过我还是没把代码跑通，我把那两个参数的值改得更小了，调试时sess.run loss就卡住了，没有显示报错，也没计算结果，并且也没有跳出运算，不知道是哪里出了问题。

[fuyi02]

this is the official version of deeplab: https://github.com/tensorflow/models/tree/master/research/deeplab you‘d best change the code.