running out of memory

tensorflow 1.9 ==>GPU
Jetson TX2 ==> 7.7 memory

Is there a way to train an object detector on this board! do you know a model that doesn’t swallow the Memory.

I tried to minimize the parameters of the config file(batch,queue,num-readers,…) and images(dim) but didn’t affect.

the irony is that the same model is training on my 6 ram 1.5 ghz laptop (tf 1.9 cpu) but it takes ages.