python - theano CPU running out of memory: what is wrong? -


i run simple network theano on server , got out-of-memory error, not sure reason. asking because unlikely because using memory.

here reasons:

first, according post, when running gpu result in problems caused no support of virtual memory, running cpu, should fine.

second, build network first layer matrix 100k 10, , second layer 10 1, it's 1m numbers model. far, tried 1000 data points together, if machine load data together, , initialize layers together, there should @ 110m float numbers. used float32, on 64bit machine. according post, each number takes 60bytes @ most. so, whole initialization takes 6gb memory. if there variate different resources take memory, don't understand why cannot run on 128gb ram server.

can suggest should into?

just in case asks code, here is.

what size minibatches? need remember activations take space in memory too.


Comments

Popular posts from this blog

amazon web services - S3 Pre-signed POST validate file type? -

c# - Check Keyboard Input Winforms -